JP4779431B2 - Object recognition device - Google Patents

Description

  The present invention relates to an object recognition device and a recognition method for recognizing an object.

  One method for recognizing a three-dimensional object is to acquire data representing a distance image of a certain point on the object and recognize the object from the obtained distance image. The distance image is an image in which the three-dimensional position information on the surface of the object to be imaged is divided and arranged in a grid pattern, and the divided image is called a pixel.

  FIG. 25 is a diagram illustrating a state in which object distance image data is acquired with respect to the object recognition apparatus 1 that performs distance measurement using a pulse laser. The distance image data is data representing the position of an object or the like. In FIG. 25, the object 2 is represented by being divided into a lattice shape, and each mass of the lattice indicates a range irradiated with the pulse laser, and one unit of a pixel whose distance is measured. It corresponds.

  In this way, the object recognition apparatus divides the entire object into pixel units, and repeatedly obtains distance image data for the entire object by irradiating pulse laser. However, the distance image data itself is merely a collection of points indicating the positions of the points reflected by the pulse laser, and it is not possible to recognize a three-dimensional object by itself. Therefore, in order to recognize a three-dimensional object, a desired object must be extracted from the distance image data thus obtained. Here, extracting the object means grasping the outer shape of the object based on the acquired distance image data.

As a method for recognizing a three-dimensional object, the outline of the object is first extracted from the acquired distance image data, and then the distance image data included in the extracted outline and the object recognition apparatus in advance. There is a method for recognizing an object by matching a stored three-dimensional template that is data of the object to be recognized by the object recognition device (see Non-Patent Document 1 and Non-Patent Document 2). ).

  In any of the methods described in Non-Patent Document 1 and Non-Patent Document 2, in the matching, without considering which point of the acquired distance image data and which point are on the same plane of the object, A method is used in which it is determined whether or not all the points of the extracted distance image data match the three-dimensional template. However, in these methods, the amount of calculation required for collating the acquired distance image data with the three-dimensional template is enormous, and it takes time to recognize the object. It was difficult to recognize.

  In recognizing the object from the distance image data, the present invention extracts information related to the edge that is a part of the object from the acquired distance image data, and uses the information related to the edge to determine the plane of the object. Provided are an object recognition device and a recognition method for quick extraction.

  An object recognition device mounted on a moving body that extracts information related to an edge that is a part of the object from the acquired distance image data, and extracts a plane of the object using the information related to the edge. And a recognition method.

  According to the present invention, in recognizing an object from the acquired distance image data, information related to an edge such as an edge line that is a part of the object is first extracted, and the object related to the edge is used using the information related to the edge. Since the plane of the object is extracted, the processing of the distance image data for extracting the plane can be performed efficiently, the amount of calculation used for processing the data can be reduced, and particularly for an object that moves at high speed. Even if it exists, it can extract correctly.

Further, even when the object recognition device is mounted on a moving body and the object recognition device moves at high speed, the object can be accurately recognized.

IEEE TRANSACTION ON IMAGE PROCESSING, VOL. 10 NO. 4 APRIL, 2001 “Model-Based Target Recognition in Pulsed Ladar Image“ Proceedings of SPIE Volume. 2748 Laser Radar Technology and Applications p272-282 “Identification of vehicle targets from low-cost lader seeker image”

  The present invention provides extraction means for extracting information related to an edge of the target object based on distance image data representing the position of at least a part of the target object, and recognition for recognizing the target object based on the information related to the edge An object recognition device comprising means.

  Further, the present invention provides distance image data acquisition means for acquiring a plurality of distance image data representing the positions of at least some of the objects, and the distance image data satisfying a first condition among the plurality of acquired distance image data. First extraction means for extracting the information, second extraction means for extracting information relating to the edge of the object based on the distance image data that satisfies the extracted first condition, and based on the information It is a target object recognition apparatus provided with the recognition means which recognizes the said target object, It is characterized by the above-mentioned.

  Further, the present invention is a launching means for emitting light to an object, an incident means for entering the light reflected by the object, an incident light level measuring means for measuring the level of the incident light, Distance image data acquisition means for acquiring a plurality of distance image data representing the position of the object from the incident light, and extracting the distance image data satisfying a first condition from the plurality of acquired distance image data First extraction means; second extraction means for extracting information relating to the edge of the object based on the extracted range image data and the level satisfying the first condition; and information relating to the edge An object recognition apparatus comprising a recognition means for recognizing the object based on the object.

  Furthermore, it is preferable that the object recognition apparatus is an object recognition apparatus in which the information related to the edge is an edge line of the object.

