Target detection device and method
The target object detection device improves accuracy by extracting opposite ends of a target in a horizontal direction and generating a frame around a representative point, addressing the issue of incorrect end identification in existing systems.
Patent Information
- Application Number
- JP2021001957
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2021-01-08
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2041-01-08
AI Technical Summary
Existing target object detection systems risk inaccurately identifying the rear end of a target as the front end, leading to incorrect positioning and shape detection of the target.
A target object detection device that extracts opposite ends of a target in a horizontal direction from an image, sets a representative point between these ends, and generates a frame around this point to accurately determine the target's position, using an imaging unit, edge extraction, and coordinate detection units.
The device accurately detects the target's position without mistaking the front end for the rear end, enhancing detection precision.
Smart Images

Figure 0007593116000001 
Figure 0007593116000002 
Figure 0007593116000003
Abstract
Description
[Technical field]
[0001] The present invention relates to a target detection device and a target detection method that are mounted on a moving body and detect objects existing around the moving body as targets. [Background technology]
[0002] In recent years, there is a demand for recognition of surrounding conditions due to demands for safety technology and autonomous driving technology. In particular, recognition of surrounding conditions is important in cases of collision avoidance and lane changing in moving bodies such as vehicles and robots. For this reason, moving bodies are often equipped with object detection devices (target detection devices) that detect objects in order to recognize surrounding conditions (see, for example, Patent Document 1). [Prior art documents] [Patent documents]
[0003] [Patent Document 1] Special Publication No. 2019-526781 Summary of the Invention [Problem to be solved by the invention]
[0004] However, when detecting a target based on the detection results of a radar device, identifying the rear end from among the target mainly from the perspective of collision avoidance, and detecting the position and shape of the target by generating a rectangular frame forward from this identified rear end as the presence area of the target, there is a risk that the front end will be mistaken for the rear end when identifying the rear end, and the detected position of the target will differ from the actual position of the object.
[0005] The present invention has been made in consideration of the above circumstances, and an object of the present invention is to provide a target detection device and a target detection method that can detect the position of a target with higher accuracy. [Means for solving the problem]
[0006] As a result of various investigations, the inventors of the present invention have found that the above object can be achieved by the present invention described below. That is, a target detection device according to one aspect of the present invention is a device mounted on a moving body and detecting an object existing around the moving body as a target, comprising: an imaging unit that generates an image; an end extraction unit that extracts a first one end and a first other end of a target that are opposed to each other in a horizontal direction from the image generated by the imaging unit; a coordinate detection unit that sets points at the first one end and the first other end extracted by the end extraction unit that are the same distance from a second one end or a second other end in a vertical direction as a first one end point and a first other end point, respectively, and sets a target representative point that represents a position of the target between the set first one end point and first other end point, and determines coordinates of the target representative point in a predetermined coordinate system; and a target region generation unit that generates a predetermined frame centered on the target representative point set by the coordinate detection unit as an existence region of the target. Preferably, in the above-mentioned target detection device, the coordinate detection unit includes a representative point setting unit that sets points at the first one end and the first other end extracted by the end extraction unit that are at the same distance from the second one end or the second other end in the vertical direction as a first one end point and a first other end point, and sets a target representative point that represents a position of the target between the set first one end point and the first other end point, a measurement unit that measures each distance and each direction from the host aircraft at each of the first one end point and the first other end point set by the representative point setting unit, and a coordinate processing unit that determines each coordinate of the first one end point, the first other end point and the target representative point in a predetermined coordinate system based on the each distance and each direction from the host aircraft at each of the first one end point and the first other end point measured by the measurement unit. Preferably, the measurement unit is a radar device. Preferably, in the above-mentioned target object detection device, the predetermined frame is a rectangle, and the predetermined frame is generated so that a target representative point set by the coordinate detection unit coincides with an intersection of diagonals in the rectangle.
