JP6084810B2 - Tracking processing apparatus and tracking processing method - Google Patents

Description

  The present invention relates to a tracking processing device and a tracking processing method used for target tracking processing.

  2. Description of the Related Art A tracking device that tracks a target (TT: Target Tracking) using a radar signal or the like reflected from the target is known (see, for example, Patent Documents 1 to 3). The radar signal is obtained by a scanning operation using a radar antenna. The radar signal includes data for specifying positions of a plurality of targets. The obtained radar signal is output to the tracking device. The tracking device specifies the position of the target being tracked using the above data. The tracking device estimates the movement of the target being tracked using information such as the specified position.

  The apparatus described in Patent Literature 1 acquires detection data by observing positions and the like of a plurality of targets. This device calculates the track of the tracking target using detection data or the like, and then holds the track data. In addition, this device associates detection data with wakes.

Japanese Patent No. 3044296 ([Claim 1]) Japanese Patent No. 4488185 Japanese Utility Model Publication No. 5-23172

  By the way, the tracking device periodically refers to the motion state of the target being tracked. Then, the tracking device estimates the motion state of the tracking target at the latest time using the past calculation result and the information included in the radar signal at the latest time. Specifically, for example, the tracking device calculates a position at the latest time as an estimated position based on a past calculation result for a target being tracked. Next, the tracking device determines that the target closest to the estimated position among the plurality of targets specified by the radar signal at the latest time point is the tracking target. The tracking device estimates the movement of the tracking target at the latest time point using the observed data for the tracking target.

  However, there is a case where a wave or clutter generated by the navigation of the tracking vessel is generated in the vicinity of the estimated position. In this case, the tracking device may erroneously determine the wave or clutter as a tracking target. When such an erroneous determination occurs, the tracking device performs a tracking process for a wave or clutter that is different from the original target, and cannot perform an accurate tracking process.

  In addition, a prediction error and an observation error occur with respect to the position of the tracking target. The prediction error refers to a deviation amount between the estimated position and the actual position of the tracking target in the tracking process. Further, the observation error refers to the amount of deviation between the position specified based on the radar signal and the actual position for the tracking target. Therefore, it is conceivable that the prediction error is derived according to a preset model, and the observation error is derived according to a preset model. However, in such a configuration, it is necessary to model each of the prediction error and the observation error, and it is very difficult to create a model that can withstand practical use.

  In view of the above circumstances, an object of the present invention is to provide a tracking processing device and a tracking processing method capable of reliably and easily detecting a tracking target from among a plurality of targets.

(1) In order to solve the above-described problem, a tracking processing device according to an aspect of the present invention is a tracking processing device for tracking a target selected as a tracking target, and includes a detection unit and likelihood determination. A movement estimation unit, and an area determination unit . The detection unit is configured to detect feature information including representative point information for a plurality of targets. The likelihood determining unit is configured to calculate a likelihood corresponding to the representative point to be tracked for at least some of the plurality of representative points. The motion estimation unit is configured to estimate a motion state of the tracking target using the feature information of the target specified as the tracking target. The area determination unit is configured to determine an area. The likelihood determination unit specifies the tracking target from among the plurality of targets based on the calculated likelihood.

Furthermore , the detection unit detects, as additional feature information, feature information other than the representative point information for at least some of the plurality of targets. The likelihood determination unit calculates a likelihood corresponding to the feature information of the tracking target for the additional feature information, and determines the tracking target from a plurality of the targets based on the likelihood. Identify. The detection unit detects an echo signal that is a reflected wave from the plurality of targets including the tracking target with respect to a transmission signal, detects an echo image specified by the echo signal, and detects the echo image Are detected as additional feature information. The area determination unit determines whether the area of the echo image to be tracked exceeds a predetermined threshold value. When the area is equal to or less than a predetermined threshold, the motion estimation unit calculates the motion state using the feature information of the tracking target at the time of estimating the motion state, and the area When is exceeding a predetermined threshold value, the motion state is calculated without using the feature information of the tracking target at the time of calculating the motion state.

( 2 ) More preferably, the said detection part detects each said feature information using the echo signal which is the reflected wave in each said target with respect to a transmission signal. The additional feature information includes at least the shape information of the echo image specified by the echo signal, the level of the echo signal, the information specifying the surrounding state of each target, and the Doppler shift amount related to the echo signal. Including one.

( 3 ) Preferably, each said feature information is information represented by a numerical value, and the said likelihood determination part calculates each said likelihood using a probability density function for every said feature information.

( 4 ) More preferably, the probability density function is set for each feature information. The value used as the average value in each probability density function is a value estimated in advance before the likelihood is calculated for the corresponding feature information. The likelihood determining unit uses the corresponding feature information as an independent variable in each probability density function. The likelihood determination unit calculates a dependent variable that depends on the independent variable as the likelihood.

( 5 ) More preferably, the likelihood determination unit calculates an overall likelihood by multiplying the likelihoods of the same target by each other. Moreover, the said likelihood determination part specifies the said target with the said total likelihood highest among each said target as the said tracking object.

( 6 ) More preferably, the said area determination part is comprised so that the said area may be determined using a probability density function. The value used as the average value of the probability density variable is a value estimated in advance for the area before being determined by the area determination unit. The predetermined threshold is set based on the average.

( 7 ) In order to solve the above-described problem, a tracking processing method according to an aspect of the present invention is a tracking processing method for tracking a target selected as a tracking target, including a detection step and likelihood determination. A step, a motion estimation step, and an area determination step . In the detection step, feature information including representative point information is detected for a plurality of targets. In the likelihood determination step, the likelihood corresponding to the representative point to be tracked is calculated for at least some of the representative points. In the motion estimation step, the motion state of the tracking target is estimated using the feature information of the target specified as the tracking target. In the area determining step, the area is determined. In the likelihood determining step, the tracking target is specified from among the plurality of targets based on the calculated likelihood. Furthermore, in the detection step, feature information other than the representative point information is detected as additional feature information for at least some of the plurality of targets. In the likelihood determination step, for the additional feature information, a likelihood corresponding to the feature information of the tracking target is calculated, and the tracking target is selected from a plurality of the targets based on the likelihood. Identify. In the detection step, an echo signal that is a reflected wave at the plurality of targets including the tracking target with respect to the transmission signal is detected, an echo image specified by the echo signal is detected, and the echo image is detected Are detected as additional feature information. In the area determining step, it is determined whether or not the area of the echo image to be tracked exceeds a predetermined threshold value. In the motion estimation step, when the area is equal to or smaller than a predetermined threshold, the motion state is calculated using the feature information of the tracking target at the time of estimating the motion state, and the area When is exceeding a predetermined threshold value, the motion state is calculated without using the feature information of the tracking target at the time of calculating the motion state.

  According to the present invention, it is possible to provide a tracking processing device and a tracking processing method that can reliably and easily detect a tracking target from among a plurality of targets.

1 is a block diagram of a radar apparatus including a tracking processing apparatus according to an embodiment of the present invention. It is a typical top view for demonstrating the relationship between the own ship and a target object echo image. It is a data list for demonstrating the data extracted regarding a target echo image (target). It is a typical top view which shows the target echo image detected by the echo detection part. It is a flowchart for demonstrating an example of the flow of a process about the probability density function update in a function update part. It is a graph which shows an example of the function updated in a function update part about a representative point. It is a graph which shows an example of the function updated by the function update part about the area of a tracking target. It is a flowchart for demonstrating an example of the flow of a process in a likelihood calculation part. It is a flowchart for demonstrating an example of the flow of a process in an area determination part. It is a mimetic diagram for explaining the case where fusion has not occurred about a target echo image and other target echo images. It is a mimetic diagram for explaining the case where fusion has occurred about a target echo image and other target echo images. It is a schematic diagram which shows the state in which excessive blur has arisen in the target object echo image. It is a flowchart for demonstrating an example of the flow of a process by the exercise | movement estimation part. It is a schematic diagram for demonstrating the process in a motion estimation part. It is a schematic diagram for demonstrating the process in a motion estimation part.

  Hereinafter, embodiments for carrying out the present invention will be described with reference to the drawings. The present invention can be widely applied as a tracking processing apparatus that tracks a target selected as a tracking target. Hereinafter, a target set as a tracking target is referred to as a “tracking target”. In the following, the same or corresponding parts in the drawings are denoted by the same reference numerals, and description thereof will not be repeated.

