JP4617905B2 - Monitoring device and method, recording medium, and program - Google Patents

Description

  The present invention relates to a monitoring apparatus and method, a recording medium, and a program, and more particularly, to a monitoring apparatus and method, a recording medium, and a program that can more accurately monitor the intrusion of an intruder.

  In recent years, with the increasing awareness of security against the diversification of crimes, various monitoring devices have been introduced into companies, ordinary households, various facilities, and the like. As a method for detecting the intrusion of an intruder into the monitoring area of these monitoring devices, there are a method for binarizing information, a method using pattern recognition, and a method using contour perception.

  The method of binarizing information is to first output data from various imaging devices or sensors that monitor the monitoring area in the monitoring area where the intruder has entered and the state where the intruder has not entered. Get multiple sample values in advance. Next, based on those sample values, a threshold value for dividing the value of the output data from the imaging device or sensor into two groups, the state where the intruder is invading and the state where the intruder is not intruding, is determined. Then, it is a method of determining which group the output data value from the imaging device or sensor falls in based on the determined threshold and detecting the intrusion of the intruder into the monitoring area.

  The method using pattern recognition is to register the feature quantity (for example, shape) of an expected intruder in advance based on the value of output data from an imaging device or sensor that monitors the monitoring area. This is a method for detecting the intrusion of an object that matches a feature quantity that has entered into the monitoring area (see, for example, Patent Document 1).

The method using contour perception is based on the value of output data from an imaging device or sensor that monitors a monitoring area. This is a method for detecting an intruder by determining that the changing portion is an edge portion (contour portion) of the intruder.
Japanese Patent Laid-Open No. 10-285581

  However, in the case of the method of binarizing information, it is difficult to classify into two groups, intrusion state and non-intrusion state, under various variables such as movement of intruder, surrounding environment, noise, As a result, there is a problem that the possibility of missing an intruder or determining that the intruder has entered when there is no intrusion increases. In addition, since it is difficult to classify into two groups, intrusion state and non-intrusion state, for example, a threshold is determined so as to give priority to preventing oversight of intrusion, or false detection of intrusion is prevented. When the threshold value is determined so as to prioritize the situation, the determination of the threshold value is not limited to a user's arbitrary judgment, and there is a problem that the intruder cannot be detected accurately because of lack of objectivity.

  Further, the method using pattern recognition has a problem that it is not possible to detect the intrusion of an intruder having a feature quantity that is not registered in advance. In addition, it is difficult to register feature quantities that take into account various variable factors such as rotation, tilt, overlap, shape change, and ambient brightness change of intruders. However, when trying to detect the intrusion of an intruder more accurately, there is a problem that the data amount of the registered feature amount becomes enormous and the detection processing algorithm becomes complicated.

  Furthermore, in the method using contour perception, it is difficult to distinguish whether the sudden change in the value of output data from the imaging device or sensor is caused by the contour of the intruder or noise. was there. In order to separate the abrupt change in the output data value caused by the contour of the intruder from the abrupt change in the output data value caused by the noise with higher accuracy and to remove the noise, a detection process is performed. There was a problem that the algorithm of became complicated.

  The present invention has been made in view of such a situation, and makes it possible to more easily and accurately detect an intrusion of an object.

Monitoring apparatus of the present invention, is output from the sensors monitoring the monitoring area, based on the monitoring data composed of the corresponding unit data to each of the sensing units obtained by dividing the monitored area into a plurality of regions, to the monitoring area An average value of first unit data, which is unit data constituting first monitoring data, which is monitoring data output from a sensor in a monitoring device that monitors intrusion of objects, in a state where no object has entered the monitoring area And standard deviation, or first difference unit data of first difference monitoring data composed of first difference unit data which is a difference at predetermined time intervals of first unit data corresponding to the same detection unit The average value and standard deviation of the image are calculated, and no image has entered the monitoring area, and a predetermined pattern of video or illumination is projected onto the monitoring area. In the state, the first correlation matrix indicating the correlation between the second unit data that is the unit data constituting the second monitoring data that is the monitoring data output from the sensor, or the second corresponding to the same detection unit Criteria for calculating a second correlation matrix indicating the correlation between the second difference unit data of the second difference monitoring data constituted by the second difference unit data which is a difference at predetermined time intervals of the unit data of When monitoring the space calculation means and the intrusion of the object into the monitoring area, the first space between the first space constituted by the first monitoring data and the third monitoring data which is the monitoring data output from the sensor. The Mahalanobis distance of 1 is calculated using the average value and standard deviation of the first unit data and the first correlation matrix, or the same as the second space configured by the first difference monitoring data Between the third difference monitoring data constituted by the third difference unit data that is the difference for each time interval of the third unit data that is the unit data constituting the third monitoring data corresponding to the knowledge unit. Mahalanobis distance calculating means for calculating the second Mahalanobis distance using the average value and standard deviation of the first difference unit data and the second correlation matrix, and the first Mahalanobis distance or the second Mahalanobis distance And determining means for determining intrusion of the object into the monitoring area based on the distance.

  The standardized unit data obtained by standardizing the third unit data using the average value and standard deviation of the first unit data corresponding to the same detection area as the third unit data, or the third difference unit data, Based on the standardized difference unit data standardized using the average value and the standard deviation of the first difference unit data corresponding to the same detection area as the difference unit data 3, there is a high possibility that something has entered the monitoring area 1 The apparatus may further include a detection unit that detects a change region having at least two predetermined sizes, and the Mahalanobis distance calculation unit may calculate the first Mahalanobis distance or the second Mahalanobis distance in the change region.

  The reference space calculation means includes the second unit data between the second unit data corresponding to the detection units in the monitoring area and in the frame-shaped area in contact with the outer periphery of the monitoring area among the correlation coefficients constituting the correlation matrix. And calculating the first Mahalanobis distance or the second Mahalanobis distance in the frame-like region while calculating only the correlation coefficient between the difference unit data of the other and the other correlation coefficients as 0. Can be.

  The reference space calculation means includes a second unit data or second difference unit data corresponding to the detection unit and a predetermined centering on the detection unit for each detection unit among the correlation coefficients constituting the correlation matrix. Only the correlation coefficient between the second unit data or the second difference unit data corresponding to the detection unit included in the size area is calculated, and the other correlation coefficients are set to 0. be able to.

Monitoring method of the present invention, is output from the sensors monitoring the monitoring area, based on the monitoring data composed of the corresponding unit data to each of the sensing units obtained by dividing the monitored area into a plurality of regions, to the monitoring area In a monitoring method of a monitoring device that monitors intrusion of objects, first unit data that is unit data constituting first monitoring data that is monitoring data output from a sensor in a state in which no object has entered the monitoring area Of the first difference monitoring data constituted by the first difference unit data that is a difference at predetermined time intervals of the first unit data corresponding to the same detection unit The average value and standard deviation of the difference unit data are calculated, and there is no object invading the monitoring area, and the image or illumination of the predetermined pattern is the monitoring area Corresponds to the first correlation matrix indicating the correlation between the second unit data that is the unit data constituting the second monitoring data that is the monitoring data output from the sensor or the same detection unit in the projected state A second correlation matrix indicating a correlation between the second difference unit data of the second difference monitoring data configured by the second difference unit data which is a difference at predetermined time intervals of the second unit data to be A reference space calculation step to calculate, a first space configured by first monitoring data when monitoring intrusion into the monitoring area, and third monitoring data that is monitoring data output from the sensor A first Mahalanobis distance between them is calculated using the average value and standard deviation of the first unit data and the first correlation matrix, or the second Mahalanobis distance configured by the first difference monitoring data The third difference monitoring data configured by the third difference unit data that is the difference for each time interval between the third unit data that is the unit data constituting the third monitoring data corresponding to the same detection unit as the space A Mahalanobis distance calculating step of calculating a second Mahalanobis distance between the first Mahalanobis distance using the average value and standard deviation of the first difference unit data and the second correlation matrix; And a determination step of determining intrusion of the object into the monitoring area based on the Mahalanobis distance of 2.

The program recorded on the recording medium of the present invention is based on monitoring data that is output from a sensor that monitors the monitoring area and that is composed of unit data corresponding to each of the detection units obtained by dividing the monitoring area into a plurality of areas. The first monitoring data, which is monitoring data output from the sensor in a state in which nothing is intruding into the monitoring area, is a program for monitoring processing of the monitoring device that monitors the intrusion of the object into the monitoring area. It is composed of the average value and standard deviation of the first unit data, which is the unit data to be configured, or the first difference unit data which is the difference at predetermined time intervals of the first unit data corresponding to the same detection unit. Calculating the average value and the standard deviation of the first difference unit data of the first difference monitoring data, and the state in which nothing has entered the monitoring area, and A first correlation indicating a correlation between second unit data that is unit data constituting second monitoring data that is monitoring data output from the sensor in a state in which a video or illumination of a predetermined pattern is projected onto the monitoring region. Or the second difference unit data of the second difference monitoring data constituted by the second difference unit data which is the difference at predetermined time intervals of the second unit data corresponding to the same detection unit. A reference space calculation step for calculating a second correlation matrix indicating a correlation between the first space, a first space constituted by the first monitoring data when monitoring the intrusion of the object into the monitoring area, and output from the sensor Calculating the first Mahalanobis distance between the monitoring data and the third monitoring data using the average value and standard deviation of the first unit data and the first correlation matrix, or The third difference, which is the difference for each time interval between the second space constituted by one difference monitoring data and the third unit data that is the unit data constituting the third monitoring data corresponding to the same detection unit Mahalanobis distance calculation which calculates the 2nd Mahalanobis distance between the 3rd difference monitoring data comprised by unit data using the average value and standard deviation of 1st difference unit data, and a 2nd correlation matrix And a determination step of determining whether an object has entered the monitoring area based on the first Mahalanobis distance or the second Mahalanobis distance.

Program of the present invention, is output from the sensors monitoring the monitoring area, based on the monitoring data composed of unit data corresponding to each of the detection units obtained by dividing the monitored area into a plurality of regions, ones of the monitoring area First monitoring data, which is monitoring data output from a sensor in a state in which a monitoring processing computer of a monitoring device that monitors intrusion performs monitoring processing, and nothing has entered the monitoring area 1st unit data which is unit data constituting the average value and standard deviation, or first difference unit data which is a difference at predetermined time intervals of the first unit data corresponding to the same detection unit The average value and the standard deviation of the first difference unit data of the first difference monitoring data to be calculated are calculated, and no state has entered the monitoring area The correlation between the second unit data, which is the unit data constituting the second monitoring data, which is the monitoring data output from the sensor, in the state where the image or the illumination of the predetermined pattern is projected on the monitoring region. Second correlation monitoring data configured by second difference unit data that is a difference at predetermined time intervals of the first correlation matrix or second unit data corresponding to the same detection unit. A reference space calculating step for calculating a second correlation matrix indicating a correlation between the difference unit data, a first space constituted by the first monitoring data when monitoring the entry of the object into the monitoring area, and a sensor Whether the first Mahalanobis distance from the third monitoring data, which is the monitoring data to be output, is calculated using the average value and standard deviation of the first unit data, and the first correlation matrix Alternatively, the second space which is the difference between the second space constituted by the first difference monitoring data and the third unit data which is the unit data constituting the third monitoring data corresponding to the same detection unit. 2nd Mahalanobis distance between 3rd difference monitoring data comprised by 3 difference unit data is calculated using the average value and standard deviation of 1st difference unit data, and a 2nd correlation matrix The method includes a Mahalanobis distance calculation step and a determination step of determining intrusion of the object into the monitoring area based on the first Mahalanobis distance or the second Mahalanobis distance.