  Furthermore, it is desirable that the information related to the edge is an object recognition device that is a vertex of the object.

  Furthermore, it is desirable that the first condition is an object recognition device that is determined based on a distance to the object.

  Further, the second extraction means extracts the edge line candidates of the plane constituting the object based on the distance image data and the light level that satisfy the first condition, and among the edge line candidates, It is desirable that the target object recognition apparatus be a means for extracting edge line candidates that satisfy the second condition.

  Furthermore, it is desirable that the second condition is an object recognition device that is determined based on the number of the distance image data included in the solid figure including the edge line candidate.

  Further, the object recognition device has plane extraction means for extracting a plane constituting the object based on information on the edge, and the recognition means is based on the plane extracted by the plane extraction means. And an object recognition device that recognizes the object.

  A setting unit configured to set a virtual plane based on the information relating to the edge; and a plane extracting unit configured to extract a plane that is a virtual plane satisfying a third condition among the virtual planes set in the setting unit. Preferably, the recognition means is an object recognition device that recognizes the object based on the plane extracted by the plane extraction means.

  Furthermore, it is desirable that the third condition is an object recognition device that is determined based on the number of the distance image data included in the three-dimensional figure including the virtual plane.

  Furthermore, it is desirable that the recognition means is an object recognition device provided with means for recognizing an object based on a combination of the planes extracted by the plane extraction means.

  The apparatus further comprises storage means for recording a template of the object, and collation means for collating the plane with the template, wherein the recognition means recognizes the object based on the plane that matches the template in the collation means. It is desirable that the object recognition apparatus includes means for performing the above-described operation.

  Further, the present invention is an object recognition device that extracts information related to an edge of the object based on distance image data representing the position of the object and recognizes the object based on the information. It is characterized by.

  Further, the object includes a GPS and an inertial sensor for recognizing the position and orientation of the own device, and the distance image data is data based on the position and orientation of the own device recognized by the GPS and the inertial sensor. It is desirable to be an object recognition device.

  The present invention also includes an extraction step of extracting information related to the edge of the target object based on distance image data representing the position of the target object, and a recognition step of recognizing the target object based on the information. It is an object recognition method.

  Further, the present invention provides a distance image data acquisition step for acquiring a plurality of distance image data representing the position of an object, and a step of extracting the distance image data satisfying a first condition from the plurality of acquired distance image data. A first extraction step, a second extraction step for extracting information relating to an edge of the object based on the extracted distance image data satisfying the first condition, and the object based on the information. An object recognition method comprising: a recognition step for recognizing.

Further, the present invention includes a launching step of emitting light to an object, an incident step in which the light reflected by the object is incident, an incident light level measuring step of measuring the level of the incident light, A distance image data acquisition step for acquiring a plurality of distance image data representing the position of the object from the incident light, and extracting the distance image data satisfying a first condition from the plurality of acquired distance image data A first extraction step; a second extraction step for extracting information relating to the edge of the object based on the distance image data and the level satisfying the extracted first condition; and information relating to the edge And a recognition step for recognizing the object on the basis of the object recognition method.

  A block diagram of the object recognition apparatus of the present embodiment is shown in FIG. The control unit 11 outputs a firing time signal To that indicates the time at which the laser emitting unit 12 emits the pulse laser Po to the laser emitting unit 12 and the distance measuring circuit unit 15. Further, a firing intensity signal Vo representing the intensity of the pulse laser Po is output from the control unit 11 to the laser emitting unit 12.

  The laser emitting unit 12 emits a pulse laser Po to the scanning optical system 13 in accordance with the input emission time signal To and the emission intensity signal Vo.

  The scanning optical system 13 scans the pulse laser Po emitted from the laser emitting unit 12 in the biaxial direction, emits the pulse laser Po to the outside, and enters the pulse laser Pi reflected by the object. The incident pulse laser Pi is output to the photoelectric converter 14. Further, the scanning optical system 13 outputs the firing angle (Φo, ψo) representing the AZ angle Φ and the EL angle Ψ of the scanning angle when the pulse laser Po is emitted to the coordinate conversion unit 171 described later.

  The photoelectric conversion unit 14 converts the input pulse laser Pi into an electric pulse Ei. The converted electric pulse Ei is output to the distance measuring circuit unit 15 and the incident light level measuring unit 16.