  FIG. 1 is a block diagram of a radar apparatus 1 including a tracking processing apparatus 3 according to an embodiment of the present invention. The radar apparatus 1 according to the present embodiment is a marine radar provided in a vessel such as a fishing boat. The radar apparatus 1 is mainly used for detecting a target such as another ship. Further, the radar apparatus 1 is configured to be able to track a target selected as a tracking target. The radar apparatus 1 is configured to be able to track a plurality of tracking targets at the same time. The radar apparatus 1 is configured to estimate the motion state of the tracking target. In the present embodiment, the radar apparatus 1 calculates an estimated velocity vector of the tracking target as the motion state. The estimated speed vector is a vector indicating the traveling direction and the traveling speed estimated for the tracking target. The radar apparatus 1 displays the estimated velocity vector of the tracking target on the screen. Hereinafter, a ship provided with the radar device 1 is referred to as “own ship”.

  As shown in FIG. 1, the radar device 1 includes an antenna unit 2, a tracking processing device 3, and a display 4.

  The antenna unit 2 includes an antenna 5, a receiving unit 6, and an A / D conversion unit 7.

  The antenna 5 is a radar antenna capable of transmitting a pulsed radio wave having strong directivity. The antenna 5 is configured to receive an echo signal that is a reflected wave from a target. That is, the echo signal of the target is a reflected wave at the target with respect to the transmission signal from the antenna 5. The radar apparatus 1 measures the time from when a pulsed radio wave is transmitted to when an echo signal is received. Thereby, the radar apparatus 1 can detect the distance r to the target. The antenna 5 is configured to be able to rotate 360 ° on a horizontal plane. The antenna 5 is configured to repeatedly transmit and receive radio waves while changing the transmission direction of pulsed radio waves (changing the antenna angle). With the above configuration, the radar apparatus 1 can detect a target on a plane around the ship over 360 °.

  In the following description, an operation from transmission of a pulsed radio wave to transmission of the next pulsed radio wave is referred to as “sweep”. The operation of rotating the antenna 360 ° while transmitting / receiving radio waves is called “scan”. In the following, a certain scan is referred to as an “n scan”, and a scan immediately before the n scan is referred to as an “n−1 scan”. Similarly, m scans before n scans are referred to as “nm scans”. Note that n and m are natural numbers. In the following, the operation of the radar apparatus 1 at the n-scan time point will be described unless the operation is particularly described.

  The receiving unit 6 detects and amplifies the echo signal received by the antenna 5. The reception unit 6 outputs the amplified echo signal to the A / D conversion unit 7. The A / D converter 7 samples an analog echo signal and converts it into digital data (echo data) consisting of a plurality of bits. Here, the echo data includes data for specifying the intensity (signal level) of the echo signal received by the antenna 5. The A / D converter 7 outputs the echo data to the tracking processing device 3.

  The tracking processing device 3 is configured to identify a target selected from a plurality of targets as a tracking target and to perform tracking processing of the tracking target. More specifically, the tracking processing device 3 is configured to calculate an estimated speed vector of the tracking target, an estimated position of the tracking target, and the like.

  The tracking processing device 3 is configured using hardware including a CPU, a RAM, a ROM (not shown), and the like. The tracking processing device 3 is configured by using software including a tracking processing program stored in the ROM.

  The tracking processing program is a program for causing the tracking processing device 3 to execute the tracking processing method according to the present invention. The hardware and software are configured to operate in cooperation. Thereby, the tracking processing device 3 can function as the sweep memory 8, the signal processing unit 9, the echo detection unit 10, the tracking processing unit 11, and the like. The tracking processing device 3 is configured to perform processing described below for each scan.

  The tracking processing device 3 includes a signal processing unit 9, an echo detection unit (detection unit) 10, and a tracking processing unit 11.

  The signal processing unit 9 removes interference components included in the echo data and unnecessary waveform data by performing filter processing or the like. The signal processing unit 9 is configured to detect feature information of echo data related to the target echo image. The signal processing unit 9 outputs the processed echo data to the echo detection unit 10.

  The echo detector 10 is configured to detect a target echo image and detect feature information of echo data related to the target echo image. That is, the echo detection unit 10 includes a target echo image detection unit and a feature information extraction unit. The signal processing unit 9 and the echo detection unit 10 constitute a detection unit for detecting target feature information.

  The echo detector 10 obtains the distance r to the position corresponding to the echo data based on the read address when the echo data is read from the signal processor 9. Further, data indicating which direction the antenna 5 is currently facing (antenna angle θ) is output from the antenna 5 to the echo detector 10. With the above configuration, when the echo detection unit 10 reads the echo data, the echo detection unit 10 can acquire the position corresponding to the echo data with polar coordinates of the distance r and the antenna angle θ.

  The echo detector 10 is configured to detect whether or not a target exists at a position corresponding to the echo data. For example, the echo detector 10 determines the signal level at the position corresponding to the echo data, that is, the signal intensity. The echo detector 10 determines that a target exists at a position where the signal level is equal to or higher than a predetermined threshold level value.

  Next, the echo detector 10 detects a range where the target is present. For example, the echo detection unit 10 detects a group of areas where the target is present as an area where the target echo image is present. In this way, the echo detector 10 detects the target echo image based on the echo data. The outline shape of the target echo image substantially matches the outline shape of the target. However, due to noise included in the echo data, the outline shape of the target echo image is slightly different from the outline shape of the target. Next, the echo detector 10 extracts feature information related to the target echo image using the echo data.

  FIG. 2 is a schematic plan view for explaining the relationship between the ship 100 and the target echo image 120. In FIG. 2, the target echo image 120 is illustrated as a rectangular image. In FIG. 2, a target 130 specified by the target echo image 120 is shown. In FIG. 2, the outline shape of the target 130 is displayed in a state that matches the target echo image 120.

  As shown in FIGS. 1 and 2, in the polar coordinate system, a straight line distance from the ship position M1 is indicated as a distance r with respect to the ship position M1 as the position of the ship 100. Is shown as an angle θ. In the present embodiment, the ship position M <b> 1 corresponds to the position of the antenna 5. When extracting the representative point P of the target echo image 120, the echo detection unit 10 uses the partial image 110 of the ring-shaped portion centered on the ship position M1. The image 110 is an image of a region surrounded by the first straight line 111, the second straight line 112, the first arc 113, and the second arc 114.

  The first straight line 111 is a straight line passing through the point closest to the own ship position M1 in the rear edge 120a of the target echo image 120 and the own ship position M1. The second straight line 112 is a straight line passing through the point closest to the ship position M1 in the front edge 120b of the target echo image 120 and the ship position M1. The first arc 113 is an arc passing through the portion 120c closest to the ship position M1 in the target echo image 120. The center of curvature of the first arc 113 is the ship position M1. The second arc 114 is an arc passing through the portion 120d farthest from the ship position M1 in the target echo image 120. The second arc 114 is concentric with the first arc 113.

  FIG. 3 is a data list for explaining data extracted with respect to the target echo image 120 (target 130). As shown in FIGS. 2 and 3, in the present embodiment, the signal processing unit 9 and the echo detection unit 10 cooperate with each other to obtain the following 12 pieces of data for the target echo image 120 (target 130): Extracted as feature information data. That is, the echo detector 10 extracts the following 12 pieces of text data based on the echo data and the image data of the target echo image 120. In the present embodiment, the twelve text data are flag data 201, distance rp data 202, end angle θe data 203, angular width θw data 204, and leading edge distance rn data 205. The data 206 of the last edge distance rf, the data 207 of the area ar (shape information), the coordinate data 208 of the representative point P, the data 209 of the echo level ec, and the adjacent distance (specify the state around the target) Information) ad data 210, Doppler shift amount ds data 211, and time tm data 212.

  The above flag is a flag for specifying whether or not the target echo image 120 is in a predetermined state, for example. This flag is configured to be set to “1” or “0”.

  The distance rp is a linear distance from the ship position M1 to the representative point P of the target echo image 120. In the present embodiment, the representative point P is a centroid point of the image 110. The end angle θe is the antenna angle θ described above at the time when the detection of the target echo image 120 is completed. The angle width θw is the width of the target echo image 120 in the angular direction around the ship position M1. The angle width θw is also an angle formed by the first straight line 111 and the second straight line 112. The forefront edge distance rn is the distance between the portion 120c of the target echo image 120 and the ship position M1. The last edge distance rf is the distance between the portion 120d of the target echo image 120 and the ship position M1. The area ar is an area in the partially shaped image 110 of the ring-shaped portion, and is treated as the area of the target echo image 120 in the present embodiment.