In the monitoring apparatus, the monitoring method, the recording medium, and the program of the present invention, the monitoring data is composed of unit data corresponding to each detection unit that is output from a sensor that monitors the monitoring area and divides the monitoring area into a plurality of areas. In the monitoring of the intrusion of the object into the monitoring area based on the monitoring data, the first unit data constituting the first monitoring data that is the monitoring data output from the sensor when no object has entered the monitoring area 1st difference monitoring data comprised by the average value and standard deviation of 1 unit data, or the 1st difference unit data which is the difference for every predetermined time interval of the 1st unit data corresponding to the same detection unit The average value and the standard deviation of the first difference unit data are calculated, a state in which no object has entered the monitoring area, and an image or a predetermined pattern A first correlation matrix indicating a correlation between second unit data that is unit data constituting second monitoring data that is monitoring data output from the sensor in a state in which the illumination is projected onto the monitoring region; or The second indicating the correlation between the second difference unit data of the second difference monitoring data constituted by the second difference unit data that is the difference at predetermined time intervals of the second unit data corresponding to the same detection unit. When the correlation matrix of 2 is calculated and the intrusion of the object into the monitoring area is monitored, the first space constituted by the first monitoring data and the third monitoring data which is the monitoring data output from the sensor A first Mahalanobis distance between the average value and the standard deviation of the first unit data and the first correlation matrix, or a second differential data configured by the first difference monitoring data. Sky And third difference monitoring data constituted by third difference unit data which is a difference for each time interval of third unit data which is unit data constituting third monitoring data corresponding to the same detection unit, and The second Mahalanobis distance between the first Mahalanobis distance and the second Mahalanobis distance is calculated using the average value and standard deviation of the first difference unit data and the second correlation matrix. Based on this, the intrusion of the object into the monitoring area is determined.

  According to the present invention, it is possible to detect an intrusion of an object. Further, according to the present invention, it is possible to detect an intrusion of an object more easily and accurately.

  Embodiments of the present invention will be described below. Correspondences between constituent elements described in the claims and specific examples in the embodiments of the present invention are exemplified as follows. This description is to confirm that specific examples supporting the invention described in the claims are described in the embodiments of the invention. Therefore, even if there are specific examples that are described in the embodiment of the invention but are not described here as corresponding to the configuration requirements, the specific examples are not included in the configuration. It does not mean that it does not correspond to a requirement. On the contrary, even if a specific example is described here as corresponding to a configuration requirement, this means that the specific example does not correspond to a configuration requirement other than the configuration requirement. not.

  Further, this description does not mean that all the inventions corresponding to the specific examples described in the embodiments of the invention are described in the claims. In other words, this description is an invention corresponding to the specific example described in the embodiment of the invention, and the existence of an invention not described in the claims of this application, that is, in the future, a divisional application will be made. Nor does it deny the existence of an invention added by amendment.

The monitoring device according to claim 1 (for example, the intrusion detection device 12 in FIG. 1) outputs the monitoring region output from a sensor (for example, the imaging device 11 in FIG. 1) that monitors the monitoring region to a plurality of regions. Based on monitoring data (for example, frame information) composed of unit data (for example, unit data) corresponding to each of the divided detection units (for example, the detection unit 21 in FIG. 2), the intrusion into the monitoring area The first monitoring data (for example, non-intrusive frame information) that is the monitoring data output from the sensor in a state in which the device does not enter the monitoring area. Average value and standard deviation of first unit data (for example, non-intrusive unit data) that is the unit data, or the first unit data corresponding to the same detection unit The first difference unit data of the first difference monitoring data (for example, non-intrusion difference frame information) configured by first difference unit data (for example, non-intrusion difference data) that is a difference at regular time intervals. The average value and the standard deviation are calculated, and the image or illumination in a state in which the object does not enter the monitoring area and a predetermined pattern (for example, the pattern shown in FIGS. 6 to 9) is displayed in the monitoring area. The second unit data (for example, pattern unit data) that is the unit data constituting the second monitoring data (for example, pattern frame information) that is the monitoring data output from the sensor in the state projected onto A first correlation matrix (for example, a unit data correlation matrix) indicating a correlation between the second unit data corresponding to the same detection unit The correlation between the second difference unit data of the second difference monitoring data (for example, pattern difference frame information) constituted by the second difference unit data (for example, pattern difference data) that is a difference for each time interval is obtained. When a reference space calculation unit (for example, a reference space data calculation unit 72 in FIG. 4) for calculating a second correlation matrix (for example, a difference data correlation matrix) to be shown and monitoring the entry of the object into the monitoring area, Between the 1st space (for example, unit data standard space) constituted by the 1st monitoring data, and the 3rd monitoring data (for example, monitoring frame information) which is the monitoring data outputted from the sensor A first Mahalanobis distance is calculated using an average value and standard deviation of the first unit data and the first correlation matrix, or the first difference monitoring data is calculated. Third unit data (for example, monitoring data) that is the unit data constituting the third monitoring data corresponding to the same detection unit as the second space (for example, difference data reference space) configured by data Second difference between the third difference monitoring data (for example, monitoring difference frame information) constituted by the third difference unit data (for example, monitoring difference data) which is a difference for each time interval of the unit data). Mahalanobis distance calculating means (for example, the distance calculating unit 75 in FIG. 4) for calculating the Mahalanobis distance using the average value and standard deviation of the first difference unit data and the second correlation matrix; Determination means (for example, determination unit 76 in FIG. 4) for determining the intrusion of the object into the monitoring area based on the second Mahalanobis distance or the second Mahalanobis distance. It is characterized in.

3. The monitoring device according to claim 2 (for example, the intrusion detection device 12 in FIG. 1), wherein the third unit data is the first unit data corresponding to the same detection area as the third unit data. Standardized unit data standardized using an average value and standard deviation of the first difference unit data, or an average of the first difference unit data corresponding to the same detection area as the third difference unit data. Based on the standardized difference unit data standardized using the value and the standard deviation, one or more predetermined change areas (for example, the change areas of FIG. 103), for example, a change area detection unit 74) of FIG. 4, wherein the Mahalanobis distance calculation means includes the first Mahalanobis distance in the change area. Other and calculates the second Mahalanobis distance.

In the monitoring device according to claim 3 (for example, the intrusion detection device 12 in FIG. 1), the reference space calculation unit is in the monitoring region among the correlation coefficients constituting the correlation matrix, and Only the correlation coefficient between the second unit data or the second difference unit data corresponding to the detection unit in the frame-like region (for example, the region 201 in FIG. 22) in contact with the outer periphery of the monitoring region is calculated. In addition, another correlation coefficient is set to 0, and the Mahalanobis distance calculating means calculates the first Mahalanobis distance or the second Mahalanobis distance in the frame-shaped region.

The monitoring method according to claim 5 is a detection unit (for example, the detection in FIG. 2) that is output from a sensor (for example, the imaging device 11 in FIG. 1) that monitors the monitoring area and that divides the monitoring area into a plurality of areas. A monitoring device (for example, in FIG. 1) that monitors the entry of the object into the monitoring area based on monitoring data (for example, image information) composed of unit data (for example, unit data) corresponding to each of the units 21). Intrusion detection device 12) monitoring method, wherein first monitoring data (for example, non-intrusive frame information) which is the monitoring data output from the sensor in a state where the object does not enter the monitoring area The average value and standard deviation of the first unit data (for example, non-intrusive unit data) that is the unit data constituting the first unit data, or the first unit data corresponding to the same detection unit The first difference of the first difference monitoring data (for example, non-intrusion difference frame information) configured by first difference unit data (for example, non-intrusion difference data) that is a difference at predetermined time intervals of the data The average value and standard deviation of the unit data are calculated, and the image or illumination in a state where the object does not enter the monitoring area and a predetermined pattern (for example, the patterns shown in FIGS. 6 to 9) Second unit data (for example, pattern unit) that constitutes the second monitoring data (for example, pattern frame information) that is the monitoring data output from the sensor in a state projected on the monitoring area. First correlation matrix (for example, unit data correlation matrix) indicating the correlation between the data) or the second unit data corresponding to the same detection unit Between the second difference unit data of the second difference monitoring data (for example, pattern difference frame information) constituted by the second difference unit data (for example, pattern difference data) that is a difference at every predetermined time interval When a reference space calculation step (for example, steps S3 and S7 in FIG. 5) for calculating a second correlation matrix (for example, a difference data correlation matrix) indicating the correlation, and monitoring the intrusion of the one into the monitoring area, Between the 1st space (for example, unit data standard space) constituted by the 1st monitoring data, and the 3rd monitoring data (for example, monitoring frame information) which is the monitoring data outputted from the sensor The first Mahalanobis distance is calculated using the average value and standard deviation of the first unit data and the first correlation matrix, or the first difference Third unit data (for example, the unit data constituting the third monitoring data corresponding to the same detection unit as the second space (for example, the difference data reference space) configured by the minute monitoring data (for example, Second difference with third difference monitoring data (for example, monitoring difference frame information) constituted by third difference unit data (for example, monitoring difference data) that is a difference at every time interval of (monitoring unit data). A Mahalanobis distance calculating step (for example, step S35 in FIG. 10) for calculating the Mahalanobis distance using the average value and standard deviation of the first difference unit data and the second correlation matrix; A determination step (for example, FIG. 10) for determining whether the object has entered the monitoring area based on the Mahalanobis distance or the second Mahalanobis distance. Step S36) and characterized in that it comprises a.