[0007] This type of target detection device does not identify the rear end of the target, but instead extracts a first one end and a first other end of the target that are opposed to each other in the horizontal direction based on an image, sets a target representative point representing the position of the target between the extracted first one end and first other end, and generates a predetermined frame as the presence area of the target, centered on this set target representative point.Therefore, there is no risk of mistaking the front end for the rear end, and the position of the target can be detected more accurately.
[0008] In another aspect, in the above-described target detection device, the coordinate detection unit sets the target representative point at a center position between the set first one end point and the set first other end point.
[0009] The line segment connecting the first one end point and the first other end point may be a diagonal line of the target, and in this case, the first one end point and the first other end point are considered to be both ends of the diagonal line of the target. Since the target detection device sets the target representative point at the center position between the first one end point and the first other end point, the target representative point can be set at the center position (center position) of the target, and therefore the position of the target can be detected with higher accuracy.
[0010] In another aspect, in the above-mentioned target detection device, the coordinate detection unit determines a point of the target that is closest to the aircraft between the first one end and the first other end as a closest point, and sets a point that is line-symmetric to the closest point with a line segment connecting the first one end point and the first other end point as an axis of symmetry as a virtual point, and the target area generation unit generates, as the predetermined frame, a rectangle whose vertices are the first one end point, the first other end point, the closest point, and the virtual point set by the coordinate detection unit.
[0011] Such a target detection device can detect the planar shape of the target by generating, as the predetermined frame, a rectangle having the first one end point, the first other end point, the nearest point, and the imaginary point as vertices.
[0012] A target detection method according to another aspect of the present invention is used in a moving body and detects objects existing around the moving body as targets, and includes an imaging process for generating an image, an end extraction process for extracting a first one end and a first other end of the target that are opposed to each other in a horizontal direction from the image generated in the imaging process, a coordinate detection process for setting points at the first one end and the first other end extracted in the end extraction process that are the same distance from a second one end or a second other end in a vertical direction as a first one end point and a first other end point, setting a target representative point that represents a position of the target between the set first one end point and first other end point, and determining the coordinates of the target representative point in a predetermined coordinate system, and a target area generation process for generating a predetermined frame centered on the target representative point set in the coordinate detection process as an area in which the target exists.
[0013] This type of target detection method does not identify the rear end of the target, but extracts a first one end and a first other end of the target that are opposed to each other in the horizontal direction based on an image, sets a target representative point representing the position of the target between the extracted first one end and first other end, and generates a predetermined frame as the presence area of the target centered on this set target representative point.Therefore, there is no risk of mistaking the front end for the rear end, and the target position can be detected more accurately. Effect of the Invention
[0014] The target detection device and the target detection method according to the present invention can detect the position of a target with higher accuracy. [Brief description of the drawings]
[0015] [Figure 1] 1 is a block diagram showing a configuration of a target detection device according to an embodiment; [Diagram 2] 3 is a diagram for explaining the positions of an imaging unit and a measurement unit in the target detection device. FIG. [Diagram 3] 3 is a diagram for explaining a method of detecting target information in the target detection device; FIG. [Figure 4] 4 is a flowchart showing an operation of the target detection device. [Diagram 5] FIG. 13 is a diagram for explaining a comparative example. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0016] Hereinafter, one or more embodiments of the present invention will be described with reference to the drawings. However, the scope of the invention is not limited to the disclosed embodiments. In addition, configurations with the same reference numerals in each drawing indicate that they are the same configurations, and their description will be omitted as appropriate. In this specification, when referring to a general term, a reference numeral without a subscript is used, and when referring to an individual configuration, a reference numeral with a subscript is used.