  The echo level ec indicates the intensity of an echo signal that identifies the target echo image 120. This intensity may be a peak intensity of an echo signal specifying the target echo image 120, or may be an average value of the intensity of the echo signal. In the present embodiment, the adjacent distance ad is, for example, a distance between two adjacent target echo images 120. The Doppler shift amount ds is, for example, the difference between the frequency of the pulse signal radiated from the antenna 5 and the frequency of the echo signal reflected by the target 130 specified by the target echo image 120. Based on the Doppler shift amount ds, the relative speed between the target 130 specified by the target echo image 120 and the ship 100 can be obtained. Time tm is the time when the target echo image 120 is detected. Note that the data of the echo image 120 may include a spare data area. In the present embodiment, the signal processing unit 9 and the echo detection unit 10 extract the 12 pieces of feature information for each target echo image 120. Of these twelve pieces of feature information, feature information other than the position of the representative point P is an example of “additional feature information” in the present invention. Further, these 12 pieces of feature information are all information represented by numerical values.

  An example of a plurality of target echo images 120 detected by the echo detector 10 is shown in FIG. FIG. 4 is a schematic plan view showing the target echo image 120 detected by the echo detector 10. In FIG. 4, as an example, four target echo images 120 (121, 122, 123, 124) at the time of n scanning are shown. In FIG. 4, the shape of the target echo image 120 (121, 122, 123, 124) matches the shape of the target 130 (131, 132, 133, 134), respectively.

  The target 131 specified by the target echo image 121, the target 132 specified by the target echo image 122, and the target 133 specified by the target echo image 123 are, for example, small ships. The target 134 specified by the target echo image 124 is, for example, a large ship. The echo detection unit 10 includes the representative point P1 (n) of the target echo image 121, the representative point P2 (n) of the target echo image 122, and the target echo image 123 as the representative point P at the n scan time point. The representative point P3 (n) and the representative point P4 (n) of the target echo image 124 are detected. Hereinafter, a case where the target 131 is the tracking target 140 will be described as an example.

  As shown in FIGS. 1 to 4, the echo detection unit 10 outputs feature information data 201 to 212 for each target echo image 120 to the tracking processing unit 11.

  The tracking processing unit 11 is configured to identify the tracking target 140 from among the plurality of targets 120 and perform the tracking process of the tracking target 140. The tracking target 140 is selected by the operator based on, for example, symbols indicating the plurality of targets 130 displayed on the display 4. The selection command of the tracking target 140 by the operator is issued, for example, when the operator operates an operating device (not shown). In the present embodiment, the tracking processing unit 11 is configured to perform processing on the basis of the XY coordinate system unless otherwise specified.

  The tracking processing unit 11 is configured to calculate an estimated velocity vector V1 (n) of the tracking target 140 at the n scanning time point (latest scanning time point). The tracking processing unit 11 is configured to display the estimated speed vector V1 (n) on the display 4.

  The tracking processing unit 11 includes a feature information memory 12, a selection unit 13, an association unit (likelihood determination unit) 14, an area determination unit 15, and a motion estimation unit 16.

  The feature information memory 12 is configured to store data output from the signal processing unit 9 and the echo detection unit 10. The feature information memory 12 is configured to store data related to functions f1 to f5 described later. The feature information memory 12 stores feature information data 201 to 212 for all target echo images 120 at each time point from the (n-T) scan time point to the n scan time point. The constant T is a preset value, for example, about several tens.

  The sorting unit 13 is configured to set the sorting area S1 (n). The selection area S1 (n) is an area for selecting a representative point (tracking representative point) PR (n) of the tracking target 140 at the time of n scanning. The selection unit 13 is connected to the echo detection unit 10, the association unit 14, and the motion estimation unit 16.

  The selection unit 13 receives the coordinate data of the representative points P (P1 (n) to P4 (n)) of each target echo image 120 at the n-scan time point from the echo detection unit 10. In addition, the selection unit 13 receives data of the estimated speed vector V1 (n−1) of the tracking target 140 with the (n−1) scan time as the calculation reference time. In the present embodiment, the selection unit 13 receives data of the estimated speed vector V1 (n−1) from the motion estimation unit 16.

  Next, the selection unit 13 detects the estimated position PRe (n−1) of the tracking target 140. The estimated position PRe (n-1) is a position specified by the estimated speed vector V1 (n-1). The estimated position PRe (n−1) is a position where the representative point PR (n) of the tracking target 140 is estimated to arrive at n scan points. The selection unit 13 sets a selection region S1 (n) having a predetermined radius around the estimated position PRe (n-1). FIG. 4 illustrates a state where the representative points P1 (n) and P2 (n) of the target echo images 121 and 122 are included in the selection area S1. The selecting unit 13 outputs the coordinate data 208 to the associating unit 14 for each of the representative points P1 (n) and P2 (n).

  The associating unit 14 performs a process of associating the estimated position PRe (n−1) with the representative point PR (n) of the tracking target 140 at the time of n scanning. In the present embodiment, the associating unit 14 sets the likelihood Lh100 corresponding to the representative point PR (n) of the tracking target 140 for each of the representative points P1 (n) and P2 (n) in the selection region S1 (n). , Lh200 is calculated. The associating unit 14 specifies the tracking target 140 representative point PR (n) at the n-scan time point based on the likelihoods Lh100 and Lh200.

  The associating unit 14 includes a function updating unit 21, a likelihood calculating unit 22, and a likelihood comparing unit 23.

  The function update unit 21 is configured to update a function used for the association process. In this embodiment, a probability density function (Gauss function, normal distribution function) is used as this function. In the present embodiment, the probability density function is a function having an average and a variance. In the present embodiment, the probability density function is updated based on past tracking results.

  In the present embodiment, as the feature information used for the probability density function, the position information of the representative point PR (n), the area ar information about the tracking target 140, and the echo level ec information about the tracking target 140, Information on the adjacent distance ad for the tracking target 140 and information on the Doppler shift amount ds for the tracking target 140 are included. In the present embodiment, one probability density function pdf is set.

  The probability density function pdf includes a function f1 of the dimension of the representative point PR, a function f2 of the area ar, a function f3 of the echo level ec, a function f4 of the adjacent distance ad, and a Doppler shift amount. It is a function specified by the function f5 of the dimension of ds. Each of the functions f1 to f5 is a probability density function, and the probability density function pdf is a multidimensional probability density function. The function updating unit 21 updates the probability density function pdf by updating the functions f1 to f5. In the following description, the functions f1 to f5 may be simply referred to as a function f when collectively referred to.

  Below, an example of the flow of the process which updates the probability density function pdf in the function update part 21 is demonstrated. FIG. 5 is a flowchart for explaining an example of a processing flow for updating the probability density function pdf in the function updating unit 21. The tracking processing device 3 reads and executes each step of the flowchart shown below from a memory (not shown). This program can be installed externally. The installed program is distributed in a state stored in a recording medium, for example. The processing in the tracking processing device 3 is performed at each scanning time point. In the present embodiment, when a flowchart is described, the drawings other than the flowchart are also referred to as appropriate.

  In the present embodiment, the function update unit 21 performs maximum likelihood estimation processing. Specifically, as illustrated in FIG. 5, the function update unit 21 selects feature information for updating the probability density function pdf (step S101). For example, the function updating unit 21 first selects position information of the representative point PR. That is, the function update unit 21 updates the function f1 related to the position of the representative point PR.

FIG. 6 is a graph showing an example of the function f1 updated by the function update unit 21 for the representative point PR. The function f1 has mean μ11 and μ12 and variances σ11 ² and σ12 ² . In the present embodiment, the average corresponds to an estimated value in each function f. That is, in each function f of the present embodiment, the value used as the average value is a value estimated in advance before the likelihood Lh is calculated for the corresponding feature information. The x axis in the graph of FIG. 6 indicates the x coordinate, and the y axis indicates the y coordinate. In addition, the vertical axis (z axis) of the graph of FIG. 6 indicates the likelihood Lh.

  The function updater 21 calculates the averages μ11 and μ12 in the function f1 (step S102). Specifically, the function updating unit 21 calculates the average μ11 of the x coordinate and the average μ12 of the y coordinate for the representative point PR (n−1) of the tracking target 140.