The program recorded in the recording medium according to claim 6 (for example, the removable medium 51 in FIG. 3) is output from a sensor (for example, the imaging device 11 in FIG. 1) that monitors the monitoring area. Based on monitoring data (for example, image information) composed of unit data (for example, unit data) corresponding to each of the detection units (for example, the detection unit 21 in FIG. 2) divided into a plurality of regions. This is a program for monitoring processing of a monitoring device (for example, the intrusion detection device 12 in FIG. 1) that monitors the intrusion of an object into the area, and is output from the sensor in a state where the object does not enter the monitoring area. The first unit data (for example, non-intrusion unit) that is the unit data constituting the first monitoring data (for example, non-intrusion frame information) that is the monitoring data. Data) average value and standard deviation, or first difference unit data (for example, non-intrusive difference data) that is a difference at predetermined time intervals of the first unit data corresponding to the same detection unit. A state in which the average value and standard deviation of the first difference unit data of the configured first difference monitoring data (for example, non-intrusion difference frame information) is calculated, and the monitoring object is not invaded. In addition, in a state where an image or illumination of a predetermined pattern (for example, patterns shown in FIGS. 6 to 9) is projected onto the monitoring area, second monitoring data (the monitoring data output from the sensor) For example, a first correlation matrix (e.g., a correlation between second unit data (for example, pattern unit data) that is the unit data constituting the pattern frame information) Unit data correlation matrix), or second difference unit data (for example, pattern difference data) that is a difference for each predetermined time interval of the second unit data corresponding to the same detection unit. A reference space calculating step (for example, calculating a second correlation matrix (for example, a difference data correlation matrix) indicating a correlation between the second difference unit data of second difference monitoring data (for example, pattern difference frame information) (for example, Steps S3 and S7 in FIG. 5 and when monitoring the intrusion of the object into the monitoring area, a first space (for example, a unit data reference space) constituted by the first monitoring data, and the sensor The first Mahalanobis distance to the third monitoring data (for example, monitoring frame information), which is the monitoring data to be output, is the average of the first unit data The same detection unit as a second space (for example, a difference data reference space) that is calculated using a value and a standard deviation, and the first correlation matrix, or configured by the first difference monitoring data The third difference unit data (for example, monitoring difference data) that is the difference for each time interval of the third unit data (for example, monitoring unit data) that is the unit data constituting the third monitoring data corresponding to ), The second Mahalanobis distance between the third difference monitoring data (for example, monitoring difference frame information), the average value and the standard deviation of the first difference unit data, and the second correlation. A Mahalanobis distance calculating step (for example, step S35 in FIG. 10) calculated using a matrix, and the first Mahalanobis distance or the second Mahalanobis distance. Based on the determination step the ones from entering the monitoring area (e.g., step S36 of FIG. 10), characterized in that it comprises a.

The program according to claim 7 is a detection unit (for example, a detection unit in FIG. 2) that is output from a sensor (for example, the imaging device 11 in FIG. 1) that monitors a monitoring area and that divides the monitoring area into a plurality of areas. 21) a monitoring device (for example, the intrusion in FIG. 1) that monitors the intrusion of the object into the monitoring area based on the monitoring data (for example, image information) composed of unit data (for example, unit data) corresponding to each of First monitoring data (for example, the monitoring data output from the sensor in a state in which the computer of the detection device 12) performs a monitoring process and the object does not enter the monitoring area. , The average value and standard deviation of the first unit data (for example, non-intrusive unit data) that is the unit data constituting the non-intrusive frame information), or the same First difference monitoring data (for example, no difference) constituted by first difference unit data (for example, non-intrusion difference data) that is a difference at predetermined time intervals of the first unit data corresponding to the detection unit. The average value and the standard deviation of the first difference unit data of the intrusion difference frame information) are calculated, the state in which the object does not enter the monitoring area, and a predetermined pattern (for example, FIG. 6 to FIG. 9). The unit data constituting the second monitoring data (for example, pattern frame information) that is the monitoring data output from the sensor in the state where the image or illumination of the pattern shown in FIG. A first correlation matrix (for example, unit data correlation matrix) indicating a correlation between certain second unit data (for example, pattern unit data), or the same Second difference monitoring data (for example, pattern difference frame) constituted by second difference unit data (for example, pattern difference data) that is a difference at predetermined time intervals of the second unit data corresponding to the knowledge unit. A reference space calculating step (for example, steps S3 and S7 in FIG. 5) for calculating a second correlation matrix (for example, a difference data correlation matrix) indicating a correlation between the second difference unit data of information), and the monitoring When monitoring the intrusion of the object into the area, a first space constituted by the first monitoring data (for example, a unit data reference space) and a third monitoring that is the monitoring data output from the sensor A first Mahalanobis distance between the data (for example, monitoring frame information), an average value and a standard deviation of the first unit data, and the first correlation matrix Or the second space configured by the first difference monitoring data (for example, the difference data reference space) and the third monitoring data corresponding to the same detection unit Third difference monitoring data (for example, monitoring difference data) configured by third difference unit data (for example, monitoring difference data) that is a difference for each time interval of third unit data (for example, monitoring unit data) that is unit data. , A Mahalanobis distance calculating step (for example, calculating a second Mahalanobis distance with respect to the monitoring difference frame information) using the average value and standard deviation of the first difference unit data and the second correlation matrix (for example, Based on the step S35) of FIG. 10 and the first Mahalanobis distance or the second Mahalanobis distance, the entry of the object into the monitoring area Determining step (e.g., step S36 of FIG. 10), characterized in that it comprises a.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

  FIG. 1 is a diagram showing an embodiment of a monitoring system 1 to which the present invention is applied. The monitoring system 1 is configured to include an imaging device 11 and an intrusion detection device 12.

  The imaging device 11 is, for example, a camera using a CCD (Charge Coupled Device) or a CMOS (Complementary Metal Oxide Semiconductor) as an imaging element. The imaging device 11 is installed at a predetermined position, and photographs the situation in the monitoring area (viewing angle) as needed. The imaging device 11 outputs image information of the captured image of the monitoring area to the intrusion detection device 12.

  For example, as illustrated in FIG. 2, the imaging apparatus 11 includes k detection units 21-1 to 21 -k (n rows × m columns) obtained by dividing a monitoring region into regions of a predetermined size (where k = n Xm) is output to the intrusion detection device 12 as image information composed of image data of the image. For example, the detection units 21-1 to 21-k are regions corresponding to one pixel of an image captured by the imaging device 11. Hereinafter, the detection units 21-1 to 21-k are simply referred to as detection units 21 when it is not necessary to individually distinguish them. Hereinafter, the image data corresponding to each detection unit 21 constituting the image information is also referred to as unit data.

  As will be described later with reference to FIGS. 5 and 13, the intrusion detection device 12 detects the intrusion of an intruder such as a person, an animal, or an object into the monitoring area based on image information input from the imaging device 11. To do.

  FIG. 3 is a block diagram illustrating a functional configuration example of the intrusion detection device 12 of FIG. The intrusion detection device 12 includes a CPU (Central Processing Unit) 31, a ROM (Read Only Memory) 32, a RAM (Random Access Memory) 33, an input unit 34, an output unit 35, a recording unit 36, a communication unit 37, and a drive 38. Configured to include. The CPU 31, ROM 32, and RAM 33 are connected to each other via an internal bus 41, and the input unit 34, output unit 35, recording unit 36, communication unit 37, and drive 38 are mutually connected via an input / output interface 42. It is connected. In addition, the imaging device 11 is connected to the input / output interface 42. Further, the internal bus 41 and the input / output interface 42 are connected to each other.

  The CPU 31 receives processing commands and data input by the user using the input unit 34 via the input / output interface 42 and the internal bus 41, and is stored in the ROM 32 based on the input processing commands. Various processes are executed in accordance with a program stored in the storage unit 36 or a program loaded into the RAM 33 from the recording unit 36.

  The ROM 32 stores basically fixed data among programs used by the CPU 31 and calculation parameters.

  The RAM 33 stores programs used in the execution of the CPU 31, parameters and data that change as appropriate during the execution.

  The input unit 34 includes, for example, a button, a switch, a keyboard, or a mouse, and is operated when the user inputs various commands to the intrusion detection device 12.

  The output unit 35 includes a display such as a CRT (Cathode Ray Tube) or an LCD (Liquid Crystal Display), a speaker, and the like, and outputs images and sounds processed by the CPU 31.

  The recording unit 36 is composed of a hard disk or the like, and records or reproduces a program executed by the CPU 31 and information.

  The communication unit 37 includes, for example, a modem and a terminal adapter, and performs communication processing with an external information processing apparatus via a network including the Internet.

  The drive 38 is connected to an input / output interface 42 as necessary, and a removable medium 51 including a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory is appropriately mounted, and a program read from the medium is read by the recording unit 36. To be installed.

  FIG. 4 is a block diagram illustrating a part of a configuration example of functions realized by the CPU 31 that executes a predetermined program. When the CPU 31 executes the program, an image preprocessing unit 71, a reference space data calculation unit 72, a data management unit 73, a change area detection unit 74, a distance calculation unit 75, and a determination unit 76 are realized. The image preprocessing unit 71 is further configured to include a noise filter unit 81, an image information acquisition unit 82, and an interframe difference calculation unit 83.

  The noise filter unit 81 removes noise from the image information input from the imaging device 11 via the input / output interface 42 and the internal bus 41, and supplies the image information from which the noise has been removed to the image information acquisition unit 82. The noise filter unit 81 is, for example, an arithmetic filter such as a median filter, a neighborhood average filter, or a Sobel filter. Further, the noise filter unit 81 may be configured by hardware such as a low-pass filter.

  The image information acquisition unit 82 separates the image information for each frame, and sends the separated image information to the inter-frame difference calculation unit 83, the reference space data calculation unit 72, or the change region detection unit 74 as necessary. Or is recorded in the RAM 33 or the recording unit 36. Hereinafter, information for one frame of image information is also referred to as frame information.

  For each frame information, the inter-frame difference calculation unit 83 sets the unit data value corresponding to each detection unit 21 to the same detection unit 21 of the previous frame information by a predetermined number of frames (for example, set by the user). The difference from the corresponding unit data value is calculated. That is, the difference for each predetermined time interval (frame interval) of the unit data corresponding to the same detection unit 21 is calculated. The inter-frame difference calculation unit 83 supplies data indicating the calculated difference value (hereinafter referred to as difference data) to the reference space data calculation unit 72 or the change area detection unit 74 as necessary, or the RAM 33 or recording. Or recording in the unit 36. Hereinafter, information composed of difference data for one frame is also referred to as difference frame information.

  As will be described later with reference to FIG. 5, the reference space data calculation unit 72 is in a state where an intruder has not entered the monitoring area (hereinafter referred to as a non-intrusion state) before starting monitoring by the monitoring system 1. , Unit data (hereinafter, non-intrusive unit) corresponding to each detection unit 21 in frame information (hereinafter, referred to as non-intrusive frame information) based on image information (hereinafter, referred to as non-intrusive image information) output from the imaging device 11. Difference data corresponding to each detection unit 21 in the difference frame information (hereinafter referred to as non-intrusion difference frame information) based on the average value and standard deviation of the data and the non-intrusion image information (hereinafter referred to as non-intrusion difference frame information). Mean) and standard deviation.