[0017] The target detection device in the embodiment is a device mounted on a moving body and detects objects (objects visible from the moving body within a predetermined range) existing around the moving body as targets. The target detection device includes an imaging unit that generates an image, an end extraction unit that extracts a first one end and a first other end of a target that are opposed to each other in the horizontal direction from the image generated by the imaging unit, a coordinate detection unit that sets each point at the first one end and the first other end extracted by the end extraction unit at the same distance from a second one end or a second other end in the vertical direction as a first one end point and a first other end point, sets a target representative point representing the position of the target between the set first one end point and the set first other end point, and obtains coordinates of the target representative point in a predetermined coordinate system, and a target region generation unit that sets a predetermined frame centered on the target representative point set by the coordinate detection unit as an existence region of the target. Such a target detection device will be described in more detail below.
[0018] Fig. 1 is a block diagram showing a configuration of a target detection device in an embodiment. Fig. 2 is a diagram for explaining the arrangement positions of an imaging unit and a measurement unit in the target detection device. Fig. 3 is a diagram for explaining a method of detecting target information in the target detection device. Fig. 3A is a plan view showing a schematic diagram of a situation of a host vehicle and another vehicle traveling on a road with one lane in each direction. Fig. 3B is a diagram showing a schematic diagram of an image generated by imaging another vehicle traveling ahead of the host vehicle shown in Fig. 3A.
[0019] A target detection device D in the embodiment includes an imaging unit 1, a coordinate detection unit 2, a control processing unit 3, and a storage unit 4, as shown in FIG.
[0020] The imaging unit 1 is mounted on a moving body, connected to the control processing unit 3, and is a device that captures an image of the surroundings of the target detection device D in a predetermined imaging range and generates an image thereof according to the control of the control processing unit 3. The imaging unit 1 outputs the generated image to the control processing unit 3. The moving body is a device that can change its own position, such as a vehicle or a robot. The vehicle is, for example, a parts transport vehicle in a factory and an automobile. The imaging unit 1 is mounted on a vehicle so as to capture an image of the front (traveling direction, traveling direction). For example, in this embodiment, as shown in FIG. 2, the imaging unit 1 is disposed in the vehicle VC (hereinafter, for convenience of explanation, in order to distinguish it from the preceding vehicle, appropriately) on the ceiling surface (inner surface of the roof) near the windshield, with its imaging direction (optical axis direction) facing diagonally downward so as to capture an image of the front of the vehicle VC in a predetermined imaging range (angle of view) R1, including the road surface, within the vehicle VC. The imaging unit 1 is, for example, a digital camera including an imaging optical system that forms an optical image of a subject (object) on a predetermined imaging plane, an area image sensor that is arranged with its light receiving surface aligned with the imaging plane and converts the optical image of the subject into an electrical signal, and an image processing unit that processes the output of the area image sensor to generate image data that represents an image of the subject, etc. The digital camera may be a visible light camera or an infrared camera.
[0021] The coordinate detection unit 2 is a device that sets each point at the first one end and the first other end, which are extracted by the end extraction unit 32 described later, at the same distance from the second one end or the second other end in the vertical direction as the first one end point and the first other end point, sets a target representative point representing the position of the target between the set first one end point and the first other end point, and obtains the coordinates of the target representative point in a predetermined coordinate system. In this embodiment, the coordinate detection unit 2 sets the target representative point at the center position between the set first one end point and the first other end point. That is, the target representative point is a point that divides the line segment connecting the first one end point and the first other end point into two equal parts. The coordinate detection unit 2 includes, for example, a measurement unit 21, a representative point setting unit 22 (33), and a coordinate processing unit 23 (34). Such a coordinate detection unit 2 will be described in more detail later.