  The distance measuring circuit unit 15 performs a calculation represented by Ti-To using the incident time signal Ti obtained from the input electric pulse Ei and the emission time signal To representing the emission time of the pulse laser Po, and performs a calculation represented by Ti-To. The time from when the laser Po is emitted until the pulse laser Pi is incident is obtained. If c is the speed of light, the distance Ro to the object can be obtained by calculation represented by (Ti−To) × c / 2. The obtained distance Ro is output to the coordinate conversion unit 171.

  The incident light level measurement unit 16 measures the incident light level Vi of the pulse laser Pi from the input electric pulse Ei. The measured incident light level Vi is output to the edge candidate extraction unit 173.

  The coordinate conversion unit 171 calculates the distance image data (Ro, Φo, Ψo) including the input launch angle (Φo, Ψo) and the distance Ro to the object from the polar coordinates (Ro, Φo, Ψo), Xo = Rocos ΨosinΦo, Yo. = Rocos [Psi] ocos [Phi] o, and Zo = Rosin [Psi] o (Xo, Yo, Zo). The relationship between polar coordinates and Cartesian coordinates is as shown in FIG. The distance image data (Xo, Yo, Zo) subjected to coordinate conversion and the distance Ro to the object are output to the distance image data extraction unit 172 in association with each other.

  In the present invention, a sufficient amount of distance image data is necessary to extract the plane constituting the object. Therefore, the pulse laser Po is repeatedly emitted and reflected while changing the emission angle (Φo, Ψo). A plurality of the acquired distance image data are acquired by entering a pulse laser Pi. Therefore, the coordinate conversion unit 171 performs coordinate conversion on the plurality of acquired distance image data, and outputs the result to the distance image data extraction unit 172.

  The distance image data extraction unit 172 extracts distance image data (Xo, Yo, Zo) corresponding to the distance Ro included in the distances Rd to Rd + ΔR in the specific range.

  The value of ΔR representing the range of distances to be extracted is set to a value that does not cause an error in consideration of the ratio between ΔR and the distance Rd. For example, if Rd = 1000 [m] and ΔR = 10 [m], the difference in incident light level due to distance, that is, the difference in incident light level at distances Rd and Rd + ΔR is several percent. Is not a problem. The extracted distance image data (Xo, Yo, Zo) is output to the edge candidate extraction unit 173 and the plane determination unit 175.

  The edge candidate extraction unit 173 uses the incident light level Vi corresponding to the input distance image data (Xo, Yo, Zo) to represent an edge line or vertex edge corresponding to a side or vertex of a plane constituting the object. Edge candidates that are distance image data candidates are extracted. The measurement points satisfying the condition that the incident light level Vi is a predetermined value, for example, ½ or less of the maximum value of the incident light level of the distance image data, are extracted as candidates constituting a part of the edge. The reason for determining whether or not an edge candidate is based on the difference in the incident light level is that if the plane is different, the incident light level is different depending on the incident range and angle of the laser incident from the object recognition device. It is. In FIG. 3, since the range in which the object 3 is irradiated on the laser 4 becomes smaller in the order of (a), (b), and (c), the incident light level is in the order of (a), (b), and (c). Get smaller. Further, the incident light level varies depending on the angle of the incident laser, and is so large that the light is irradiated at an angle close to perpendicular to the surface. In FIG. 4, the incident light level decreases in the order of (a), (b), and (c). In particular, when the object has a rectangular parallelepiped shape, the two adjacent surfaces are almost perpendicular, so the difference in the incident light level incident on both surfaces becomes very large.

  The edge candidate extraction unit 173 determines whether or not the number of distance image data in a range sufficient for edge line extraction has been extracted. If it is determined that the number of distance image data is not sufficient, the distance image data extraction unit 172 Request edge candidate data again. The requested distance image data extraction unit 172 increases the distance range by ΔR, extracts distance image data corresponding to the distance Ro satisfying the condition of Rd + ΔR to Rd + 2ΔR, and thereafter, until the sufficient number of data is extracted. And the above operation is repeated.

  When it is determined that a sufficient number of distance image data has been extracted, the edge candidate extraction unit 173 extracts edge candidate lines from the edge candidate data. For example, as shown in FIG. 5, first, an arbitrary point is extracted from the distance image data, and that point is set as an edge candidate point D1. Then, assuming an edge candidate line L1 that connects the edge candidate point D2 that is closest to the edge candidate point D1, the other edge candidate data is in a space of a three-dimensional figure such as a predetermined cylinder centered on the edge candidate line L1. When a predetermined number or more exist, the edge candidate line L1 is determined to be an actual edge line. The size of the space of the solid figure such as the cylindrical shape and the predetermined number of edge candidate data are determined in advance in consideration of the number of firings of the pulse laser Po, the range of firing angles (Φo, Ψo), and the like. The straight line expression representing the edge candidate line may be a straight line equation in general geometric mathematics.