  Further, the reference space data calculation unit 72 is output from the imaging device 11 in a non-intrusive state and in a state in which a predetermined pattern of illumination or video, which will be described later with reference to FIGS. Average value and standard deviation of unit data (hereinafter referred to as pattern unit data) corresponding to each detection unit 21 in frame information (hereinafter referred to as pattern frame information) based on image information (hereinafter referred to as pattern image information); Alternatively, an average value and a standard deviation of difference data (hereinafter referred to as pattern difference data) corresponding to each detection unit 21 in difference frame information (hereinafter referred to as pattern difference frame information) based on the pattern image information are calculated. In addition, the reference space data calculation unit 72 is a correlation matrix (hereinafter referred to as a unit data correlation matrix) indicating a correlation between pattern unit data corresponding to each detection unit 21 or pattern difference data corresponding to each detection unit 21. A correlation matrix (hereinafter referred to as a difference data correlation matrix) indicating the correlation between the two is calculated.

  The reference space data calculation unit 72 indicates the average value and standard deviation of the non-intrusive unit data and the data indicating the unit data correlation matrix, or the average value and standard deviation of the non-intrusive differential data, and the differential data correlation matrix. Data is supplied to the data management unit 73.

  The data management unit 73 stores the average value and standard deviation of the non-intrusive unit data and the data indicating the unit data correlation matrix, or the average value and standard deviation of the non-intrusive differential data and the data indicating the differential data correlation matrix. The data is recorded in the recording unit 36 and managed, and supplied to the distance calculation unit 75 as necessary.

  As will be described later with reference to FIG. 10, the change area detection unit 74 is a unit based on image information (hereinafter referred to as monitoring image information) output from the imaging device 11 after monitoring by the monitoring system 1 is started. A detection unit 21 having a large change in data (hereinafter referred to as monitoring unit data) or difference data (hereinafter referred to as monitoring difference data) is detected, and an area (hereinafter referred to as a predetermined size) centered on the detected detection unit 21. , Referred to as a change region). The change area detection unit 74 supplies data indicating the range of the detected change area to the data management unit 73 and the distance calculation unit 75. In addition, the change area detection unit 74 supplies monitoring unit data or monitoring difference data corresponding to the detection unit 21 included in the change area to the distance calculation unit 75.

  As will be described later with reference to FIG. 10, the distance calculation unit 75 includes a space composed of non-intrusive frame information (non-intrusive image information) (hereinafter referred to as a unit data reference space) and a monitoring unit for one frame. A Mahalanobis distance in a change area between information composed of data (hereinafter referred to as monitoring frame information), or a space composed of non-intrusive differential frame information (hereinafter referred to as differential data reference space), and 1 A Mahalanobis distance in a change area between information composed of monitoring difference data for frames (hereinafter referred to as monitoring difference frame information) is calculated. The distance calculation unit 75 supplies data indicating the calculated Mahalanobis distance value to the determination unit 76.

  As will be described later with reference to FIG. 10, the determination unit 76 determines whether or not an intruder has entered the monitoring area by comparing the Mahalanobis distance supplied from the distance calculation unit 75 with a predetermined detection threshold. To do. When the determination unit 76 determines that an intruder has entered the monitoring area, information (hereinafter referred to as intrusion detection information) notifying that the intruder has been detected in the monitoring area is transmitted to the internal bus 41 and the input / output. The data is supplied to the output unit 35 via the interface 42.

  Next, processing executed by the monitoring system 1 will be described with reference to FIGS.

  First, reference space creation processing executed by the monitoring system 1 will be described with reference to the flowchart of FIG. This process is started, for example, when the user operates the input unit 34 to input a reference space creation process command before starting monitoring by the monitoring system 1.

  In step S <b> 1, the image information acquisition unit 82 acquires image information of a non-intrusive monitoring area. Specifically, the imaging device 11 outputs image information (non-intrusive image information) of an image obtained by imaging a non-intrusive monitoring area to the intrusion detection device 12. The non-intrusive image information input to the intrusion detection device 12 is input to the noise filter unit 81 via the input / output interface 42 and the internal bus 41, and after the noise is removed by the noise filter unit 81, the image information acquisition unit 82. The image information acquisition unit 82 separates the non-intrusive image information frame by frame, records the separated frame information (non-intrusive frame information) in the RAM 33 or the recording unit 36, and supplies the frame information to the inter-frame difference calculation unit 83. To do.

  In step S <b> 2, the interframe difference calculation unit 83 calculates interframe difference data of the non-intrusive frame information. Specifically, the interframe difference calculation unit 83 reads from the RAM 33 or the recording unit 36 the non-intrusive frame information P (P is a natural number) frames before the non-intrusive frame information newly input from the image information acquisition unit 82. put out. This value P is preset by the user, for example. The inter-frame difference calculation unit 83 calculates non-intrusion difference data that is a difference between the latest non-intrusion frame information corresponding to the same detection unit 21 and the non-intrusion unit data of the non-intrusion frame information before the P frame. The interframe difference calculation unit 83 supplies non-intrusion difference frame information configured by one frame of non-intrusion difference data to the reference space data calculation unit 72. The reference space data calculation unit 72 records the non-intrusive difference frame information in the RAM 33 or the recording unit 36.

  The processes of steps S1 and S2 are repeatedly executed until, for example, a predetermined amount of non-intrusive difference frame information is accumulated or a predetermined time elapses. Note that the period for collecting the non-intrusive differential frame information is preferably set to a period in which all the expected changes in the environment of the monitoring area (for example, changes in external light, vibrations, etc.) are included.

Hereinafter, the number of collected non-intrusive differential frame information is ns, and the non-intrusive differential data corresponding to the i-th detection unit 21 in the j-th non-intrusive differential frame information is Xs _ij .

In step S3, the reference space data calculation unit 72 calculates the average value and standard deviation of the non-intrusion difference data. Specifically, the reference space data calculation unit 72 reads all the non-intrusion difference frame information from the RAM 33 or the recording unit 36, and calculates the average value and standard deviation for each non-intrusion difference data corresponding to each detection unit 21. . Hereinafter, the average value of the non-invasion differential data corresponding to the i-th detection units 21 and ms _i, the standard deviation and [sigma] s _i.

  In step S4, the data management unit 73 records the average value and standard deviation of the non-intrusive difference data. Specifically, the data management unit 73 acquires data indicating the average value and standard deviation of the non-intrusion difference data of each detection unit 21 from the reference space data calculation unit 72, and records the data indicating the average value and standard deviation. The data is recorded in the unit 36.

  In step S5, the image information acquisition unit 82 acquires the image information of the monitoring area in a state where a predetermined pattern is projected. Specifically, the user projects (projects) illumination whose illumination has been changed in a predetermined pattern using a lighting device (not shown) onto a monitoring area, or uses a video device (not shown) to form a predetermined pattern. Is projected onto the monitoring area.

  6 to 9 show examples of illumination or video patterns projected onto the monitoring area. For example, as shown in FIG. 6, two different colors or brightness of horizontal stripes of illumination or video are projected onto the monitoring area, and the intervals between the stripes are changed randomly, or indicated by arrows in the figure. Each stripe is moved at random speed from top to bottom. In addition, for example, as shown in FIG. 7, two types of different colors or brightness of vertical stripes of illumination or video are projected onto the monitoring area, and the intervals between the stripes are changed randomly, or arrows in the figure Each stripe is moved at random speed from left to right as shown in FIG.

  Further, for example, as shown in FIG. 8, two types of different stripes of illumination with different colors or brightness or images are projected onto the monitoring area, and the intervals between the stripes are randomly changed. Each stripe is moved at a random speed from left to right as indicated by the arrows in the figure. Also, for example, as shown in FIG. 9, two different colors or brightness of a checkered pattern illumination or video is projected onto the monitoring area, and the interval between the stripes is changed randomly, or an arrow in the figure. Each stripe is moved at random speed from left to right or from top to bottom as shown in FIG.

  The imaging device 11 outputs to the intrusion detection device 12 image information (pattern image information) of an image obtained by imaging a monitoring area in a non-intrusive state and in a state where the illumination of the pattern shown in FIGS. To do. The pattern image information input to the intrusion detection device 12 is input to the noise filter unit 81 via the input / output interface 42 and the internal bus 41, and after the noise is removed by the noise filter unit 81, the image information acquisition unit 82. To be supplied. The image information acquisition unit 82 records frame information (pattern frame information) obtained by separating the non-intrusive image information for each frame in the RAM 33 or the recording unit 36 and supplies the frame information to the inter-frame difference calculation unit 83.

  In step S6, the inter-frame difference calculation unit 83 calculates the inter-frame difference data of the pattern frame information. Specifically, the inter-frame difference calculation unit 83 reads pattern frame information P (P is a natural number) frames before the pattern frame information newly input from the image information acquisition unit 82 from the RAM 33 or the recording unit 36. This value P is the same value as in the case of calculating difference data between frames of the non-intrusive frame information in step S2. The inter-frame difference calculation unit 83 calculates pattern difference data that is the difference between the latest pattern frame information corresponding to the same detection unit 21 and the pattern unit data of the pattern frame information before the P frame. The inter-frame difference calculation unit 83 supplies pattern difference frame information configured by pattern difference data for one frame to the reference space data calculation unit 72. The reference space data calculation unit 72 records the pattern difference frame information in the RAM 33 or the recording unit 36.

  The processes in steps S5 and S6 are repeatedly executed until, for example, a predetermined amount of pattern difference data is accumulated or a predetermined time elapses.

Hereinafter, the number of collected pattern difference frame information is np, and the pattern difference data corresponding to the i-th detection unit 21 in the j-th pattern difference frame information is Xp _ij .

In step S7, the reference space data calculator 72 calculates a correlation matrix. Specifically, the reference space data calculation unit 72 reads all the pattern difference frame information from the RAM 33 or the recording unit 36, and calculates an average value and a standard deviation for each pattern difference data corresponding to each detection unit. Hereinafter, the average value of the pattern difference data corresponding to the i-th detection units 21 and mp _i, the standard deviation and .sigma.p _i.

  Next, the reference space data calculation unit 72 calculates a correlation matrix R (difference data correlation matrix) indicating a correlation between pattern difference data corresponding to each detection unit 21 based on the following equations (1) to (3). Is calculated.

x_ij＝(Xp_ij−mp_i)÷σp_i ・・・(2)
r_ij＝(x_i1 x_j1＋x_i2 x_j2＋・・・＋x_inp x_jnp)÷np ・・・(3)
ただし、1≦i≦k，1≦j≦k

x _ij = (Xp _ij −mp _i ) ÷ σp _i (2)
r _ij = (x _i1 x _j1 + x _i2 x _j2 + ... + x _inp x _jnp ) ÷ np (3)
However, 1 ≦ i ≦ k, 1 ≦ j ≦ k

The correlation coefficient r _ij is a correlation coefficient indicating the correlation between the pattern difference data corresponding to the i th detection unit 21 and the pattern difference data corresponding to the j th detection unit 21.

In step S8, the data management unit 73 records the correlation matrix calculated in step S7, and the reference space creation process ends. Specifically, the data management unit 73 acquires data indicating each element (correlation coefficient r _ij ) of the correlation matrix R from the reference space data calculation unit 72, and records data indicating each element of the correlation matrix R as a recording unit 36.

  Next, the monitoring process executed by the monitoring system 1 will be described with reference to the flowchart of FIG.