[0022] The storage unit 4 is connected to the control processing unit 3, and is a circuit that stores various predetermined programs and various predetermined data under the control of the control processing unit 3. The various predetermined programs include, for example, a control program for controlling each of the units 1, 2, and 4 of the target detection device D in accordance with the function of each unit; an edge extraction program for extracting a first one end and a first other end of a target that are opposed to each other in the horizontal direction from an image generated by the imaging unit 1; a representative point setting program for setting each point at the first one end and the first other end extracted by the edge extraction program at the same distance from the second one end or the second other end in the vertical direction as a first one end point and a first other end point, and setting a target representative point that represents the position of the target between the set first one end point and the first other end point; a coordinate processing program for determining the coordinates of the first one end point, the first other end point, and the target representative point in a predetermined coordinate system based on the distances and directions from the vehicle at each of the first one end point and the first other end point measured by the measurement unit 21; and a target area generation program for generating a predetermined frame centered on the target representative point set by the representative point setting program as the existence area of the target. The various types of predetermined data include data necessary for executing each program, such as the shape of the predetermined frame. The storage unit 4 includes, for example, a ROM (Read Only Memory) which is a non-volatile storage element, an EEPROM (Electrically Erasable Programmable Read Only Memory) which is a rewritable non-volatile storage element, and the like. The storage unit 4 also includes a RAM (Random Access Memory) which serves as a so-called working memory of the control processing unit 3 and stores data generated during the execution of the predetermined programs.
[0023] The control processing unit 3 is a circuit for controlling each of the units 1, 2, and 4 of the target detection device D in accordance with the function of each unit, and detecting objects existing around the vehicle VC as targets. The control processing unit 3 is configured to include, for example, a CPU (Central Processing Unit) and its peripheral circuits. By executing the control processing program, the control processing unit 3 functionally includes a control unit 31, an edge extraction unit 32, a representative point setting unit 33 (22), a coordinate processing unit 34 (23), and a target area generation unit 35.
[0024] The control unit 31 controls each of the units 1, 2, and 4 of the target detection device D in accordance with the function of each unit, and is in charge of overall control of the target detection device D.
[0025] The edge extraction unit 32 extracts a first one end and a first other end that face each other in the horizontal direction of a target from an image generated by the imaging unit 1. The edge extraction unit 32 includes, for example, a pre-processing unit 321 and an extraction unit 322. The pre-processing unit 321 and the extraction unit 322 are functionally configured in the control processing unit 3.
[0026] The pre-processing unit 321 generates a silhouette image by filling the image area of a target (predetermined object) with a single color from the image generated by the imaging unit 1. For example, a known target detection technique used in collision detection can be used for such image processing. Alternatively, the pre-processing unit 321 may be configured to include a machine learning model that has been machine-learned to generate a silhouette image by filling the image area of a target with a single color from the image generated by the imaging unit 1. For example, a neural network used in deep learning such as a convolutional neural network (CNN) can be used for such a machine learning model. Alternatively, the pre-processing unit 321 may be configured to extract edges by applying an edge filter to the image generated by the imaging unit 1, extract straight lines by applying a Hough transform of straight lines to the extracted edges, and generate a silhouette image by filling the approximately rectangular shape formed by the extracted straight lines with a single color as the image area of the target. Note that an image may be generated in which the approximately rectangular shape formed by the extracted straight lines is superimposed on the image generated by the imaging unit 1.
[0027] The extraction unit 322 extracts a first one end and a first other end of a target, which are opposed to each other in the horizontal direction, from the silhouette image generated by the pre-processing unit 321. In the silhouette image, the image area of the target is filled with a single color, so that the extraction unit 322 can easily extract both ends that are opposed to each other in the horizontal direction in the image area of the target as the first one end and the first other end, respectively.
[0028] For example, as shown in Fig. 3A, when the host vehicle VC, the first other vehicle Ob1, and the second other vehicle Ob2 are traveling on a road with one lane in each direction, the first other vehicle Ob1 is traveling ahead of the host vehicle VC in the same lane as the host vehicle VC, and the second other vehicle Ob2 is traveling diagonally ahead of the host vehicle VC in the oncoming lane, an image area POb of the first other vehicle Ob1 shown in Fig. 3B is formed in the silhouette image generated by the pre-processing unit 321 of the edge extraction unit 32 from the image generated by the imaging unit 1 of the host vehicle VC. Note that, in Fig. 3B, for convenience of explanation, the image area POb of the first other vehicle Ob1 is not filled in with a single color. In such a case, the extraction unit 322 extracts both ends (the left end and the right end when viewed in a plan view of Figure 3B) EL1 and ER1 facing each other in the horizontal direction in the image area POb of the first other vehicle Ob1 as the first one end (left end) EL1 and the first other end (right end) ER1, respectively.