  In addition, when recognizing a specific target object such as an automobile, the length of the edge line can be assumed to some extent. Therefore, in order to perform a quick extraction process, the edge line size is determined based on the assumed target size. It is also possible to preset the length to be within a predetermined value range. Similarly, when detecting a specific object such as an automobile, the range of the position where the plane exists can also be assumed to be within the predetermined range, so the range where the edge line exists can be set to be limited to the predetermined range. it can.

  The edge line L1 is extracted by this process, and vertex edges V1 and V2 representing the vertexes of the plane, which are distance image data at both ends of the edge line L1, are obtained. The position of the vertex edge that is the end point of the edge line may be obtained by expanding or contracting the length of the solid figure and obtaining the increase or decrease of the number of edge candidate data included in the solid figure at that time. Alternatively, it may be obtained from the distribution of the position and number of edge candidate data. The obtained edge candidate line L1 and vertex edges V1 and V2 are output from the edge candidate extraction unit 173 to the virtual plane calculation unit 174.

  As illustrated in FIG. 6, the virtual plane calculation unit 174 sets a plurality of virtual plane candidates having the obtained edge line L <b> 1 as one side, and outputs the information to the plane determination unit 175. The plane equation may be a plane equation in general geometric mathematics.

  The plane determination unit 175 sets a solid figure (for example, a thin rectangular parallelepiped) F1, F2, and F3 including a plurality of virtual planes S1, S2, S3... Based on information output from the virtual plane calculation unit 174, and sets the solid figure Each of them calculates how much distance image data output from the distance image data extraction unit 172 is included. It is determined that a plane included in a solid figure including a lot of distance image data has a higher possibility of a real plane. For example, when the three-dimensional figure F2 includes the largest amount of distance image data, the virtual plane S2 included in the three-dimensional figure F2 is determined as one of the existing planes constituting the object and extracted.

  Here, for example, assuming that the edge line of one plane such as a cube or a rectangular parallelepiped is also the edge line of another plane, the plane determination unit 175 performs virtual plane calculation so that the plane can be efficiently extracted. At least two planes may be simultaneously extracted from a plurality of virtual planes based on the information output from the unit 174. At this time, as a condition for extracting a plurality of planes, for example, it is conceivable that the number of distance image data included in the solid figure including the planes is equal to or greater than a predetermined value. Further, if at least one plane is extracted in the plane determination unit 175, at least four edge lines and four vertex edges constituting the extracted plane are extracted. And the virtual plane calculation unit 174 may set a new virtual plane using the information. Further, in the above description, the edge candidate extraction unit 173 obtains the edge line and the vertex edge, but first obtains the vertex edge as described above without obtaining the edge line, and outputs the vertex edge to the virtual plane calculation unit 174. The virtual plane calculation unit may set a virtual plane based on the vertex edge.

  In reality, an object such as an automobile is hardly composed of a complete plane. Therefore, in order to appropriately extract an object, it is necessary to extract not only a plane but also a substantially plane including a slightly concave or convex portion. Therefore, even if there is a difference in the incident light level of the laser pulse that is reflected by the same substantially flat surface due to the unevenness of the substantially flat surface and is incident on the object recognition device, the object recognition device recognizes the difference as an error. For example, by setting the edge candidate extraction unit 173 so as to be within a range, it is possible to extract even a substantially flat surface. Note that the plane described in this specification includes a substantially plane.

  Next, a flowchart in the present embodiment is shown in FIG. First, Ro obtained by the ranging circuit unit 15 from the emission time signal To output from the control unit 11 and the electric pulse Eo converted from the pulse laser Po in the photoelectric conversion unit 14, and the scanning optical system 13. Distance image data represented by polar coordinates (Ro, Φo, Ψo) is acquired based on the output launch angles (Φo, Ψo) (S101).

  The coordinate converter 171 converts the distance image data from polar coordinates (Ro, Φo, Ψo) to Cartesian coordinates (Xo, Yo, Zo) represented by Xo = Rocos ΨosinΦo, Yo = Rocos ΨocosΦo, and Zo = Rosin Ψo, The coordinate image-converted distance image data (Xo, Yo, Zo) and the distance Ro to the object are associated with each other and output to the distance image data extraction unit 172 (S102).

  The distance image data extraction unit 172 extracts, from the acquired distance image data, distance image data in which the distance Ro to the object is included in the distances Rd to Rd + ΔR in the specific range. The value of ΔR representing the range of distance to be extracted is set to a value that does not cause an error in consideration of the ratio between ΔR and the distance Rd (S103).