  In step S31, the image information acquisition unit 82 acquires image information of the monitoring area. Specifically, the imaging device 11 outputs image information (monitoring image information) of an image obtained by imaging the monitoring area to the intrusion detection device 12. The monitoring image information input to the intrusion detection device 12 is input to the noise filter unit 81 via the input / output interface 42 and the internal bus 41, and after the noise is removed by the noise filter unit 81, the image information acquisition unit 82. To be supplied. The image information acquisition unit 82 records frame information obtained by separating the monitoring image information for each frame (hereinafter referred to as monitoring frame information) in the RAM 33 or the recording unit 36 and supplies it to the inter-frame difference calculation unit 83.

  In step S32, the inter-frame difference calculation unit 83 calculates the inter-frame difference data of the monitoring image information. Specifically, the interframe difference calculation unit 83 reads the monitoring frame information P (P is a natural number) frames before the monitoring frame information newly input from the image information acquisition unit 82 from the RAM 33 or the recording unit 36. This value P is the same value as the case of calculating difference data between frames of the non-intrusive frame information in step S2 of FIG. The inter-frame difference calculation unit 83 calculates monitoring difference data that is a difference between the latest monitoring frame information corresponding to the same detection unit 21 and the monitoring unit data of the monitoring frame information before the P frame. The inter-frame difference calculation unit 83 supplies monitoring difference frame information constituted by monitoring difference data for one frame to the reference space data calculation unit 72. The reference space data calculation unit 72 supplies the monitoring difference frame information to the change region detection unit 74.

Hereinafter, the monitoring differential data corresponding to the i-th detection unit 21 and Xw _i.

In step S33, the change area detection unit 74 identifies a change area. Specifically, for example, the change area detection unit 74 first uses the average value ms _i of the non-intrusion difference data and the standard deviation σs _i based on the following equation (4) to determine the monitoring difference data Xw _i. Standardized data (hereinafter referred to as standardized monitoring difference data) Xcw _i (i = 1, 2,..., K) is obtained.

Xcw _i = (Xw _i −ms _i ) ÷ σs _i (4)

The change area detection unit 74 standardizes the detection unit 21 having the largest absolute value of the standardized monitoring difference data Xcw _i and the monitoring difference data using the average value ms _i of the non-intrusive difference data and the standard deviation σs _i in the difference data reference space. The detection unit 21 having the largest absolute value is detected. The change area detection unit 74 includes the detected detection unit 21 and identifies an area included in the monitoring area, as a change area, of an area having a predetermined size centered on the detection unit 21.

  For example, as shown in FIG. 11, when the intruder 102 enters the monitoring area 101 and the standardized monitoring difference data corresponding to the detection unit 21 included in the area into which the intruder 101 has entered becomes maximum, the detection is performed. The area 103 including the unit 21 is specified as the change area. In the following description, the number of detection units 21 included in the change area is assumed to be q.

In step S34, the distance calculation unit 75 calculates an inverse matrix of the correlation matrix. Specifically, the change area detection unit 74 supplies information indicating the range of the change area to the data management unit 73 and the distance calculation unit 75. The data management unit 73 reads only the correlation coefficient r _ij corresponding to the detection unit 21 included in the change area from the correlation coefficient r _ij constituting the correlation matrix R from the recording unit 36, and reads the read correlation coefficient r _ij is supplied to the distance calculation unit 75. The distance calculator 75 uses the acquired correlation coefficient r _ij to generate a correlation matrix Rp corresponding to the detection unit 21 in the change region, and further calculates an inverse matrix Rp ⁻¹ of the correlation matrix Rp.

In step S35, the distance calculation unit 75 calculates the Mahalanobis distance between the difference data reference space and the monitoring difference frame information. Specifically, the change area detection unit 74 supplies the normalized difference data Xcw _i of the detection unit 21 included in the changing area to the distance calculation unit 75. The distance calculation unit 75 calculates the Mahalanobis distance D between the difference data reference space and the difference monitoring frame information in the change area based on the following equations (5) and (6). Strictly speaking, the distance calculation unit 75 includes a space constituted by non-intrusion difference data corresponding to the detection unit 21 included in the change area and monitoring difference data corresponding to the detection unit 21 included in the change area. The Mahalanobis distance D between the measured data is calculated.

In addition, each element c _i (i = 1, 2,..., Q) of the matrix C in Expression (6) sequentially arranges standardized difference data Xcw _i corresponding to q detection units 21 included in the change region. The element r _ij of the correlation matrix Rp, which is a q-order square matrix, corresponds to the correlation coefficient between detection units constituting the matrix C, and can be cited from the elements of the correlation matrix R.

In addition, when calculating the Mahalanobis distance D, a correlation matrix R indicating a correlation between detection units measured using a predetermined pattern prepared instead of a correlation matrix between detection units generated from information obtained by measuring a non-intrusion state. (Correlation coefficient r _ij ) is used for the following reason. That is, when the unit data (non-intrusion unit data) corresponding to each detection unit 21 hardly changes in the non-intrusion state, the non-intrusion difference data that is a difference between predetermined frames of the non-intrusion unit data is 0 or 0. It is likely that close values will continue. In that case, the actual correlation between the difference data corresponding to each detection unit 21 may not be sufficiently reflected in the correlation coefficient between the non-intrusion difference data.

Therefore, for example, as described above, in a non-intrusive state, the predetermined pattern shown in FIGS. 6 to 9 is projected onto the monitoring area, and the correlation between the difference data corresponding to each detection unit 21 appears significantly. The correlation matrix R (correlation coefficient r _ij ) is calculated based on the difference data (pattern difference data) collected in step (1). Thus, the correlation matrix R (correlation coefficient r _ij ) more accurately reflects the correlation of the difference data corresponding to each detection unit 21, and the correlation matrix R (correlation coefficient r _ij ) is The Mahalanobis distance D calculated by using the value is a more accurate and reliable value.

For example, when the state of the monitoring area (for example, illuminance, etc.) changes frequently, the correlation coefficient r _ij based on the non-intrusive difference data sufficiently reflects the correlation of the difference data corresponding to each detection unit 21. In this case, the correlation matrix R may be calculated using non-intrusive difference data.

  Further, by obtaining the Mahalanobis distance D not in the entire monitoring area but in a change area that is a part of the monitoring area, there are the following advantages. That is, for example, in the monitoring area 101, when the luminance changes in a distributed manner in a plurality of areas as shown in FIG. 12, and when the luminance changes in one area as shown in FIG. If the changed luminance values are substantially the same and the total areas of the areas where the luminance changes are substantially the same, the Mahalanobis distance D between the difference data reference space and the monitoring difference frame information in the monitoring area 101 is approximately There is a tendency to be equal. That is, based on the calculated Mahalanobis distance D, the state of the monitoring area 101 shown in FIG. 12 and the state of the monitoring area 101 shown in FIG.

  On the other hand, the state shown in FIG. 12 and the state shown in FIG. 13 are calculated based on the calculated Mahalanobis distance D by calculating the Mahalanobis distance D in the change area around the detection unit 21 where the monitoring difference data has changed the most. Can be more accurately distinguished. Further, by calculating the Mahalanobis distance D in the change area, the calculation amount is reduced and the speed of detecting the intruder is increased as compared with the case of calculating the Mahalanobis distance in the entire monitoring area.

  For example, when the monitoring area is small or when it is not necessary to distinguish between the states shown in FIGS. 12 and 13, the processing in step S33 is omitted and the Mahalanobis distance D in the entire monitoring area is set. You may make it calculate. In addition, for example, the size of the change region may be obtained based on experimental results so that a state that needs to be detected can be distinguished from a state that does not need to be detected.

  In step S36, the determination unit 76 determines whether the Mahalanobis distance D is greater than a predetermined detection threshold. Specifically, the determination unit 76 acquires data indicating the Mahalanobis distance D from the distance calculation unit 75. When the determination unit 76 compares the Mahalanobis distance D with a preset detection threshold value and determines that the Mahalanobis distance D is equal to or less than the predetermined detection threshold value, that is, the monitoring area may be in a non-intrusive state. When it is determined that the property is high, the process returns to step S31, and the processes after step S31 described above are repeatedly executed. That is, monitoring of the monitoring area is continuously executed.

  If it is determined in step S36 that the Mahalanobis distance D is greater than the predetermined detection threshold, that is, if it is determined that there is a high possibility that an intruder has entered the monitoring area, the process proceeds to step S37.

  A method for determining the detection threshold will be described later with reference to FIG.

  In step S <b> 37, the determination unit 76 notifies the user of the intruder entry. Specifically, the determination unit 76 supplies intrusion detection information to the output unit 35 via the internal bus 41 and the input / output interface 42. The output unit 35 displays, for example, an image for notifying the intruder into the monitoring area on the display or outputs a sound for notifying the intrusion from the speaker, so that the intruder enters the monitoring area. To be notified.

  Thereafter, the process returns to step S31, and the processes after step S31 described above are repeatedly executed. That is, monitoring of the monitoring area is continuously executed.

  In the above description, the monitoring area is monitored based on the difference data corresponding to each detection unit 21. However, the monitoring area is monitored using unit data instead of the difference data. May be. In this case, the Mahalanobis distance between the unit data reference space and the monitoring frame information is calculated based on the average value and standard deviation of the non-intrusive unit data and the correlation matrix calculated based on the pattern unit data, and the monitoring area What is necessary is just to detect the intrusion of the intruding object into. In this case, for example, when the state of the monitoring area (for example, illuminance, etc.) frequently changes, the correlation between the unit data corresponding to each detection unit 21 is sufficiently reflected in the correlation coefficient based on the non-intrusive unit data. In this case, the correlation matrix may be calculated using non-intrusive unit data.

  Also, by reducing the frame interval P for obtaining the difference data (by shortening the time interval), the detection accuracy of intruders that change faster is improved, and by increasing the frame interval P (by increasing the time interval). This improves the accuracy of detecting an intruder that enters or changes over a long period of time. Further, by monitoring the monitoring area using unit data instead of the difference data, the intrusion of the intruder can be detected by directly comparing the initial state (non-intrusion state) and the current monitoring state.

Here, an example of how to determine the detection threshold will be described with reference to FIG. The distribution (probability density function) of the Mahalanobis distance D calculated based on the above equation (5) substantially follows a standard normal distribution having an average value of 0 and a standard deviation σ ₀ of 1 as represented by a line 151 (formula (5) defines Mahalanobis square distance D ² and assumes that the variation in D ² is random, D ² follows a chi-square distribution, and the positive part of D follows a standard normal distribution.) For example, when the detection threshold is 5 which is five times the standard deviation σ ₀ of the Mahalanobis distance D in the non-intrusion state, that is, when the Mahalanobis distance D exceeds 5, there is a possibility that an intruder has entered the monitoring area. If it is determined that there is an intruder, the probability that the Mahalanobis distance D in the non-intrusion state is a value that is more than five times the standard deviation σ ₀ from the average value, that is, the intruder has entered the monitoring area despite the non-intrusion state. The probability that it is determined that there is a possibility (false detection) is 2.9 × 10 ⁻⁷ .