[0029] The coordinate detection unit 2, which includes the measurement unit 21, the representative point setting unit 22 (33), and the coordinate processing unit 23 (34), will be further described.
[0030] The representative point setting unit 22 (33) sets each point at the first one end and the first other end extracted by the end extraction unit 32, which are at the same distance from the second one end or the second other end in the vertical direction, as a first one end point and a first other end point, and sets a target representative point representing the position of the target between the set first one end point and first other end point. In this embodiment, the representative point setting unit 22 (33) sets the target representative point at a center position between the set first one end point and first other end point. The distance from the second one end or the second other end is set appropriately in advance.
[0031] For example, in the example shown in Fig. 3, as shown in Fig. 3B, the representative point setting unit 22 (33) sets the points PL1, PR2 at the same distance from the second one end (upper end) or the second other end (lower end) in the vertical direction at the one end (left end) EL1 and the other end (right end) ER1 extracted by the end extraction unit 32 as the first one end point (left end point) PL1 and the first other end point (right end point) PR1. Then, the representative point setting unit 22 (33) sets the target representative point PM1 between the set first one end point PL1 and first other end point PR1, in this embodiment, at the center position between the one end point PL1 and the other end point PR1.
[0032] The measurement unit 21 is connected to the control processing unit 3, and is a device that measures the distances and directions from the vehicle (own vehicle VC) at one end point and the other end point set by the representative point setting unit 22 (33) under the control of the control processing unit 3. The measurement unit 21 outputs the measurement results to the control processing unit 3. In this embodiment, the measurement unit 21 is, for example, a radar device that is provided at the front end of the vehicle VC as shown in FIG. 2 and measures an object in front of the vehicle VC in a predetermined measurement range R2. The radar device includes, for example, a transmission unit that transmits a millimeter wave band transmission wave while scanning the measurement range R2, a reception unit that receives a reflected wave of the transmission wave reflected by an object, and a signal processing unit that determines the direction of the object and the distance to the object based on the transmission wave and the reflected wave. The signal processing unit determines the direction of the object from the transmission direction of the transmission wave in scanning the measurement range R2 for each direction (each point of the measurement range R2) by scanning, and determines the distance to the object based on the time difference between the transmission timing of the transmission wave and the reception timing of the reflected wave (TOF (Time-Of-Flight) method). The radar device is not limited to this configuration and may be any appropriate type of radar device. For example, the radar device may be a laser radar device that uses laser light instead of a millimeter wave band, and the radar device may be a device that has multiple receiving antennas and determines the direction of the object from the phase difference of the reflected waves received by the multiple receiving antennas. The measurement unit 21 may also be a stereo camera that generates a distance image from a pair of stereo images.
[0033] The coordinate processing unit 23 (34) determines the coordinates of the one end point, the other end point, and the target representative point in a predetermined coordinate system based on the distances and directions from the vehicle of the vehicle measured by the measurement unit 21 at each of the one end point and the other end point. The predetermined coordinate system (global coordinate system) may be any coordinate system with an arbitrary point as the coordinate origin, such as an XYZ orthogonal coordinate system or an rθφ polar coordinate system. For example, the predetermined coordinate system is an XYZ orthogonal coordinate system with the center position of the vehicle VC as the coordinate origin, the front-rear direction as the X axis, the width direction as the Y axis, and the height direction as the Z axis. The measurement unit 21 performs measurements in its own coordinate system (local coordinate system) with the measurement unit 21 as the coordinate origin, so the coordinate processing unit 23 (34) converts the measurement results of the measurement unit 21 into coordinate values of the XYZ orthogonal coordinate system based on the arrangement position of the measurement unit 21 and the position of the coordinate origin of the XYZ orthogonal coordinate system. Each pixel (each point) of the silhouette image generated by processing the image generated by the imaging unit 1 as described above by the pre-processing unit 321 and the measurement results of the measurement unit 21 (each position represented by direction and distance) are associated with each other by so-called sensor fusion, which combines the image of the camera and the measurement results of the radar device. The coordinate processing unit 23 (34) obtains the coordinates of each of the one end point and the other end point from the measurement results of the measurement unit 21 that are associated by sensor fusion, and obtains the coordinates of the target representative point based on the obtained coordinates of each of the one end point and the other end point.