  The edge candidate extraction unit 173 extracts edge candidate data using the incident light level Vi corresponding to the obtained distance image data (S104). Further, the edge candidate extraction unit 173 determines whether or not the number of distance image data in a range sufficient for edge line extraction has been extracted. If it is determined that the distance image data is not sufficient, the process of S103 is performed again. At this time, the distance range is increased by ΔR, and distance image data corresponding to the distance Ro satisfying the condition of Rd + ΔR to Rd + 2ΔR is extracted. Therefore, the distance range is increased by ΔR until it is determined in S105 that a sufficient number of distance image data has been extracted, and the operations of S103 to S105 are repeated (S105).

  When it is determined that a sufficient range of distance image data has been extracted, the edge candidate extraction unit 173 extracts edge lines from the edge candidate data extracted in S104.

  The edge line L1 extracted by this extraction process is output to the virtual plane calculation unit 174 (S106).

  The virtual plane calculation unit 174 sets a plurality of virtual plane candidates having the obtained edge line L1 as one side, and outputs the information to the plane determination unit 175 (S107).

The plane determination unit 175 sets solid figures F1, F2, F3... Including a plurality of virtual planes S1, S2, S3... Based on the information output from the virtual plane calculation unit 174, and each set solid figure is S103. The amount of distance image data obtained in step 1 is calculated. It is determined that there is a higher possibility of an actual plane as the plane included in the three-dimensional figure including more distance image data (S108).
As described above, in the present invention, when recognizing an object, first, an edge line of the object is extracted from the acquired distance image data, and a plane having the edge line as one side is extracted. The amount of calculation used to process the distance image data is reduced. Moreover, since the object can be recognized quickly by reducing the amount of calculation, the object can be accurately recognized even when the object moves at high speed.

FIG. 8 shows an example in which the object recognition apparatus according to the present invention is used for detection of an automobile. Presence of a vehicle passing through a predetermined range by irradiating a laser from the laser irradiation unit 5 of the object recognition device provided to irradiate a laser on a range where the vehicle passes the road to recognize the plane a and the plane b And the outer shape can be recognized immediately. In addition, it is possible to grasp the position of the automobile by installing the object recognition apparatus in a parking lot. Furthermore, the object recognition apparatus can perform intrusion and position determination of an airplane at an airfield or airport, and can also be used for defense applications such as detection of a vehicle such as a tank in a specific place.

  The object recognition apparatus of a present Example is represented with the block diagram of FIG. Control unit 21, laser emitting unit 22, scanning optical system 23, photoelectric conversion unit 24, ranging circuit unit 25, incident light level measurement unit 26, coordinate conversion unit 271, distance image data extraction unit 272, edge candidate extraction unit 273 and the virtual plane calculation unit 274 are the control unit 11, the laser emitting unit 12, the scanning optical system 13, the optical-electrical conversion unit 14, the distance measuring circuit unit 15, the incident light level measurement unit 16, and the coordinate conversion in the first embodiment, respectively. Since the functions are the same as those of the unit 171, the distance image data extraction unit 172, the edge candidate extraction unit 173, and the virtual plane calculation unit 174, description thereof is omitted to avoid redundancy. In the present embodiment, in addition to the configuration of the first embodiment, a three-dimensional graphic creation unit 276 is provided. Thereby, since the data processing unit 27 outputs a three-dimensional image composed of a plurality of surfaces, the object can be specified more accurately.

  The plane determination unit 275 outputs information about the plane extracted by determining that there is a high possibility of an existing plane to the three-dimensional figure creation unit 276.

  The three-dimensional figure creation unit 276 stores the plane information input from the plane determination unit 275 in a recording unit (not shown). When a plurality of planes are stored, a three-dimensional image obtained by combining the plurality of planes is created and output.

  Next, a flowchart in the present embodiment is shown in FIG. Steps S201 to S208 are the same as steps S101 to S108 in the first embodiment, respectively, and the description thereof is omitted to avoid redundancy.

  The three-dimensional graphic creation unit 276 stores the plane extracted in S208, and when a plurality of planes are stored, creates a three-dimensional graphic combining these planes (S209).

FIG. 11 is an example in which a vehicle is detected using the object recognition apparatus in the present embodiment. In the object recognition apparatus according to the present embodiment, for example, a three-dimensional image composed of the plane c, the plane d, and the plane e of the object 6 is extracted from the object. The extracted three-dimensional image information is grasped as shown in FIG. By obtaining this three-dimensional image, it becomes possible to roughly assume the shape of the entire object, and the accuracy of recognizing the object is improved.