In a normal distribution, the value of 2.9 × 10 ^-7, which is the probability that a value that is more than 5 times the standard deviation from the average value, appears, for example, the reliability of the safety device according to IEC (International Electro technical Commission) 61696-3 It is a value that is regarded as having no malfunction in the standard, and in fact, the probability of an accident per the number of flights of an airplane is almost equal to this value. Therefore, when the detection threshold is set to 5, it can be considered that there is almost no possibility that it is determined that there is a possibility that an intruder has entered the monitoring area in spite of the non-intrusion state.

On the other hand, a line 152 indicates the distribution (probability density function) of the Mahalanobis distance D detected when an intruder in the monitoring area (hereinafter referred to as intruder A) enters. 40 the average m _t of the Mahalanobis distance D in the distribution indicated by line 152, when the standard deviation sigma _t was 7, the probability that the Mahalanobis distance D is less than 5 the detection threshold, i.e., the intruder A in the monitoring area The possibility of being missed in spite of the intrusion is 2.9 × 10 ⁻⁷ because the detection threshold value 5 is away from the average value 40 of the distribution indicated by the line 152 by 5 × σ _t . Can be regarded as almost none.

That is, if the average value m _t and standard deviation sigma _t of the distribution of the Mahalanobis distance D in a state where the intruder A has entered the monitoring area satisfies the relationship of formula (7), without erroneous detection of the entrance of the intruder A It can be considered that it can be detected almost certainly. In other words, in the monitoring system 1, when the detection threshold is set to 5, an intruder that has an average value and a standard deviation of the Mahalanobis distance D when entering the monitoring area satisfies the relationship of Expression (7) is detected. It can be assured that it can be detected almost certainly. Therefore, it is possible to identify an intruder that can be detected by the monitoring system 1 in advance.

m _t ≧ 5 + σ _t × 5 (7)

  Note that the above is an example of how to determine the detection threshold value. Naturally, the detection threshold value may be determined based on other reasons such as safety.

  Next, referring to FIG. 15 to FIG. 18, in the monitoring area in which pixels as detection units are arranged in 3 rows × 3 columns shown in FIG. 15, the difference between the difference data reference space and the monitoring difference frame information is set. When obtaining the Mahalanobis distance, the Mahalanobis distance calculated using the detected monitoring difference data as it is, that is, the actual Mahalanobis distance and the feature quantity extracted from the detected monitoring difference data, and the Mahalanobis calculated using the extracted feature quantity The results of experiments comparing distances will be described.

  FIG. 15 is a diagram illustrating a monitoring region used in the experiment. The detection units in the monitoring area are the detection unit a11, the detection unit a12, and the detection unit a13 from the left in the first row, and the detection unit a21, the detection unit a22, and the detection unit a23 from the left in the second row. From the left, the detection unit is a detection unit a31, the detection unit a32, and the detection unit a33.

  FIG. 16 is a table showing the value of difference data (non-intrusion difference data) between predetermined frames of unit data (pixel data) corresponding to each detection unit detected in the non-intrusion state. The number of detections of the difference data is 5, the average value of the difference data corresponding to each detection unit is shown in the second column from the right, and the standard deviation of the difference data is shown in the rightmost column. For example, as shown in the third row from the top of the table, for the detection unit a11, the detection value of the first difference data is 1, the detection value of the second difference data is 5, and the detection value of the third difference data. 5, the detection value of the fourth difference data is 3, the detection value of the fifth difference data is 1, the average value of the difference data is 3, and the standard deviation is 2.00. Hereinafter, a space constituted by the difference data shown in FIG. 16 is referred to as a difference data reference space.

  FIG. 17 shows the difference data value corresponding to one detection unit as a value obtained by adding 10 times the standard deviation of the difference data to the average value of the difference data in the difference data reference space for each detection unit. It is a table | surface which shows the Mahalanobis distance between the monitoring difference frame information comprised by the difference data which made the value of corresponding difference data become the average value of difference data in difference data reference space, and difference data reference space. For example, in the first monitoring difference frame information, the difference data value corresponding to the detection unit a11 is set to 23, which is 10 times the standard deviation 2 to the average value 3 in the difference data reference space, and corresponds to another detection unit. The value of the difference data is 3 which is the average value in the reference space.

  Then, the Mahalanobis distance between the difference data reference space and the monitoring difference frame information was calculated by the following three methods. The value of the Mahalanobis distance shown in FIG. 17 is a square value of the Mahalanobis distance calculated based on the equation (5).

  As a first method, the Mahalanobis distance was calculated based on the above-described equation (5) using the difference data shown in FIG. 16 and the difference data shown in FIG. 17 as they are. That is, the Mahalanobis distance calculated by the first method indicates the actual Mahalanobis distance between the difference data reference space and the monitoring difference frame information. In the table of FIG. 17, the value of the Mahalanobis distance calculated by the first method is shown in the third row from the bottom labeled “direct calculation” in the second column from the left. For example, the first Mahalanobis distance value calculated by the first method is 33.12.

  As a second method, the detection units arranged in the same row in the horizontal direction of the monitoring area shown in FIG. 15 are taken as a set, and the maximum value, the minimum value, and the average value of the difference data of each set are extracted as feature amounts. The Mahalanobis distance was calculated based on the above-described equation (5) using the feature amount of the difference data. That is, the monitoring area shown in FIG. 15 is divided into three sets of detection units a11 to a13, a21 to a23, and a31 to a33, and the maximum value, the minimum value, and the average value of the difference data for each time of each set The Mahalanobis distance between the space constituted by and the data constituted by the maximum value, the minimum value, and the average value of the difference data for each set of difference data shown in FIG. 17 was calculated. In the table of FIG. 17, the Mahalanobis distance value calculated by the second method is shown in the second row from the bottom labeled “Calculated by lateral feature value” in the second column from the left. . For example, the first Mahalanobis distance calculated by the second method is 30.50.

  As a second method, the detection units arranged in the same column in the vertical direction of the monitoring area shown in FIG. 15 are taken as a set, and the maximum value, the minimum value, and the average value of the difference data of each set are extracted as feature amounts. The Mahalanobis distance was calculated based on the above-described equation (5) using the feature amount of the difference data. That is, the monitoring area shown in FIG. 15 is divided into three sets of detection units a11 to a31, a12 to a32, and a13 to a33, and the maximum value, the minimum value, and the average value of the difference data for each time of each set The Mahalanobis distance between the space constituted by and the data constituted by the maximum value, the minimum value, and the average value of the difference data for each set of difference data shown in FIG. 17 was calculated. In the table of FIG. 17, the Mahalanobis distance value calculated by the third method is shown in the bottom row labeled “Calculated by the feature amount in the vertical direction” in the second column from the left. For example, the value of the first Mahalanobis distance calculated by the third method is 104.63.

  FIG. 18 is a graph showing the distribution of Mahalanobis distance values calculated by the first to third methods. The horizontal axis indicates the Mahalanobis distance calculated based on the monitoring difference frame information (difference data) in FIG. 17, and the vertical axis indicates the square value of the calculated Mahalanobis distance. Line 201 shows the distribution of Mahalanobis distance values calculated by the first method, line 202 shows the distribution of Mahalanobis distance values calculated by the second method, and line 203 calculates by the third method. Shows the distribution of Mahalanobis distance values.

  Even if only the difference data corresponding to one detection unit is uniformly changed so as to add 10 times the standard deviation to the average value of the difference data in the reference space, as shown by lines 202 and 203, The Mahalanobis distance calculated using the quantity varies depending on the detection unit in which the difference data is changed. On the other hand, the Mahalanobis distance calculated using the difference data as it is, as shown by the line 201, has almost no variation and is almost constant.

  Therefore, as in the monitoring system 1 described above, the calculation using the difference data or the unit data as it is is more accurate than the calculation using the feature amount extracted from the difference data or the unit data, for example. As a result, the reliability of the detection of the intrusion of the intruder in the monitoring area becomes higher.

  As described above, it is possible to more easily and accurately detect the intruder intrusion into the monitoring area without registering the expected feature quantity of the intruder in advance or using a complicated algorithm. .

  Next, with reference to FIGS. 19 to 23, a calculation amount for calculating the correlation matrix R and a method for reducing the data amount will be described.

As described above, when the monitoring area is monitored using the difference data or the unit data as they are, for example, when 600 vertical x 800 horizontal detection units 21 are arranged in the monitoring area, one detection unit 21 is provided. There is a correlation coefficient r _ij with the difference data or unit data corresponding to the other 479999 detection units 21, and the size of the correlation matrix R in the equation (1) is 480000. It becomes row x 480000 columns.

  By the way, the correlation of difference data or unit data becomes weaker as the position of the corresponding detection unit 21 is separated. Therefore, the calculation amount for calculating the correlation matrix R can be reduced by limiting the range for obtaining the correlation (correlation coefficient) of the difference data or the unit data.

  For example, as shown in FIG. 19, when only the correlation between adjacent detection units a1 to a8 is considered for the detection unit z1, the correlation coefficient between a maximum of 8 detection units per detection unit You can ask for. In addition, as shown in FIG. 20, when only the correlation between the detection units a1 to a8 adjacent to the detection unit z1 and the detection units b1 to b16 adjacent to the outside of the detection units a1 to a8 is considered, What is necessary is just to obtain | require the correlation coefficient between a maximum of 24 detection units per detection unit. Further, as shown in FIG. 21, for the detection unit z1, adjacent detection units a1 to a8, detection units b1 to b16 adjacent to the outside of the detection units a1 to a8, and adjacent to the outside of the detection units b1 to b16 When only the correlation with the detection units c1 to c24 to be considered is considered, it is only necessary to obtain correlation coefficients between a maximum of 48 detection units per detection unit.

  In general, if the correlation coefficient is 0.7 or more, two events have a strong correlation, and if the correlation coefficient is less than 0.3, the two events are regarded as uncorrelated. For example, when the detection unit 21 corresponds to a region imaged by one light receiving element of the CCD, each light receiving element captures an image independently, and a video signal of a predetermined format is generated by combining output signals from each light receiving element. As a result, the dependency between adjacent light receiving elements exists as a finite value, but the unit data or difference data corresponding to two different detection units 21 except for a failure such as an abnormal stroke between the light receiving elements. It can be considered that there is almost no strong correlation with a correlation coefficient of 0.7 or more. That is, the correlation coefficient between unit data or difference data corresponding to two different detection units 21 can be regarded as 0.7 or less.

  Accordingly, the correlation coefficient between unit data or difference data corresponding to a detection unit (for example, detection unit z1 and detection unit c10 in FIG. 21) sandwiching two detection units is 0.343 (= 0.7 to the third power). ) Can be considered as follows. Furthermore, the correlation coefficient between the unit data corresponding to the detection unit 21 located between the three detection units or the difference data is 0.2401 (= 0.7 4 <0.3) at the maximum, and is regarded as uncorrelated. be able to. Therefore, for example, the detection unit z1 in FIG. 21 can be regarded as uncorrelated with the detection units adjacent to the outside of the detection units c1 to c24.