尚、Δｄｘ_ｉは、任意のｉスキャン時点における、代表点ＰＲ（ｉ）のｘ座標の予測誤差である。より具体的には、Δｄｘ_ｉは、ｉスキャン時点における、代表点ＰＲ（ｉ）のｘ座標の（観測値−予測値）である。この場合の観測値とは、ｉスキャン時点でエコー検出部１０によって検出された、代表点ＰＲ（ｉ）のｘ座標である。また、予測値とは、（ｉ−１）スキャン時点において運動推定部１６で予測された、代表点ＰＲ（ｉ）のｘ座標である。したがって、平均μ１１は、（ｎ−Ｔ）スキャン時点から（ｎ−１）スキャン時点までの各スキャン時点における、代表点ＰＲのｘ座標の予測誤差を平均した値である。 In the present embodiment, the average μ11 is calculated by the following formula (1).

Δdx _i is a prediction error of the x coordinate of the representative point PR (i) at an arbitrary i scan time. More specifically, Δdx _i is (observed value−predicted value) of the x coordinate of the representative point PR (i) at the time of i scan. The observed value in this case is the x coordinate of the representative point PR (i) detected by the echo detector 10 at the time of i scan. The predicted value is the x coordinate of the representative point PR (i) predicted by the motion estimation unit 16 at (i-1) scan time. Therefore, the average μ11 is a value obtained by averaging the prediction errors of the x coordinate of the representative point PR at each scan time from the (n−T) scan time to the (n−1) scan time.

  The average μ12 is calculated in the same manner as the average μ11. That is, the average μ12 is a value obtained by averaging the prediction errors of the y coordinate of the representative point PR at each scan time from the (n−T) scan time to the (n−1) scan time.

Next, the function updater 21 calculates variances σ11 ² and σ12 ² of the function f1 (step S103). That is, the function update unit 21 calculates the variance σ11 ^{2 in} the x-axis direction and the variance σ12 ² in the y-axis direction of the function f1.

In the present embodiment, dispersion Shiguma11 ² is calculated by the following equation (2).

Distributed Shiguma12 ² is calculated in the same manner as the average Myu11. That is, the dispersion Shiguma12 ² replaces the Derutadx _i of formula (2) in Derutady _i, and is calculated by the equation obtained by replacing the μ11 to Myu12. As described above, the function update unit 21 calculates the averages μ11 and μ12 and the variances σ11 ² and σ12 ² . The function updating unit 21 updates the function f1 by calculating the averages μ11 and μ12 and the variances σ11 ² and σ12 ² .

  Next, the function update unit 21 determines whether all of the functions f1 to f5 have been updated (step S104). In this case, the functions f2, f3, f4f, and f5 have not been updated yet (NO in step S104). Therefore, the function update unit 21 repeats the processes of steps S101 to S104.

  Specifically, the function updating unit 21 selects feature information for updating the probability density function pdf (step S101). In this case, for example, the function updating unit 21 selects the area ar information as the feature information. That is, the function update unit 21 updates the function f2.

FIG. 7 is a graph showing an example of the function f2 updated by the function updating unit 21 with respect to the area ar of the tracking target 140 at the n-scan time point. As shown in FIG. 7, the function f2 has an average .mu.2, variance .sigma. @ 2 ^2. The horizontal axis of the graph in FIG. 7 indicates the area ar. The vertical axis of the graph in FIG. 7 indicates the likelihood Lh.

  The function updating unit 21 calculates the average μ2 in the function f2 (Step S102). The average μ2 is calculated by the function updating unit 21 as follows, for example. That is, the function update unit 21 first refers to the feature information memory 12. As a result, the function updating unit 21 obtains the data 207 of the area ar in the target 131 set as the tracking target 140 for each scan time from the (n-1) scan time to the (n-T) scan time. refer. Then, the function updating unit 21 calculates the average value of the area ar from the (n−1) scan time to the (n−T) scan time. The function updating unit 21 sets this average value as an average μ2. The function updating unit 21 may set the area ar of the target 131 at the time of (n−1) scanning as an average μ2.

Next, the function updating unit 21 calculates the variance .sigma. @ 2 ² function f2 (step S103). The function update unit 21 performs, for example, maximum likelihood estimation processing. Thus, the function updating unit 21 calculates the variance .sigma. @ 2 ² functions f2. Thus, the function update unit 21, by calculating the average μ2 and variance .sigma. @ 2 ^2, and updates the function f2.

  Next, the function update unit 21 determines whether all of the functions f1 to f5 have been updated (step S104). In this case, the functions f3, f4, and f5 have not been updated yet (NO in step S104). Therefore, the function update unit 21 repeats the processes of steps S101 to S104.

  Specifically, the function updating unit 21 selects feature information for updating the probability density function pdf (step S101). In this case, for example, the function update unit 21 selects the echo level ec as the feature information. That is, the function update unit 21 updates the function f3. The function f3 has the same shape as the function f2, although not shown.

  The function updating unit 21 calculates the average μ3 in the function f3 (Step S102). The average μ3 of the echo level ec in this case is calculated by the function updating unit 21 as follows, for example. That is, the function update unit 21 first refers to the feature information memory 12. As a result, the function update unit 21 stores the echo level ec data of the target 131 identified as the tracking target 140 for each scan time from the (n-1) scan time to the (n-T) scan time. 209 is referred to. Then, the function updating unit 21 calculates the average value of the echo level ec from the (n−1) scan time to the (n−T) scan time. The function updating unit 21 sets this average value as an average μ3. The function updating unit 21 may set the echo level ec for the target 131 at the time of (n−1) scanning as an average μ3.

Next, the function updating unit 21 calculates the variance .sigma.3 ² function f3 (step S103). The function update unit 21 performs, for example, maximum likelihood estimation processing. Thus, the function updating unit 21 calculates the variance .sigma.3 ² functions f3. Thus, the function update unit 21, by calculating the average μ3 and variance .sigma.3 ^2, to update the function f3.

  Next, the function update unit 21 determines whether all of the functions f1 to f5 have been updated (step S104). In this case, the functions f4 and f5 are not yet set (NO in step S104). Therefore, the function update unit 21 repeats the processes of steps S101 to S104.

  Specifically, the function updating unit 21 selects feature information for updating the probability density function pdf (step S101). In this case, for example, the function updating unit 21 selects the adjacent distance ad as the feature information. That is, the function update unit 21 sets the function f4. The function f4 has the same shape as the function f2, although not shown.

  The function updating unit 21 calculates the average μ4 in the function f4 (Step S102). The average μ4 is calculated by the function updating unit 21 as follows, for example. That is, the function update unit 21 first refers to the feature information memory 12. Thereby, the function updating unit 21 sets the data of the adjacent distance ad for the target 131 identified as the tracking target 140 for each scan time point from the (n-1) scan time point to the (n-T) scan time point. 210 is referred to. Then, the function updating unit 21 calculates the average value of the adjacent distances ad from the (n−1) scan time to the (n−T) scan time. The function updating unit 21 sets this average value as an average μ4. The function updating unit 21 may set the adjacent distance ad for the target 131 at the time of (n−1) scanning as an average μ4.

Next, the function updating unit 21 calculates the variance? 4 ² function f4 (step S103). The function update unit 21 performs, for example, maximum likelihood estimation processing. Thus, the function updating unit 21 calculates the variance? 4 ² functions f4. Thus, the function update unit 21, by calculating the average μ4 and variance? 4 ^2, and updates the function f4.

  Next, the function updater 21 determines whether or not all the functions f1 to f5 have been set (step S104). In this case, the function f5 has not been updated yet (NO in step S104). Therefore, the function update unit 21 repeats the processes of steps S101 to S104.

  Specifically, the function updating unit 21 selects feature information for updating the probability density function pdf (step S101). In this case, for example, the function update unit 21 updates the Doppler shift amount ds as the feature information. That is, the function update unit 21 updates the function f5. The function f5 is the same as the function f2 although not shown.

  The function updating unit 21 calculates the average μ5 in the function f5 (Step S102). In this case, the average μ5 of the adjacent distance ad is calculated by the function updating unit 21 as follows, for example. That is, the function update unit 21 first refers to the feature information memory 12. Thereby, the function updating unit 21 sets the Doppler shift amount ds for the target 131 identified as the tracking target 140 for each scan time point from the (n-1) scan time point to the (n-T) scan time point. Refer to the data 211. Then, the function updating unit 21 calculates the average value of the Doppler shift amount ds from the (n−1) scan time to the (n−T) scan time. The function updating unit 21 sets this average value as an average μ5. The function updating unit 21 may set the Doppler shift amount ds for the target 131 at the time of (n−1) scanning as an average μ5.

Next, the function updating unit 21 calculates the variance Shiguma5 ² function f5 (step S103). The function update unit 21 performs, for example, maximum likelihood estimation processing. Thus, the function updating unit 21 calculates the variance Shiguma5 ² function f5. Thus, the function update unit 21, by calculating the average μ5 and variance Shiguma5 ^2, to update the function f5.