  The object recognition apparatus of a present Example is represented with the block diagram of FIG. Control unit 31, laser emitting unit 32, scanning optical system 33, photoelectric conversion unit 34, distance measuring circuit unit 35, incident light level measuring unit 36, coordinate conversion unit 371, distance image data extraction unit 372, edge candidate extraction unit 373 and virtual plane calculation unit 374 are the control unit 11, laser emitting unit 12, scanning optical system 13, photoelectric conversion unit 14, ranging circuit unit 15, incident light level measurement unit 16, coordinates in Example 1, respectively. Since it has the same function as the conversion unit 171, the distance image data extraction unit 172, the edge candidate extraction unit 173, and the virtual plane calculation unit 174, description thereof is omitted to avoid redundancy. In the present embodiment, in addition to the configuration of the first embodiment, a planar template storage unit 377 and a planar template matching unit 378 are provided. As a result, the plane obtained by acquiring the data and the template prepared in advance can be compared to identify the object, and in particular, the size of the plane obtained from the measurement of the object and the direction of the object When it is known, it is possible to quickly identify an object.

  The plane determination unit 375 outputs the plane information extracted by determining that the possibility of an existing plane is high to the plane template matching unit 378.

  The planar template storage unit 377 includes one or a plurality of templates, which are assumed in advance as the plane of the object, depending on the type of the object, in the storage unit, and outputs the template to the planar template matching unit 378.

  The plane template collation unit 378 collates the input plane information with the template, and outputs the plane information determined to match.

  As described above, according to the object recognition device of the present embodiment, the object is identified by collating a plane corresponding to a part of the object outline with the template, so that the object is recognized from the entire object outline. In addition, it is possible to reduce the amount of calculation required for the process for recognizing the object.

  Next, a flowchart in the present embodiment is shown in FIG. The steps S301 to S308 are the same as the steps S101 to S108 in the first embodiment, respectively, and the description thereof is omitted to avoid redundancy.

  The plane template collation unit 378 collates the plane extracted in S308 with a template prepared in advance, and outputs information on the plane determined to match. (S310).

  FIG. 15 is an example of an apparatus for detecting an automobile using the object recognition apparatus in the present embodiment. In this apparatus, for example, it is assumed that the plane f and the plane g can be extracted from the measurement point data of the object 6. These two planes are collated with templates α, β, and γ prepared according to the difference in vehicle type as shown in FIG. 16, and it is determined that the plane f and the plane g coincide with the template γ. At this time, the vehicle type can be recognized by the size and position of the template γ.

For example, when a vehicle is recognized by the object recognition apparatus according to the present embodiment, the range of the size of the vehicle can generally be assumed in advance without recognizing the entire outline of the automobile and extracting the outline of the automobile that is the object. So, for example, for the two planes in front of the outline of the car, the size, depth, spacing, etc. of the plane are compared with the template data prepared in advance, whether there is a car, Classification can be performed, and the number of vehicles entering and leaving the parking lot, information on the types of vehicles entering and leaving the vehicle, and the like can be quickly grasped.

  The configuration of the fourth embodiment will be described with reference to FIG. Laser emitting unit 42, scanning optical system 43, photoelectric conversion unit 44, distance measuring circuit unit 45, incident light level measuring unit 46, distance image data extracting unit 472, edge candidate extracting unit 473, virtual plane calculating unit 474, plane The determination unit 475 is the laser emitting unit 12, the scanning optical system 13, the optical-electrical conversion unit 14, the distance measuring circuit unit 15, the incident light level measuring unit 16, the distance image data extracting unit 172, and the edge candidate in the first embodiment. Since it has the same function as the extraction unit 173, the virtual plane calculation unit 174, and the plane determination unit 175, its description is omitted to avoid redundancy. In this embodiment, in addition to the configuration of the first embodiment, an inertial sensor 48 and a GPS 49 are provided. In the first to third embodiments, the position of the object recognition device is fixed. However, in this embodiment, the object recognition device is mounted on a moving body, and the position changes. According to the present embodiment, since the position and orientation of the device itself can be grasped by the inertial sensor 48 and the GPS 49, the object can be quickly recognized even when the object recognition device moves at a high speed. .

  The control unit 41 outputs, to the inertial sensor 48 and the GPS 49, in addition to the laser emitting unit 42 and the distance measuring circuit unit 45, the laser emitting unit 42 instructs the laser emitting unit 42 to emit the pulse laser.