Accordingly, each detection unit 21 corresponds to difference data or unit data corresponding to the detection unit 21 and 48 detection units 21 within a range of 7 rows × 7 columns centering on the detection unit 21. Even if only the correlation coefficient r _ij between the difference data or the unit data is calculated and the other correlation coefficients are set to 0, the calculated correlation matrix R is the unit data corresponding to each detection unit 21 or It can be considered that the correlation between the difference data is reflected almost accurately. In this case, as shown in Expression (8), the correlation matrix R is a symmetric band matrix having a band having a predetermined width.

（ただし、t＝k−s＋１）

(However, t = k−s + 1)

As a result, if only the correlation coefficient r _ij in the band of the upper triangular portion excluding the diagonal component of the correlation matrix R is recorded, the correlation matrix R can be generated and recorded when necessary. The amount of data to keep can be reduced.

In addition, when an intruder enters the monitoring area, it is guaranteed that it always passes through the outer periphery of the monitoring area. As shown in FIG. 22, a predetermined width in contact with the outer periphery of the monitoring area, as shown in FIG. If only the frame-shaped area 201 is set as the monitoring target, only the correlation between the difference data or the unit data corresponding to the detection unit 21 included in the area 201 needs to be taken into consideration. The number of relations r _ij can be reduced.

FIG. 23 is a diagram showing the upper left corner of the area 201 in FIG. One rectangle to which a number in the area 201 is assigned represents one detection unit, and the number assigned to each detection unit indicates an array number of the detection unit. The array number here is an element number of a matrix used when the difference data or unit data of the detection unit is arranged in a column matrix (or row matrix) or the correlation coefficient r _ij is arranged in the correlation matrix R. .

  In the example shown in FIG. 23, seven detection units are arranged in the width direction of each side of the region 201, array number 1 is assigned to the detection unit in the upper left corner, and the detection units on the same column as array number 1 are assigned. , Array numbers 2 to 7 are assigned in order in the width direction of the upper side of the area 201 (from the top to the bottom), and the detection units in the next column (the column on the right) are similarly arranged from the top to the bottom. Numbers 8 to 14 are assigned, and similarly, array numbers are assigned to the detection units so as to repeat in the width direction of the frame of the region 201 up to the rightmost column on the upper side of the region 201.

  By assigning array numbers in this way, for example, for one detection unit, correlation coefficients with difference data or unit data corresponding to a maximum of 24 detection units in which the array numbers before and after the detection unit are continuous. 21, as shown in FIG. 21, it is within the range of 7 rows × 7 columns centering on the detection unit, and within the region 201 (more strictly speaking, within one side of the region 201). The correlation coefficient with the difference data or the unit data corresponding to the included detection unit can be obtained.

  For example, in the case of the detection unit of SEQ ID NO: 25, 24 detection units (detection units of SEQ ID NOS: 1 to 24) in which the sequence numbers before the detection unit of SEQ ID NO: 25 are continuous, and the detection units of SEQ ID NO: 25 The correlation coefficient with the difference data or the unit data corresponding to 24 detection units (detection units of array numbers 25 to 49) with subsequent array numbers is obtained, and in the case of the detection unit of array number 77, array number 77 24 detection units with consecutive array numbers (detection units with array numbers 53 to 76) and 24 detection units with consecutive array numbers after the detection unit with array number 77 (array) The correlation coefficient with the difference data or unit data corresponding to the detection units (numbers 78 to 101) is obtained, and the correlation coefficient with the difference data or unit data corresponding to the other detection units is set to 0. It may be Re.

  Thereby, the width of the band of the correlation coefficient R can be reduced to 49, and the amount of data to be recorded can be further reduced.

  Further, for example, if a correlation between difference data or unit data corresponding to a maximum of 48 detection units in which the sequence numbers before and after each detection unit are continuous is obtained, the detection unit is further changed. For each of the detection units included in the range of 7 rows × 7 columns as the center, it is within the range of 7 rows × 7 columns around the detection units and in the region 201 (more strictly speaking, It becomes possible to obtain up to the correlation coefficient between the difference data or the unit data corresponding to the detection unit included in one side of the region 201.

  Further, for example, a correlation with difference data or unit data corresponding to a detection unit of an adjacent portion of each side of the region 201 may be considered. For example, when calculating the correlation coefficient corresponding to the detection unit of the upper side of the region 201, the detection unit of the left side of the lower region 201 of the detection units of the array element numbers 7, 14, 21, 28, 35, 42, and 49 The correlation coefficient with the difference data or unit data corresponding to may be calculated.

Further, since the correlation matrix R is a symmetric matrix with a diagonal component of 1, the correlation matrix Rp in the change region is also a symmetric matrix with a diagonal component of 1. Using this property, when calculating the inverse matrix Rp ^-1 of the correlation matrix Rp of the change region, the number of calculations when calculating the Mahalanobis distance D by obtaining the inverse matrix Rp ^-1 in the form of LU decomposition Can be reduced from the level of n to the cube (where n is the order of the inverse matrix Rp ⁻¹ ) to the level of n to the square. Furthermore, when the range for obtaining the correlation coefficient is limited, and the correlation matrix R is a symmetric band matrix with a diagonal component of 1, the inverse matrix Rp ⁻¹ is obtained using the modified Cholesky method, thereby further reducing the calculation speed. The speed can be increased.

  As described above, the intrusion of the object into the monitoring area based on the monitoring data composed of the unit data corresponding to each of the detection units obtained by dividing the monitoring area into a plurality of areas, which is output from the sensor monitoring the monitoring area. In monitoring, a program for performing monitoring processing, and a first unit that is unit data constituting first monitoring data that is monitoring data output from a sensor in a state in which nothing has entered the monitoring area 1st difference monitoring which comprises the 1st correlation matrix which shows the correlation between data, or the 1st difference unit data which is the difference for every predetermined time interval of the 1st unit data corresponding to the same detection unit When the second correlation matrix indicating the correlation between the first difference unit data of the data is calculated and the entry of the object into the monitoring area is monitored, the first monitoring data configured by the first monitoring data The first Mahalanobis distance between the space and the second monitoring data that is the monitoring data output from the sensor is calculated using the first correlation matrix, or is configured by the first difference monitoring data And second difference unit data that is a difference for each time interval of second unit data that is unit data constituting second monitoring data corresponding to the same detection unit. The second Mahalanobis distance between the two differential monitoring data is calculated using the second correlation matrix, and the intrusion into the monitoring area is based on the first Mahalanobis distance or the second Mahalanobis distance. Can be detected. In addition, it is possible to detect the intrusion of objects more easily and accurately.

  In addition, in monitoring the intrusion of the monitoring area based on the monitoring data that is output from the sensor that monitors the monitoring area and is composed of unit data corresponding to each of the detection units that divide the monitoring area into a plurality of areas, The average value and standard deviation of the first unit data that is the unit data constituting the first monitoring data that is the monitoring data output from the sensor or the same detection unit in the state where no object has entered the monitoring area An average value and a standard deviation of the first difference unit data of the first difference monitoring data constituted by the first difference unit data which is a difference at predetermined time intervals of the first unit data corresponding to At the same time, the sensor exits in a state where no object has entered the monitoring area and an image or illumination of a predetermined pattern is projected on the monitoring area. The first correlation matrix indicating the correlation between the second unit data that is the unit data constituting the second monitoring data that is the monitored data, or the predetermined second unit data corresponding to the same detection unit A second correlation matrix indicating the correlation between the second difference unit data of the second difference monitoring data constituted by the second difference unit data which is a difference for each time interval is calculated, and intrusion into the monitoring area , The first Mahalanobis distance between the first space constituted by the first monitoring data and the third monitoring data, which is the monitoring data output from the sensor, is defined as the first unit data. Or the third monitoring data corresponding to the same detection unit as that of the second space constituted by the first difference monitoring data. Composition unit The second Mahalanobis distance from the third difference monitoring data constituted by the third difference unit data that is the difference at each time interval of the third unit data that is the data is the first difference unit data. When the intrusion into the monitoring area is determined based on the first Mahalanobis distance or the second Mahalanobis distance, the average value and the standard deviation of Intrusion of things can be detected. In addition, it is possible to detect the intrusion of objects more easily and accurately.

  In the above description, an example of monitoring using image information of an image captured by the imaging device 11 has been shown. For example, a temperature distribution of a substance based on infrared radiation energy detected by an infrared thermography device or the like is shown. You may make it use the image information of the image which shows distribution of time until the laser beam and sound wave detected by the apparatus which scans a monitoring area | region while projecting image information and a laser beam and a sound wave. In this case, instead of the patterns shown in FIGS. 6 to 9, the correlation matrix is calculated based on image information acquired by moving an object having a temperature change or an uneven object in the monitoring region. do it.

  Further, in step S33 of FIG. 10, a plurality of change areas may be specified. For example, the detection unit 21 in which the standardized monitoring difference data is maximized or minimized can be detected, and a plurality of change regions can be specified for each detected detection unit 21. When specifying a plurality of change areas, for example, the maximum value, the average value, or the total value of Mahalanobis distances in the plurality of change areas is compared with a predetermined detection threshold value, so that an intruder enters the monitoring area. May be detected.

  Further, for each piece of frame information, difference data with frame information at a plurality of frame intervals is calculated, and the Mahalanobis distance is calculated using an average value, maximum value, integrated value, etc. of the plurality of difference data. Also good.

  Further, a plurality of detection units may be combined into one to calculate difference data or unit data. Furthermore, only detection units having a large correlation coefficient (strong correlation) may be combined into one detection unit.

  Further, after the difference data is calculated by the inter-frame difference calculation unit 83, the noise may be removed by the noise filter unit 81.

  The series of processes described above can be executed by hardware or can be executed by software. When a series of processing is executed by software, a program constituting the software may execute various functions by installing a computer incorporated in dedicated hardware or various programs. For example, it is installed from a network or a recording medium into a general-purpose personal computer or the like.

  As shown in FIG. 3, this recording medium is distributed to provide a program to a user separately from the computer, and is removable such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory on which the program is recorded. It is not only configured by package media including the media 51, but also configured by a ROM 32 on which a program is recorded, a hard disk included in the recording unit 36, and the like provided to the user in a state of being preinstalled in a computer. .

  The program for executing the series of processes described above is installed in a computer via a wired or wireless communication medium such as a local area network, the Internet, or digital satellite broadcasting via an interface such as a router or a modem as necessary. You may be made to do.

  Further, in the present specification, the step of describing the program stored in the recording medium is not limited to the processing performed in chronological order according to the described order, but is not necessarily performed in chronological order. It also includes processes that are executed individually.

  Further, in the present specification, the term “system” means an overall apparatus constituted by a plurality of apparatuses and means.