  Next, the function update unit 21 determines whether all of the functions f1 to f5 have been updated (step S104). In this case, all the functions f1 to f5 are updated (YES in step S104). Therefore, the function updating unit 21 outputs the data of the probability density function pdf specified by the functions f1 to f5 to the likelihood calculating unit 22 (Step S105), and ends the process. With the above configuration, the functions f1 to f5 are updated independently. The functions f1 to f5 are not fixed functions, but are functions that change at each scanning time point.

  The likelihood calculating unit 22 and the likelihood comparing unit 23 determine which of the targets 131 and 132 in the selection area S1 (n) is the tracking target 140 based on the set probability density function pdf. I do. Below, an example of the flow of the process in the likelihood calculation part 22 and the likelihood comparison part 23 is demonstrated. FIG. 8 is a flowchart for explaining an example of a processing flow in the likelihood calculating unit 22 and the likelihood comparing unit 23.

  In the flowchart illustrated in FIG. 8, the likelihood calculating unit 22 calculates the total likelihood Lh100 and 200 as the likelihood corresponding to the tracking target 140 for each target 131 and 132 in the selection region S1 (n). The overall likelihood Lh100 is a product of likelihoods Lh11, Lh12, Lh13, Lh14, and Lh15 described later. The overall likelihood Lh200 is a product of likelihoods Lh21, Lh22, Lh23, Lh24, and Lh25 described later. The likelihood comparison unit 23 identifies the tracking target 140 from the targets 131 and 132 based on the total likelihood Lh100 and 200.

  Specifically, the likelihood calculating unit 22 first selects the target 130 on which the likelihood calculation is performed (step S201). In the present embodiment, the likelihood calculating unit 22 selects the target 131 among the targets 130 (131, 132) in the selection area S1 (n). Next, the likelihood calculating unit 22 selects one of the functions f1 to f5 of the probability density function pdf (step S202).

  In this case, the likelihood calculating unit 22 first selects the function f1. Next, the likelihood calculating unit 22 calculates the likelihood Lh11 for the representative point P1 (n) of the target 131 (step S203). Specifically, the likelihood calculating unit 22 refers to the data of the function f1 and the coordinate data 208 of the representative point P1 (n) of the target 131. Then, the difference between the x coordinate of the predicted value of the representative point PR (n) estimated at the time of (n-1) scanning and the x coordinate of the representative point P1 (n) is used as an independent variable, and the function f1 (FIG. 6)). Also, the difference between the y coordinate of the predicted value of the representative point PR (n) estimated at the time of (n-1) scanning and the y coordinate of the representative point P1 (n) is used as an independent variable, and the function f1 (FIG. 6)). As a result, the dependent variable in the function f1 is calculated as the likelihood Lh11. That is, the likelihood Lh11 corresponding to the representative point PR (n) of the tracking target 140 is calculated for the representative point P1 (n) of the target 131.

  Next, the likelihood calculating unit 22 determines whether all the likelihoods Lh11 to Lh15 related to the target 131 have been calculated (step S204). In this case, the likelihoods Lh12 to Lh15 have not been calculated yet (NO in step S204). Therefore, the likelihood calculating unit 22 repeats the processes of steps S202 to S204.

  Specifically, the likelihood calculating unit 22 selects one from the functions f1 to f5 of the probability density function pdf (step S202).

  In this case, the likelihood calculating unit 22 selects the function f2. Next, the likelihood calculating unit 22 calculates the likelihood Lh12 as the likelihood corresponding to the area ar of the tracking target 140 for the area ar of the target 131 (step S203). Specifically, the likelihood calculating unit 22 refers to the data of the function f2 and the data 207 of the area ar of the target 131. Then, the likelihood calculating unit 22 substitutes the area ar of the target 131 at the time of n scanning as an independent variable into the function f2. Thereby, the dependent variable in the function f2 is calculated as the likelihood Lh12. That is, the likelihood Lh12 for the echo level ec of the target 131 is calculated.

  Next, the likelihood calculating unit 22 determines whether all the likelihoods Lh11 to Lh15 related to the target 131 have been calculated (step S204). In this case, the likelihoods Lh13 to Lh15 have not been calculated yet (NO in step S204). Therefore, the likelihood calculating unit 22 repeats the processes of steps S202 to S204.

  Specifically, the likelihood calculating unit 22 selects one from the functions f3 to f5 (step S202).

  In this case, the likelihood calculating unit 22 selects the function f3. Next, the likelihood calculating unit 22 calculates the likelihood Lh13 as the likelihood corresponding to the echo level ec of the tracking target 140 for the echo level ec of the target 131 (step S203). Specifically, the likelihood calculating unit 22 refers to the data of the function f3 and the data 209 of the echo level ec of the target 131. Then, the likelihood calculating unit 22 substitutes the echo level ec for the target 131 at the n-scan time point into the function f3 as an independent variable. Thereby, the dependent variable in the function f3 is calculated as the likelihood Lh13. That is, the likelihood Lh13 of the echo level ec for the target 131 is calculated.

  Next, the likelihood calculating unit 22 determines whether all the likelihoods Lh11 to Lh15 related to the target 131 have been calculated (step S204). In this case, the likelihoods Lh14 to Lh15 have not been calculated yet (NO in step S204). Therefore, the likelihood calculating unit 22 repeats the processes of steps S202 to S204.

  Specifically, the likelihood calculating unit 22 selects one from the functions f4 and f5 (step S202).

  In this case, the likelihood calculating unit 22 selects the function f4. Next, the likelihood calculating unit 22 calculates the likelihood Lh14 as the likelihood corresponding to the adjacent distance ad of the tracking target 140 for the adjacent distance ad (step S203). Specifically, the likelihood calculating unit 22 refers to the data of the function f4 and the data 210 of the adjacent distance ad for the target 131. Then, the likelihood calculating unit 22 substitutes the adjacent distance ad with respect to the target 131 at the time of n scans into the function f4 as an independent variable. Thereby, the dependent variable in the function f4 is calculated as the likelihood Lh14. That is, the likelihood Lh14 is calculated for the adjacent distance ad in the target 131.

  Next, the likelihood calculating unit 22 determines whether all the likelihoods Lh11 to Lh15 related to the target 131 have been calculated (step S204). In this case, the likelihood Lh15 is not calculated (NO in step S204). Therefore, the likelihood calculating unit 22 repeats the processes of steps S202 to S204.

  Specifically, the likelihood calculating unit 22 selects the function f5 (step S202).

  Next, the likelihood calculating unit 22 calculates the likelihood Lh15 as the likelihood corresponding to the Doppler shift amount ds of the tracking target 140 for the Doppler shift amount ds (step S203). Specifically, the likelihood calculating unit 22 refers to the data of the function f5 and the data 211 of the Doppler shift amount ds of the target 131. Then, the likelihood calculating unit 22 assigns the Doppler shift amount ds as an independent variable to the function f5. As a result, the dependent variable in the function f5 is calculated as the likelihood Lh15. That is, the likelihood Lh15 of the Doppler shift amount ds for the target 131 is calculated.

  In the flowchart shown in FIG. 8, the likelihood calculating unit 22 calculates each of the likelihoods Lh (Lh11 to Lh15) individually. However, the likelihood calculating unit 22 may collectively calculate the likelihoods Lh (Lh11 to Lh15) by matrix calculation.

  Next, the likelihood calculating unit 22 determines whether all of the likelihoods Lh11 to Lh15 have been calculated (step S204). In this case, likelihoods Lh11 to Lh15 are calculated (YES in step S204). Therefore, the likelihood calculating unit 22 calculates the total likelihood Lh = likelihood Lh11 × Lh12 × likelihood Lh13 × h14 × Lh15 for the target 131 (step S205).

  Next, the likelihood calculating unit 22 determines whether or not the total likelihood is calculated for all the targets 130 (131, 132) in the selection region S1 (n) (step S206). In this case, the total likelihood Lh200 for the target 132 is not calculated (NO in step S206).

  Therefore, the likelihood calculating unit 22 repeats the processes of steps S201 to S205. That is, the likelihood calculating unit 22 performs the same process as the process of calculating the likelihoods Lh11, Lh12, Lh13, Lh14, and Lh15 for the target 131. Accordingly, the likelihood calculating unit 22 calculates the likelihoods Lh21, Lh22, Lh23, Lh24, and Lh25 for the target 132. The likelihoods Lh21, Lh22, Lh23, Lh24, and Lh25 are calculated in the same manner as the likelihoods Lh11, Lh12, Lh13, Lh14, and Lh15, respectively. Next, the likelihood calculating unit 22 calculates the overall likelihood Lh200 by multiplying the likelihoods Lh21, Lh22, Lh23, Lh24, and Lh25 (step S205).