[0034] For example, in the example shown in Figure 3, the coordinate processing unit 23 (34) determines the coordinates of one end point PL1 and the other end point PR1 from the measurement results of the measurement unit 21 that are associated by sensor fusion, and determines the coordinates of the target representative point PM1 based on the determined coordinates of one end point PL1 and the other end point PR1.
[0035] The target area generating unit 35 generates a predetermined frame centered on the target representative point set by the coordinate detecting unit 2 as the target existence area. The shape and size (magnitude) of the predetermined frame are appropriately set in advance. In this embodiment, the representative point setting unit 22 (33) of the coordinate detecting unit 2 determines the point of the target that is closest to the aircraft as the closest point between the first one end and the first other end. More specifically, the representative point setting unit 22 (33) determines the point of the target that is closest to the aircraft as the closest point, and determines whether or not the determined closest point exists between the first one end and the first other end, thereby determining the closest point between the first one end and the first other end. The representative point setting unit 22 (33) sets a point that is line-symmetrical to the closest point with a line segment connecting the first one end point and the first other end point as an axis of symmetry as a virtual point. The coordinate processing unit 23 (34) of the coordinate detection unit 2 obtains the coordinates of the closest point and the set virtual point. Then, the target area generation unit 35 generates a rectangle having the first one end point, the first other end point, the closest point, and the virtual point set by the coordinate detection unit 2 as vertices, as the predetermined frame.
[0036] For example, in the example shown in FIG. 3, the representative point setting unit 22 (33) determines a point of the first other vehicle Ob1 that is closest to the host vehicle VC between the first one end EL1 and the first other end ER1 as the closest point PC1, as shown in FIG. 3A and FIG. 3B, and sets a point PV1 that is line-symmetric to the closest point PC1 with the line segment α connecting the first one end point PL1 and the first other end point PR1 as the axis of symmetry α as a virtual point PV1, as shown in FIG. 3A, the coordinate processing unit 23 (34) determines the coordinates of the closest point PC1 and the set virtual point PV1, and the target area generation unit 35 generates a rectangle AR1 as the predetermined frame (the presence area of the first other vehicle Ob1) having the first one end point PL1, the first other end point PR1, the closest point PC1, and the virtual point PV1 set by the coordinate detection unit 2 as its vertices, as shown in FIG. 3A.
[0037] The control processing unit 3 and the storage unit 4 can be configured by a computer called an ECU (Electronic Control Unit).
[0038] In the above description, the measurement range R2 of the measurement unit 21 is a surface, but it may also be a straight line along the horizontal direction at a predetermined height (for example, the height at which the measurement unit 21 is disposed), in which case the distance from the second other end (lower end) is set to the predetermined height, and the predetermined coordinate system (global coordinate system) is an XY Cartesian coordinate system.
[0039] Next, the operation of this embodiment will be described with reference to a flowchart shown in FIG 5.