  The inertial sensor 48 converts the rotation angles (κo, φo, ωo) of the three axes of roll, pitch, and yaw at the time of the input launch time signal To that represents the attitude of the object recognition device to the coordinate conversion of the data processing unit 47. Output to the unit 471.

  The GPS 49 outputs the position (xo, yo, zo) of the object recognition device at the time represented by the input launch time signal To to the coordinate conversion unit 471.

  The coordinate conversion unit 471 inputs the distance Ro, the firing angle (Φo, Ψo) at the time of laser pulse emission, the position (xo, yo, zo) and the posture (κo, φo, ωo) of the object recognition device. Further, the position (xi, yi, zi) and posture (κi, φi, ωi) of the object recognition device at the time of incidence of the laser pulse are input. Then, as shown in FIG. 18, the acquired distance image data (Ro, Φo, Ψo) is converted into the position (xo, yo, zo) and posture (κo, φo, ωo) of the object recognition device at the time of laser pulse emission. The position (xi, yi, zi) and the posture (κi, φi, ωi) of the object recognition device at the time of laser pulse incidence are corrected by the existing method and coordinate conversion is performed. The reflection position (Xo, Yo, Zo) is obtained.

  Subsequently, the flowchart in the present embodiment is the same as the flowchart in the first embodiment. However, for S402, the coordinate conversion unit 471 uses the acquired distance image data, and the coordinate conversion unit 471 uses the acquired distance image data (Ro, Φo, Ψo) as the position (xo, yo, zo) and posture (κo, φo, ωo) and the position (xi, yi, zi) and posture (κi, φi, ωi) of the object recognition device at the time of incidence of the laser pulse. The reflection position (Xo, Yo, Zo) of the pulse laser in Cartesian coordinates is obtained by correcting and converting the coordinates.

Further, as shown in FIG. 19, the data processing unit 57 may include a three-dimensional figure creation unit 576 as in the second embodiment. Alternatively, as illustrated in FIG. 20, the data processing unit 67 may include a planar template storage unit 677 and a planar template matching unit 678 as in the third embodiment.

  The configuration of the fifth embodiment will be described with reference to FIG. Control unit 71, laser emitting unit 72, scanning optical system 73, photoelectric conversion unit 74, incident light level measurement unit 76, distance image data extraction unit 772, edge candidate extraction unit 773, virtual plane calculation unit 774, plane determination unit Reference numeral 775 denotes a control unit 11, a laser emitting unit 12, a scanning optical system 13, an incident light level measurement unit 16, a distance image data extraction unit 172, an edge candidate extraction unit 173, a virtual plane calculation unit 174, and a plane in the first embodiment. The function is the same as that of the determination unit 175, and the description thereof is omitted to avoid redundancy. In this embodiment, an incident time measuring section 75 is provided instead of the distance measuring circuit section 15 in the first embodiment, and a coordinate converter 771 obtains a distance Ro to the object.

  The photoelectric conversion unit 74 converts the input pulse laser Pi into an electrical pulse Ei, and outputs the electrical pulse Ei to the incident light level measurement unit 76 and the incident time measurement unit 75.

  The incident time measuring unit 75 obtains the incident time signal Ti of the pulse laser Pi from the incident electric pulse Ei. The obtained incident time is output to the coordinate conversion unit 771 of the data processing unit 77.

  The coordinate conversion unit 771 obtains the distance Ro from the incident time signal Ti and the firing time signal To input from the control unit 71 by a calculation formula of (Ti−To) × c / 2, and is input in the same manner as in the first embodiment. Distance image data (Ro, Φo, Ψo) is converted from polar coordinates (Ro, Φo, Ψo) to Cartesian coordinates (Xo, Yo, Zo) represented by Xo = Rocos ΨosinΦo, Yo = Rocos ΨocosΦo, and Zo = Rosin Ψo. As shown in FIG. 22, the data processing unit 87 may include a three-dimensional graphic creation unit 876 as in the second embodiment. Alternatively, as shown in FIG. 23, the data processing unit 97 may include a planar template storage unit 977 and a planar template matching unit 978 as in the third embodiment.

  In all the above-described embodiments, the object recognition apparatus continuously emits a pulse laser while changing the emission angle (Φo, Ψo), and the reflected pulse laser is incident, whereby the distance of the entire object is detected. Although it is configured to acquire image data, the present invention is not limited to this, and as shown in FIG. 24, the entire object is irradiated with light such as a laser at one time, and a plurality of measurement point data are acquired simultaneously. Any configuration may be used.