  When the likelihood calculating unit 22 calculates the total likelihoods Lh100 and Lh200 for all the targets 131 and 132 in the selection region S1 (n) (YES in step S206), the likelihood comparing unit 23 performs step S207. Perform the process. That is, the likelihood comparison unit 23 specifies the tracking target 140 using the total likelihoods Lh100 and Lh200 (step S207). In the present embodiment, the likelihood comparison unit 23 selects the overall likelihood Lh100 having the highest value among the overall likelihoods Lh100 and Lh200.

  In the present embodiment, as shown in FIG. 4, the estimated position PRe (n−1) of the target echo image 121 related to the target 131 and the position of the representative point P1 (n) of the target echo image 121 Are separated by a distance D1. Further, the estimated position PRe (n−1) and the position of the representative point P2 (n) of the target echo image 122 related to the target 132 are separated by a distance D2. At the time of n scanning, the distance D1 is slightly larger than the distance D2. That is, as shown in FIG. 6, likelihood Lh11 <Lh21.

  On the other hand, as shown in FIG. 7, the area ar of the target 131 is different from the average μ2 of the area ar by Δar1 at the time of n scanning. Further, at the time of n scanning, the area ar of the target 132 is different from the average μ2 of the area ar by Δar2. Δar1 is clearly smaller than Δar2. That is, Lh12> Lh22.

  In this embodiment, the likelihood of the other feature information about the target 131 and the likelihood of the other feature information corresponding to the target 132 are substantially the same at the n scan time point. That is, the likelihoods Lh13, Lh14, and Lh15 for the target 131 are substantially the same as Lh23, Lh24, and Lh25 for the target 132, respectively.

  Therefore, the total likelihood Lh100 for the target 131 is larger than the total likelihood Lh200 for the target 132 (Lh100> Lh200). In this case, the likelihood comparison unit 23 specifies the target 131 as the tracking target 140 at the n-scan time point.

  As shown in FIG. 1, after the likelihood comparison unit 23 specifies the tracking target 140, the area determination unit 15 determines the area ar of the target echo image 121 for the tracking target 140. Thereby, the area determination unit 15 determines whether or not the target echo image 121 is fused with another target echo image 120 or excessive blur occurs.

  FIG. 9 is a flowchart for explaining an example of the processing flow in the area determination unit 15. First, the area determination unit 15 reads the data of the function f2 from the function update unit 21 (step S301). The area determination unit 15 further reads out the data 207 of the area ar at the time of n scans from the feature information memory 12 for the target echo image 121 of the tracking target 140 (step S302).

Next, the area determination unit 15 determines whether or not the area ar of the target echo image 121 at the n scan time point is larger than a predetermined threshold value. The threshold th in this case is, for example, (μ2 + 3 × σ2), and is set based on the average μ2. .sigma. @ 2 is the standard deviation in the function f2, a positive square root of the variance .sigma. @ 2 ^2. When the area ar of the target echo image 121 is larger than the predetermined threshold th (YES in step S303), the area determination unit 15 adds another echo image to the target echo image 121 at the time of n scanning. It is determined that fusion with 120 or excessive blur has occurred (step S304). In this case, the area determining unit 15 treats the representative point PR (n) as not being able to be specified by the associating unit 14, and does not output the data of the representative point PR (n) to the motion estimating unit 16 (step S305). , Complete the process.

  On the other hand, when the area ar of the target echo image 121 is equal to or smaller than the predetermined threshold th (NO in step S303), the area determination unit 15 adds another object to the target echo image 121 at the n scan time point. It is determined that neither fusion with the target echo image 120 nor excessive blur has occurred (step S306). In this case, the area determination unit 15 outputs the data of the representative point PR (n) to the motion estimation unit 16 (Step S307) and completes the process.

  The above-described (a) fusion of the target echo image 121 and another target echo image 120 and (b) blurring of the target echo image 121 will be described below.

  First, (a) the reason why the target echo image 121 and another target echo image 120 (125) are fused will be described with reference to FIGS. FIG. 10 is a schematic diagram for explaining the case where the target echo image 121 and the other target echo image 125 are not fused. FIG. 11 is a schematic diagram for explaining a case where fusion occurs between the target echo image 121 and another target echo image 125. As the target 135 specified by the target echo image 125, other ships, clutter, and the like can be exemplified.

  Referring to FIGS. 1, 10, and 11, the outline shape of each target echo image 120 (121, 125) detected by the antenna unit 2 is blurred due to an observation error in the antenna unit 2. It becomes the shape. That is, the size of the outline shape of the target echo image 120 (121, 125) is larger than the size of the outline shape of the target 130 (131, 135). However, as shown in FIG. 10, when the target 131 and the target 135 are sufficiently separated, the target echo image 121 and the target echo image 125 are not fused. That is, the tracking processing unit 11 determines that the target echo image 121 and the target echo image 125 are separate targets.

  On the other hand, when the distance between the target 131 and the target 135 is small, the outline of the target echo image 121 and the outline of the target echo image 125 are merged as shown in FIG. That is, the target echo image 121 and the target echo image 125 are fused. In this case, the tracking processing unit 11 erroneously determines that the target echo image 121 and the target echo image 125 are one target echo image. As a result, the echo detector 10 detects the center point PF1 of the one target echo image as a representative point of the target echo image 121. That is, the echo detector 10 detects the center point PF1 as the representative point PR (n) of the tracking target 140. In this case, the motion estimation unit 16 performs the tracking filter process using the representative point PF1. As a result, the estimated speed vector V1 (n) calculated by the motion estimation unit 16 has a large error with respect to the actual speed vector.

  When the fusion of the target echo images 121 and 125 occurs at the n scan time point, the area ar of the target echo image 121 at the n scan time point becomes extremely large. That is, the area ar for the tracking target 140 exceeds the predetermined threshold th. Therefore, the area determination unit 15 can determine whether the target echo image 121 and another target echo image 125 are fused by determining the area ar of the tracking target.

  Next, the reason why excessive blur occurs in the (b) target echo image 121 will be described with reference to FIG. FIG. 12 is a schematic diagram illustrating a state in which the target echo image 121 is excessively blurred. Referring to FIGS. 1 and 12, the target echo image 121 is detected as a particularly blurred shape in the antenna angle θ direction due to the resolution in the antenna angle θ direction in the antenna unit 2. There is a case. As a result, the area ar of the target echo image 121 at the time of n scanning may become excessively large. In this case, the area ar of the target echo image 121 exceeds a predetermined threshold th. In this case, the position of the representative point PF2 in the target echo image 121 is greatly deviated from the position of the true representative point P1 (n) in the target echo image 121. In this case, when the motion estimation unit 16 performs the tracking filter process using the representative point PF2, the estimated speed vector V1 (n) calculated by the motion estimation unit 16 has an error with respect to the actual speed vector. Therefore, the area determination unit 15 detects whether the area ar of the target echo image 121 is excessively large by determining the area ar of the tracking target 140 (target echo image 121). If the area ar exceeds the threshold value th, the area determination unit 15 determines that excessive blur has occurred in the echo image 121.

  Next, the motion estimation unit 16 will be described. The motion estimation unit 16 is configured to perform tracking filter processing. Thereby, the motion estimation unit 16 calculates the estimated velocity vector V1 (n) of the tracking target 140 at the n-scan time point. Examples of the tracking filter include an α-β filter and a Kalman filter. The motion estimation unit 16 outputs data specifying the estimated speed vector V1 (n) to the selection unit 13 and the display 4. The data of the estimated velocity vector V1 (n) is used for the tracking filter process at the (n + 1) scan time.

  Next, an example of processing in the motion estimation unit 16 will be described. FIG. 13 is a flowchart for explaining an example of a processing flow in the motion estimation unit 16. FIG. 14 is a schematic diagram for explaining processing in the motion estimation unit 16. FIG. 14 shows a case where the target echo image 121 that specifies the tracking target 140 is neither fused with another target echo image 125 nor excessively blurred. FIG. 15 is a schematic diagram for explaining processing in the motion estimation unit 16. FIG. 15 shows a case where the target echo image 121 that specifies the tracking target 140 is fused with another target echo image 125 or excessive blur occurs.