It is a block diagram which shows one Embodiment of the monitoring system to which this invention is applied. It is a figure which shows the detection unit of a monitoring area | region. It is a figure which shows the structural example of the function of the intrusion detection apparatus of FIG. It is a block diagram which shows the structural example of the function implement | achieved by CPU of the intrusion detection apparatus of FIG. It is a flowchart explaining the reference | standard space creation process performed by the monitoring system of FIG. It is a figure which shows the example of the pattern of the illumination or image | video projected on a monitoring area | region in order to calculate a correlation matrix. It is a figure which shows another example of the pattern of the illumination or image | video projected on a monitoring area | region in order to calculate a correlation matrix. It is a figure which shows another example of the pattern of the illumination or image | video projected on a monitoring area | region in order to calculate a correlation matrix. It is a figure which shows another example of the pattern of the illumination or image | video projected on a monitoring area | region in order to calculate a correlation matrix. It is a flowchart explaining the monitoring process performed by the monitoring system of FIG. It is a figure which shows the example of the intrusion to a monitoring area | region. It is a figure which shows another example of the intrusion to a monitoring area | region. It is a figure which shows another example of the intrusion to a monitoring area | region. It is a figure explaining an example of how to determine a detection threshold. It is a figure which shows the monitoring area | region used for the comparison experiment of Mahalanobis distance. It is a table | surface which shows the example of the difference data detected from the monitoring area | region of FIG. It is a table | surface which shows the result of the comparative experiment of Mahalanobis distance. It is a graph which shows the result of the comparative experiment of Mahalanobis distance. It is a figure which shows the example of the range which considers the correlation between detection units. It is a figure which shows another example of the range which considers the correlation between detection units. It is a figure which shows another example of the range which considers the correlation between detection units. It is a figure which shows the example of a monitoring area | region. It is a figure which shows the example of the sequence number of a detection unit.

Explanation of symbols

  DESCRIPTION OF SYMBOLS 1 Monitoring system, 11 Imaging device, 12 Intrusion detection device, 21 Detection unit, 31 CPU, 32 ROM, 35 Output part, 36 Recording part, 38 drive, 51 Removable media, 71 Image pre-processing part, 72 Reference space data calculation part 73 data management unit, 74 change area detection unit, 75 distance calculation unit, 76 determination unit, 81 noise filter unit, 82 image information acquisition unit, 83 inter-frame difference calculation unit

Claims (7)

Based on monitoring data that is output from a sensor that monitors the monitoring area and that is composed of unit data corresponding to each detection unit that divides the monitoring area into a plurality of areas, the intrusion of the monitoring area is monitored. In the monitoring device,
The average value and standard deviation of the first unit data that is the unit data constituting the first monitoring data that is the monitoring data output from the sensor in a state where the object has not entered the monitoring area, Alternatively, the first difference unit data of the first difference monitoring data configured by the first difference unit data that is a difference at predetermined time intervals of the first unit data corresponding to the same detection unit. The average value and the standard deviation are calculated, and the output from the sensor is performed in a state where the object does not enter the monitoring area and a video or illumination of a predetermined pattern is projected on the monitoring area. The first correlation matrix indicating the correlation between the second unit data that is the unit data constituting the second monitoring data that is the monitoring data, or the same detection A second correlation data between the second difference unit data of the second difference monitoring data configured by the second difference unit data which is a difference at a predetermined time interval of the second unit data corresponding to the unit. Reference space calculating means for calculating a correlation matrix of 2;
When monitoring the intrusion of the object into the monitoring area, a first space between the first space constituted by the first monitoring data and the third monitoring data which is the monitoring data output from the sensor. A Mahalanobis distance of 1 is calculated using an average value and standard deviation of the first unit data and the first correlation matrix, or a second space configured by the first difference monitoring data And a third difference unit data that is a difference for each time interval of the third unit data that is the unit data constituting the third monitoring data corresponding to the same detection unit. Mahalanobis distance calculating means for calculating a second Mahalanobis distance between the difference monitoring data using an average value and standard deviation of the first difference unit data and the second correlation matrix;
And a determination unit that determines whether the object has entered the monitoring area based on the first Mahalanobis distance or the second Mahalanobis distance. Standardized unit data obtained by standardizing the third unit data using an average value and standard deviation of the first unit data corresponding to the same detection area as the third unit data, or the third difference Based on standardized difference unit data obtained by standardizing unit data using an average value and standard deviation of the first difference unit data corresponding to the same detection area as the third difference unit data, in the monitoring area And further comprising detection means for detecting one or more change areas of a predetermined size that are likely to have intruded,
The monitoring apparatus according to claim 1 , wherein the Mahalanobis distance calculating unit calculates the first Mahalanobis distance or the second Mahalanobis distance in the change region. The reference space calculation unit is configured to correspond to the second detection unit corresponding to the detection unit in a frame-shaped region in the monitoring region that is in contact with an outer periphery of the monitoring region, among the correlation coefficients constituting the correlation matrix. While calculating only the correlation coefficient between unit data or between said 2nd difference unit data, other correlation coefficients are set to 0,
The monitoring apparatus according to claim 1 , wherein the Mahalanobis distance calculating unit calculates the first Mahalanobis distance or the second Mahalanobis distance in the frame-shaped region. The reference space calculation means includes, for each of the detection units, among the correlation coefficients constituting the correlation matrix, the second unit data or the second difference unit data corresponding to the detection unit, and the detection. While calculating only the correlation coefficient between the second unit data or the second difference unit data corresponding to the detection unit included in an area of a predetermined size centered on the unit, The monitoring apparatus according to claim 1 , wherein the correlation coefficient is 0. Based on monitoring data that is output from a sensor that monitors the monitoring area and that is composed of unit data corresponding to each detection unit that divides the monitoring area into a plurality of areas, the intrusion of the monitoring area is monitored. In the monitoring method of the monitoring device,
The average value and standard deviation of the first unit data that is the unit data constituting the first monitoring data that is the monitoring data output from the sensor in a state where the object has not entered the monitoring area, Alternatively, the first difference unit data of the first difference monitoring data configured by the first difference unit data that is a difference at predetermined time intervals of the first unit data corresponding to the same detection unit. The average value and the standard deviation are calculated, and the output from the sensor is performed in a state where the object does not enter the monitoring area and a video or illumination of a predetermined pattern is projected on the monitoring area. The first correlation matrix indicating the correlation between the second unit data that is the unit data constituting the second monitoring data that is the monitoring data, or the same detection A second correlation data between the second difference unit data of the second difference monitoring data configured by the second difference unit data which is a difference at a predetermined time interval of the second unit data corresponding to the unit. A reference space calculating step of calculating a correlation matrix of 2;
When monitoring the intrusion of the object into the monitoring area, a first space between the first space constituted by the first monitoring data and the third monitoring data which is the monitoring data output from the sensor. A Mahalanobis distance of 1 is calculated using an average value and standard deviation of the first unit data and the first correlation matrix, or a second space configured by the first difference monitoring data And a third difference unit data that is a difference for each time interval of the third unit data that is the unit data constituting the third monitoring data corresponding to the same detection unit. A Mahalanobis distance calculation step for calculating a second Mahalanobis distance from the difference monitoring data using the average value and standard deviation of the first difference unit data and the second correlation matrix. And,
And a determination step of determining whether the object has entered the monitoring area based on the first Mahalanobis distance or the second Mahalanobis distance. Based on monitoring data that is output from a sensor that monitors the monitoring area and that is composed of unit data corresponding to each detection unit that divides the monitoring area into a plurality of areas, the intrusion of the monitoring area is monitored. A program for monitoring processing of a monitoring device,
The average value and standard deviation of the first unit data that is the unit data constituting the first monitoring data that is the monitoring data output from the sensor in a state where the object has not entered the monitoring area, Alternatively, the first difference unit data of the first difference monitoring data configured by the first difference unit data that is a difference at predetermined time intervals of the first unit data corresponding to the same detection unit. The average value and the standard deviation are calculated, and the output from the sensor is performed in a state where the object does not enter the monitoring area and a video or illumination of a predetermined pattern is projected on the monitoring area. The first correlation matrix indicating the correlation between the second unit data that is the unit data constituting the second monitoring data that is the monitoring data, or the same detection A second correlation data between the second difference unit data of the second difference monitoring data configured by the second difference unit data which is a difference at a predetermined time interval of the second unit data corresponding to the unit. A reference space calculating step of calculating a correlation matrix of 2;
When monitoring the intrusion of the object into the monitoring area, a first space between the first space constituted by the first monitoring data and the third monitoring data which is the monitoring data output from the sensor. A Mahalanobis distance of 1 is calculated using an average value and standard deviation of the first unit data and the first correlation matrix, or a second space configured by the first difference monitoring data And a third difference unit data that is a difference for each time interval of the third unit data that is the unit data constituting the third monitoring data corresponding to the same detection unit. A Mahalanobis distance calculation step for calculating a second Mahalanobis distance from the difference monitoring data using the average value and standard deviation of the first difference unit data and the second correlation matrix. And,
A computer-readable program comprising: a determination step of determining whether the object has entered the monitoring area based on the first Mahalanobis distance or the second Mahalanobis distance. Recording medium. Based on monitoring data that is output from a sensor that monitors the monitoring area and that is composed of unit data corresponding to each detection unit that divides the monitoring area into a plurality of areas, the intrusion of the monitoring area is monitored. A program for causing a monitoring processing computer of a monitoring device to perform monitoring processing,
The average value and standard deviation of the first unit data that is the unit data constituting the first monitoring data that is the monitoring data output from the sensor in a state where the object has not entered the monitoring area, Alternatively, the first difference unit data of the first difference monitoring data configured by the first difference unit data that is a difference at predetermined time intervals of the first unit data corresponding to the same detection unit. The average value and the standard deviation are calculated, and the output from the sensor is performed in a state where the object does not enter the monitoring area and a video or illumination of a predetermined pattern is projected on the monitoring area. The first correlation matrix indicating the correlation between the second unit data that is the unit data constituting the second monitoring data that is the monitoring data, or the same detection A second correlation data between the second difference unit data of the second difference monitoring data configured by the second difference unit data which is a difference at a predetermined time interval of the second unit data corresponding to the unit. A reference space calculating step of calculating a correlation matrix of 2;
When monitoring the intrusion of the object into the monitoring area, a first space between the first space constituted by the first monitoring data and the third monitoring data which is the monitoring data output from the sensor. A Mahalanobis distance of 1 is calculated using an average value and standard deviation of the first unit data and the first correlation matrix, or a second space configured by the first difference monitoring data And a third difference unit data that is a difference for each time interval of the third unit data that is the unit data constituting the third monitoring data corresponding to the same detection unit. A Mahalanobis distance calculation step for calculating a second Mahalanobis distance from the difference monitoring data using the average value and standard deviation of the first difference unit data and the second correlation matrix. And,
And a determination step of determining whether the object has entered the monitoring area based on the distance of the first Mahalanobis or the distance of the second Mahalanobis.

Images