[0040] When the vehicle starts to operate, the target detection device D mounted on such a vehicle executes initialization of each required unit and starts its operation. By executing the control processing program, the control processing unit 3 is functionally configured with a control unit 31, an edge extraction unit 32, a representative point setting unit 33 (22), a coordinate processing unit 34 (23) and a target area generation unit 35, and the edge extraction unit 32 is functionally configured with a pre-processing unit 321 and an extraction unit 322. Then, for example, until the ignition of the vehicle is turned off, each of the processes S1 to S9 and S11 to S13 shown in Fig. 4 below is repeatedly executed at a predetermined sampling interval that is appropriately set in advance, and the target is detected.
[0041] In FIG. 4, the target detection device D causes the imaging unit 1 to generate an image and the measurement unit 21 of the coordinate detection unit 2 to perform measurements using the control processing unit 3, and stores the generated image and the measurement results in the memory unit 4 (S1).
[0042] Next, the target detection device D judges whether or not the first one end (e.g., the left end) of the target has been extracted by the control processing unit 3 (S2). More specifically, the end extraction unit 32 generates a silhouette image from the image generated by the imaging unit 1 by the preprocessing unit 321, and judges whether or not the first one end has been extracted from the silhouette image by the extraction unit 322. If the result of this judgment is that the extraction unit 322 has extracted (could extract) the first one end (Yes), the target detection device D then executes process S3. On the other hand, if the result of the judgment is that the extraction unit 322 has not extracted (could not extract) the first one end (No), the target detection device D returns the process to process S1. That is, the process is restarted from the beginning.
[0043] In this process S3, the target detection device D determines whether or not the first other end (e.g., the right end) of the target has been extracted by the control processing unit 3. More specifically, the end extraction unit 32 determines whether or not the extraction unit 322 has extracted the first other end from the silhouette image. If the result of this determination is that the extraction unit 322 has extracted (could extract) the first other end (Yes), the target detection device D then executes process S4. On the other hand, if the result of the determination is that the extraction unit 322 has not (could not extract) the first other end (No), the target detection device D returns the process to process S1.
[0044] In this process S4, the target detection device D judges whether or not the closest point of the target has been extracted by the control processing unit 3. More specifically, the coordinate detection unit 2 judges whether or not the closest point has been found by the representative point setting unit 22 (33). If the result of this judgment shows that the representative point setting unit 22 (33) has found the closest point (Yes), the target detection device D then executes process S5. On the other hand, if the result of the judgment shows that the representative point setting unit 22 (33) has not found the closest point (No), the target detection device D returns the process to process S1.
[0045] In this process S5, the target detection device D determines the coordinates of the first one-side end point, the first other-side end point, and the closest point by the coordinate detection unit 2. More specifically, the representative point setting unit 22 (33) of the coordinate detection unit 2 sets the points at the first one-side end and the first other-side end extracted by the end extraction unit 32, which are at the same distance from the second other-side end, for example, as the first one-side end point and the first other-side end point, and the coordinate processing unit 23 (34) of the coordinate detection unit 2 determines the coordinates of the first one-side end point, the first other-side end point, and the closest point in a predetermined coordinate system based on the distances and directions from the own aircraft at the first one-side end point, the first other-side end point, and the closest point, respectively, measured by the measurement unit 21.
[0046] After the process S5, the target detection device D judges whether the first one end point and the first other end point are both ends of a diagonal line of a target. This judgment is performed, for example, based on whether the closest point exists between the first one end point and the first other end point. If the closest point exists between the first one end point and the first other end point, the target detection device D judges that the first one end point and the first other end point are both ends of a diagonal line of a target (Yes), and then executes process S7. On the other hand, if the closest point does not exist between the first one end point and the first other end point (for example, if the closest point coincides with the first one end point or the first other end point), the target detection device D judges that the first one end point and the first other end point are not both ends of a diagonal line of a target (No), and then executes process S11.
[0047] In this process S7, the target detection device D sets a target representative point by the coordinate detection unit 2, obtains its coordinates, and then executes process S8. More specifically, the representative point setting unit 22 (33) of the coordinate detection unit 2 sets the target representative point at a center position between the first one end point and the first other end point set in process S5, and the coordinate processing unit 23 (34) of the coordinate detection unit 2 obtains the coordinates of the set target representative point from the coordinates of the first one end point and the first other end point obtained in process S5.