  First, the motion estimation unit 16 reads data specifying the result of the tracking filter processing in the motion estimation unit 16 at the time of (n−1) scan (step S401). In this case, as shown in FIGS. 14 and 15, the result of the tracking filter processing is the estimated coordinates of the smooth position PRs (n−1) with respect to the tracking target 140 at the time of the (n−1) scan, and the estimated values. The coordinates of the position PRe (n−1) and the estimated speed vector V1 (n−1) are included.

  Next, the motion estimation unit 16 determines whether or not the data of the representative point PR (n) is output from the area determination unit 15 (step S402). When the data of the representative point PR (n) is output (YES in step S402), that is, in the target echo image 121, both fusion with another target echo image 125 and excessive blurring occur. If not, the motion estimation unit 16 performs a tracking process using data of the representative point PR (n) observed at the n scan time (steps S403 to S404).

  Specifically, the motion estimation unit 16 reads the coordinate data of the representative point PR (n) of the tracking target 140 from the likelihood calculation unit 22 (step S403). Thereafter, the motion estimation unit 16 performs tracking filter processing using the data read in steps S401 and S403 (step S404). Accordingly, as illustrated in FIG. 14, the motion estimation unit 16 includes the smooth position PRs (n), the estimated position PRe (n), and the estimated velocity vector V1 (n) for the target echo image 121 at the time of n scans. ) And.

  The motion estimation unit 16 outputs data specifying the result of the tracking filter process to the feature information memory 12, the selection unit 13, and the display 4 (step S405).

  On the other hand, if the data of the representative point PR (n) is not output (NO in step S402), that is, the target echo image 121 is fused with another target echo image 120 or excessive blurring occurs. If it is present (NO in step S402), the motion estimation unit 16 performs the tracking process without using the representative point PR (n) at the n-scan time point (step S404).

  In this case, as shown in FIG. 15, the estimated speed vector V1 (n) calculated by the motion estimation unit 16 has the same direction and size as the estimated speed vector V1 (n-1). The starting point of the estimated speed vector V1 (n) is the end point of the estimated speed vector V1 (n-1). Thereafter, the motion estimation unit 16 performs the process of step S405. The above is an example of the process flow in the motion estimation unit 16.

  The display 4 is a liquid crystal display capable of color display, for example. The display 4 displays each target echo image 120 on the display screen using the image data of each target echo image 120. The display 4 displays the estimated speed vector V1 (n) as an image. Accordingly, an image indicating the estimated velocity vector V1 (n) is displayed on the display screen of the display device 4 for the tracking target 140 (target echo image 121). The operator of the radar apparatus 1 can confirm the motion state of the tracking target 140 by confirming the radar image displayed on the display 4.

  As described above, according to the tracking processing device 3, the likelihood calculating unit 22 of the associating unit 14 represents the representative of the tracking target 140 with respect to the representative points P1 (n) and P2 (n) of the selection region S1 (n). Likelihoods Lh11 and Lh21 corresponding to the point PR (n) are calculated. Then, the likelihood comparison unit 23 of the associating unit 14 specifies the tracking target 140 from among the plurality of targets 131 and 132 based on the calculated likelihoods Lh11 and Lh21. As described above, the associating unit 14 can more accurately determine the representative point PR (n) of the tracking target 140, that is, the representative point P1 (n), by using a statistical method. Further, the associating unit 14 does not need to set a model considering the prediction error, the observation error, and the like for each target 130 when calculating the likelihoods Lh 11 and 21. That is, the likelihood calculating unit 22 does not need to perform an operation according to a complicated model. For this reason, the arithmetic processing in the associating unit 14 is simple. Therefore, according to the tracking processing device 3, the tracking target 140 can be reliably and easily detected from the plurality of targets 130 (131, 132).

  Further, according to the tracking processing device 3, the associating unit 14 provides feature information other than information on the representative points P1 (n) and P2 (n) for each target 131 and 132 located in the selection region S1 (n). refer. And the likelihood calculation part 22 of the correlation part 14 calculates likelihood Lh12-Lh15; Lh22-Lh25 applicable to the characteristic information of the tracking target 140. FIG. In addition, the likelihood comparing unit 23 of the associating unit 14 specifies the tracking target 140 from the targets 131 and 132 based on the calculated likelihoods Lh11 to Lh15; Lh21 to Lh25. In this manner, the associating unit 14 uses the likelihoods Lh12 to Lh15; Lh22 to Lh25 in addition to the likelihoods Lh11 and 21 for the representative points P1 (n) and P2 (n) of the targets 131 and 132, The tracking target 140 can be specified. Therefore, the tracking target 140 can be more reliably determined from the plurality of targets 131 and 132. Further, as described above, the processing using the likelihood Lh is simple arithmetic processing unlike the complicated arithmetic processing according to the complicated model. Therefore, even when likelihood determination is performed for additional feature information other than the representative point P (n), the calculation load on the associating unit 14 can be small.

  Further, for example, consider a case where the echo detection unit 10 detects a wave generated by navigation of a ship or clutter as a target. In this case, the likelihood for the area of the wave or the clutter is much smaller than the corresponding likelihoods Lh12 and 22 for the targets 131 and 132. Therefore, the likelihood comparison unit 23 can suppress erroneous determination of the wave or clutter as the tracking target 140.

  Further, for example, Patent Documents 1 and 2 described above disclose tracking processing based on multiple hypotheses based on track information as one feature information on a tracking target. On the other hand, the tracking processing device 3 specifies the tracking target 140 from among the plurality of targets 130 using the plurality of feature information. Thereby, the tracking target 140 can be more reliably determined from the plurality of targets 130.

  Further, according to the tracking processing device 3, the associating unit 14 uses at least one of area ar information, echo signal level ec, adjacent distance ad information, and Doppler shift amount ds for each echo image 120. Using the feature information, the associating unit 14 can more accurately identify the tracking target 140 from among the plurality of targets 130. More specifically, the area ar information is used as the feature information. Accordingly, the associating unit 14 can suppress erroneously determining that the target 132 having a shape different from the shape of the tracking target 140 (target 131) is the tracking target 140. For example, when the tracking target is a large ship, it is more reliably suppressed that the associating unit 14 detects the target detected from the small ship or the clutter as the tracking target. it can. In addition, an echo level ec is used as feature information. Thereby, the associating unit 14 can suppress erroneously determining that the target 130 having the echo level ec different from the tracking target 140 is the tracking target 140. Further, as the feature information, information on the adjacent distance ad is used. Thereby, the associating unit 14 can more reliably determine the tracking target 140 (target 131) and the target such as clutter. Further, the Doppler shift amount ds is used as the feature information. Thereby, the echo detection unit 10 can detect the relative speed between the ship 100 and the target 130 in the direction in which the ship 100 and the target 130 face each other. Therefore, the associating unit 14 can specify the tracking target 140 from the plurality of targets 130 in consideration of the relative speed. As a result, for example, when the tracking target 140 (target 131) and another target 130 pass each other, the likelihood comparison unit 23 determines that the other target 130 is the tracking target 140. Can be prevented.

  Further, according to the tracking processing device 3, the likelihood calculating unit 22 of the associating unit 14 calculates the likelihoods Lh11 to Lh15; Lh21 to Lh25 using the function f (f1 to f5) as the probability density function. Thus, the likelihood calculating unit 22 can calculate the likelihoods Lh11 to Lh15; Lh21 to Lh25 statistically and with a simple calculation by using the function f.

  Further, according to the tracking processing device 3, the likelihood calculating unit 22 of the associating unit 14 uses the feature information (representative points P1 (n) and P2 () detected by the echo detecting unit 10 in each function f (f1 to f5). n) coordinates, area ar, echo level ec, adjacent distance ad, Doppler shift amount ds) are used as independent variables. In addition, the likelihood calculating unit 22 calculates the dependent variables subordinate to the independent variables as the likelihoods Lh11 to Lh15; Lh21 to Lh25. As a result, based on the reason that the position of the representative point P2 (n) is close to the estimated position PRe (n-1), the representative point P2 (n) is changed to the representative point PR (n) of the tracking target 140. Identification as is prevented. That is, the associating unit 14 specifies the tracking target 140 in consideration of the likelihoods Lh12 to Lh15; Lh22 to Lh25 in other feature information other than the information of the representative points P1 (n) and P2 (n). Thereby, the associating unit 14 can more accurately identify the tracking target 140 from among the targets 131 and 132.