In the fifth embodiment, a GPS and an inertial sensor may be added to the above-described configuration. In addition, when extracting the distance image data as the edge candidate, the edge candidate extraction unit compared the ratio of the pulsed laser emission light level to the incident light level instead of extracting the distance image data based on the incident light level. The distance image data may be extracted based on the reflectance. In all the embodiments, the order of the distance image data extraction unit and the coordinate conversion unit is reversed. First, the distance image data corresponding to a specific distance range is extracted, and then the extracted data is changed from polar coordinates to Cartesian coordinates. You may make it convert. As described above, the present invention is not limited to the above-described embodiments, and modifications of the embodiments within the scope conceivable by those skilled in the art are also included in the concept of the present invention.

The block diagram explaining the target object recognition apparatus in a 1st Example. Diagram showing the relationship between polar coordinates and Cartesian coordinates A diagram showing how a laser beam is irradiated onto an object A diagram showing how a laser beam is irradiated onto an object Diagram showing how edge lines are extracted A figure showing how to extract the plane of an object The flowchart explaining the target object recognition method in 1st Example. The figure explaining the target object recognition apparatus in a 1st Example. The block diagram explaining the target object recognition apparatus in 2nd Example. The flowchart explaining the target object recognition method in 2nd Example. The figure explaining the target object recognition apparatus in 2nd Example. The figure of the plane obtained by the target object recognition apparatus in 2nd Example The block diagram explaining the target object recognition apparatus in 3rd Example. The flowchart explaining the target object recognition method in 3rd Example. The figure explaining the target object recognition apparatus in 3rd Example. The figure explaining the mode of template collation in a 3rd Example The block diagram explaining the target object recognition apparatus in a 4th Example. A diagram showing the relationship between polar coordinates and Cartesian coordinates, and the position and orientation of the object recognition device The block diagram explaining the target object recognition apparatus in a 4th Example. The block diagram explaining the target object recognition apparatus in a 4th Example. Block diagram for explaining an object recognition apparatus in the fifth embodiment Block diagram for explaining an object recognition apparatus in the fifth embodiment Block diagram for explaining an object recognition apparatus in the fifth embodiment Figure of target object recognition device irradiating target object with beam Figure of target object recognition device irradiating target object with beam

Explanation of symbols

1, 5, 7, 8, 9: Object recognition device 2, 3, 6: Object 4: Laser 11, 21, 31, 41, 51, 61, 71, 81, 91: Control unit 12, 22, 32 42, 52, 62, 72, 82, 92: Laser emitting units 13, 23, 33, 43, 53, 63, 73, 83, 93: Scanning optical systems 14, 24, 34, 44, 54, 64, 74 , 84, 94: photoelectric conversion units 15, 25, 35, 45, 55, 65: ranging circuit units 16, 26, 36, 46, 56, 66, 76, 86, 96: incident light level measuring unit 17 27, 37, 47, 57, 67, 77, 87, 97: Data processing units 171, 271, 371, 471, 571, 671, 771, 871, 971: Coordinate conversion units 172, 272, 372, 472, 572 , 672, 772, 872, 972: Distance image Data extraction unit 173, 273, 373, 473, 573, 673, 773, 873, 973: edge candidate extraction unit 174, 274, 374, 474, 574, 674, 774, 874, 974; virtual plane calculation unit 175, 275, 375, 474, 575, 675, 775, 875, 975: plane determination units 276, 576, 876: three-dimensional figure creation units 377, 677, 977: plane template storage units 378, 678, 978: plane template matching unit 48, 58, 68: Inertial sensors 49, 59, 69: GPS
75, 85, 95: Incident time measuring unit

Claims (1)

Launching means for launching light onto the object;
An incident means on which the light reflected by the object is incident;
Incident light level measuring means for measuring the level of the incident light;
Distance image data acquisition means for acquiring a plurality of distance image data representing the position of the object from the incident light;
A first extraction means for extracting the distance image data satisfying a first condition from the plurality of the obtained distance image data;
Second extraction means for extracting information related to an edge of the object based on the distance image data and the level satisfying the extracted first condition;
Recognizing means for recognizing the object based on the information on the edge, and setting means for setting a virtual plane based on the information on the edge;
Of the virtual planes set by the setting means, comprising plane extraction means for extracting a plane that is a virtual plane that satisfies a third condition,
Before SL recognition means may recognize the object based on the plane that has been extracted in the plane extraction means,
The object recognition apparatus in which the third condition is determined based on the number of the distance image data included in the solid figure including the virtual plane.

Images