  Further, according to the tracking processing device 3, the likelihood calculating unit 22 of the associating unit 14 calculates the overall likelihoods Lh100 and Lh200 for each of the targets 131 and 132. Then, the likelihood comparing unit 23 identifies the target 131 having the highest overall likelihood Lh100 and Lh200 as the tracking target 140 among the targets 131 and 132. That is, the likelihood comparison unit 23 comprehensively determines the likelihoods Lh11 to Lh15; Lh21 to Lh25 based on a plurality of feature information for each target 131 and 132. Thereby, the associating unit 14 can more reliably specify the tracking target 140 from among the targets 131 and 132.

  Further, according to the tracking processing device 3, the area determination unit 15 determines whether or not the area ar in the target echo image 121 that specifies the tracking target 140 exceeds a predetermined threshold th. In addition, when the area ar is equal to or smaller than the predetermined threshold th, the motion estimation unit 16 calculates the estimated velocity vector V1 (n) using information on the representative point PR (n) at the n-scan time point. . Further, when the area ar exceeds the predetermined threshold th, the motion estimation unit 16 uses the estimated velocity vector V1 (n) without using the information of the representative point PR (n) at the n-scan time point. ) Is calculated. With such a configuration, when information suitable for accurate motion estimation is obtained for the representative point PR (n), the motion estimation unit 16 uses the information to perform the motion of the tracking target 140. Can be calculated more accurately. On the other hand, even if the motion estimation unit 16 does not have information suitable for accurate motion estimation for the representative point PR (n), the motion estimation unit 16 calculates the motion of the tracking target 140 more accurately without using the information. it can.

  Further, the area determination unit 15 determines the area ar of the target echo image 121 of the tracking target 140 using the function f2. For this reason, even when the area ar of the target echo image 121 of the tracking target 140 changes with time due to an observation error or the like in the antenna unit 2, the area determination unit 15 makes the area ar more Accurate judgment can be made.

  Patent Document 3 discloses a configuration for dealing with a state in which a jump (jump) has occurred at an observation position to be tracked. However, with the configuration described in Patent Document 3, it cannot be said that the above-described accuracy of detecting the fusion of echo images is sufficient. On the other hand, according to the tracking processing device 3, the area determination unit 15 determines the area ar using the function f2. As a result, the area determination unit 15 can detect the fusion more reliably when the target echo image 121 that specifies the tracking target 140 is fused with another echo image 125.

  Further, according to the tracking processing device 3, the area determining unit 15 sets the threshold th based on the average μ2 in the function f2. Then, the area determination unit 15 determines the size of the area ar for the tracking target 140 with reference to the threshold value th. According to such a configuration, when a value greatly deviating from the average μ of the area ar is detected as the area ar, the area determination unit 15 uses the target echo image 121 that identifies the tracking target 140 as follows: It is determined that fusion with another target echo image 125 or excessive blur has occurred. With such a configuration, the area determination unit 15 can determine the area ar of the target echo image 121 more accurately without being affected by the observation error in the antenna unit 2.

  As mentioned above, although embodiment of this invention was described, this invention is not limited to the above-mentioned embodiment, As long as it described in the claim, various changes are possible. For example, the following modifications may be made.

(1) In the above-described embodiment, an example has been described in which area information or the like about a target echo image is used as additional feature information other than representative point information about the target. However, this need not be the case. For example, the track information of the target detected by the echo detector may be used as the additional feature information. When the wake information is used as the additional feature information, the tracking target can be identified with higher accuracy from a plurality of targets that travel in a complicated manner.

(2) In the above-described embodiment, an example in which the area information of the target echo image is used as the shape information has been described. However, this need not be the case. For example, the width of the target echo image may be used as the shape information. Further, the length of the target echo image may be used as the shape information.

(3) Further, in the above-described embodiment, an example in which the center point of the target echo image is used as the representative point has been described. However, this need not be the case. For example, the forefront edge point that is the closest point to the ship in the target echo image may be used as the representative point.

(4) Moreover, in the above-mentioned embodiment, the tracking processing apparatus demonstrated to the example the form which is a tracking processing apparatus for ships. However, the present invention is not limited to a tracking processing device for ships, but can be applied as a tracking processing device for other targets.

  The present invention can be widely applied as a tracking processing device and a tracking processing method.

3 Tracking processing device 10 Echo detection unit (detection unit)
9 Signal processor (detector)
19 Association part (likelihood determination part)
130 Target 140 Tracking target (tracking target)
Lh11, Lh21 Likelihood P corresponding to the representative point to be tracked Representative point

Claims (7)

A tracking processing device for tracking a target selected as a tracking target,
For a plurality of targets, a detection unit for detecting feature information including information on representative points;
A likelihood determination unit that calculates a likelihood corresponding to the representative point to be tracked for at least some of the plurality of representative points ;
A motion estimation unit that estimates the motion state of the tracking target using the feature information of the target identified as the tracking target;
An area determination unit for determining the area ,
The likelihood determining unit identifies the tracking target from a plurality of the targets based on the calculated likelihood ,
The detection unit detects feature information other than information on the representative point as additional feature information for at least a part of the plurality of targets,
The likelihood determination unit calculates a likelihood corresponding to the feature information of the tracking target for the additional feature information, and determines the tracking target from a plurality of the targets based on the likelihood. Identify,
The detection unit detects an echo signal that is a reflected wave from the plurality of targets including the tracking target with respect to a transmission signal, detects an echo image specified by the echo signal, and detects the echo image Is detected as additional feature information,
The area determination unit determines whether the area of the echo image to be tracked exceeds a predetermined threshold;
When the area is equal to or less than a predetermined threshold, the motion estimation unit calculates the motion state using the feature information of the tracking target at the time of estimating the motion state, and the area A tracking processing device that calculates the motion state without using the feature information of the tracking target at the time when the motion state is calculated when is exceeding a predetermined threshold . The tracking processing device according to claim 1 ,
The detection unit detects each feature information using an echo signal that is a reflected wave at each target with respect to a transmission signal,
The additional feature information includes at least the shape information of the echo image specified by the echo signal, the level of the echo signal, the information specifying the surrounding state of each target, and the Doppler shift amount related to the echo signal. A tracking processing device including one. The tracking processing device according to claim 1 or 2 , wherein
Each of the feature information is information represented by numerical values,
The said likelihood determination part calculates each said likelihood using a probability density function for every said feature information, The tracking processing apparatus characterized by the above-mentioned. The tracking processing device according to claim 3 ,
The probability density function is set for each feature information,
The value used as the average value in each probability density function is a value estimated in advance before the likelihood is calculated for the corresponding feature information,
The likelihood processing unit uses the corresponding feature information as an independent variable in each probability density function, and calculates a dependent variable subordinate to the independent variable as the likelihood. The tracking processing device according to claim 4 ,
The likelihood determining unit
Multiplying the likelihoods for the same target by each other to calculate an overall likelihood, and
A tracking processing device, wherein the target having the highest overall likelihood is specified as the tracking target among the targets. The tracking processing device according to any one of claims 1 to 5 ,
The area determination unit is configured to determine the area using a probability density function,
The value used as the average value of the probability density variable is a value estimated in advance for the area before being determined by the area determination unit,
The tracking processing device, wherein the predetermined threshold is set based on the average. A tracking processing method for tracking a target selected as a tracking target,
A detection step for detecting feature information including representative point information for a plurality of targets;
A likelihood determination step for calculating a likelihood corresponding to the representative point to be tracked for each of at least some of the representative points; and
Using the feature information of the target identified as the tracking target, estimating a motion state of the tracking target, a motion estimation step;
Including an area determination step for determining an area ,
In the likelihood determination step, the tracking target is identified from among the plurality of targets based on the calculated likelihood ,
In the detection step, for at least some of the plurality of targets, feature information other than the representative point information is detected as additional feature information,
In the likelihood determination step, for the additional feature information, a likelihood corresponding to the feature information of the tracking target is calculated, and the tracking target is selected from a plurality of the targets based on the likelihood. Identify,
In the detection step, an echo signal that is a reflected wave at the plurality of targets including the tracking target with respect to the transmission signal is detected, an echo image specified by the echo signal is detected, and the echo image is detected Is detected as additional feature information,
In the area determining step, it is determined whether or not the area of the echo image to be tracked exceeds a predetermined threshold value,
In the motion estimation unit step, when the area is equal to or less than a predetermined threshold, the motion state is calculated using the feature information of the tracking target at the time of estimating the motion state, and the A tracking processing method , wherein, when the area exceeds a predetermined threshold, the motion state is calculated without using the feature information of the tracking target at the time of calculating the motion state .

Images