[0048] In this process S8, the target detection device D sets a virtual point by the coordinate detection unit 2, obtains its coordinates, and then executes process S9. More specifically, the representative point setting unit 22 (33) of the coordinate detection unit 2 sets a point that is line-symmetric to the closest point with respect to the line segment connecting the first one end point and the first other end point set in process S5 as an axis of symmetry, as a virtual point, and the coordinate processing unit 23 (34) of the coordinate detection unit 2 obtains the coordinates of this set virtual point from the coordinates of the first one end point, the first other end point, and the closest point obtained in process S5.
[0049] In this process S9, the target detection device D generates the predetermined frame as the target existence area by the control processing unit 3, and ends this process. More specifically, the target area generating unit 35 generates, as the predetermined frame, a rectangle having the first one end point, the first other end point, the nearest point, and the virtual point as vertices.
[0050] On the other hand, in the process S11, similarly to the process S7, the target detection device D sets a target representative point by the control processing unit 3, obtains its coordinates, and then executes the process S12. More specifically, the control processing unit 3 sets the target representative point at a center position between the first one end point and the first other end point set in the process S5, and obtains the coordinates of the set target representative point from the coordinates of the first one end point and the first other end point obtained in the process S5. In this case, since the host vehicle VC is traveling directly behind the other vehicle Ob, the target representative point is set at the center position of the rear end of the other vehicle Ob.
[0051] In this process S12, in order to determine the type of the other vehicle Ob, such as a regular car or a large vehicle, the target detection device D causes the imaging unit 1 to generate an image by the control processing unit 3, determines the type of the other vehicle Ob based on the generated image, and then executes process S13. For example, a second machine learning model that has been machine-learned to determine the type of vehicle from the image generated by the imaging unit 1 is used to determine the type of vehicle in this process S12. Note that in this process S12, the image generated by the imaging unit 1 in process S1 may be used without causing the imaging unit 1 to generate an image.
[0052] In this process S13, the target detection device D, by using the control processing unit 3, generates a rectangular frame of a size (dimension) according to the vehicle type determined in process S12, by aligning the center position of the rear end short side of the rectangular frame with the target representative point obtained in process S11, and generates the target existence area from the target representative point forward in the traveling direction, and ends this process. For example, if the vehicle type is a standard car, a rectangular frame with a width (short side) of 1.8 [m] and a length (long side) of 3.5 [m] is generated. Also, for example, if the vehicle type is a large vehicle, a rectangular frame with a width of 2.5 [m] and a length of 6.0 [m] is generated.
[0053] If multiple targets are present in the image (silhouette image) generated in process S1, each of the processes S1 to S9 and S11 to S13 is executed for each target. Only a vehicle may be selected as the target for which each of the processes S1 to S9 and S11 to S13 is executed. For example, a third machine learning model that has been machine-learned to extract vehicles from the image generated by the imaging unit 1 is used to select the vehicle.
[0054] As described above, the target detection device D in the embodiment and the target detection method implemented therein do not identify the rear end of the target, but instead extract a first one end and a first other end that are opposed to each other in the horizontal direction of the target based on an image, set a target representative point that represents the position of the target between the extracted first one end and first other end, and generate a predetermined frame as the presence area of the target centered on this set target representative point, so there is no risk of mistaking the front end for the rear end, and the position of the target can be detected with greater accuracy.
[0055] FIG. 5 is a diagram for explaining a comparative example. As a comparative example, when detecting targets based on the detection results of a radar device, identifying the rear end from among them mainly from the viewpoint of collision avoidance, and generating a rectangular frame in the forward direction of travel from the identified rear end as the target's existence area to detect the position and shape of the target, as shown in FIG. 5, if the rear end BL2 of an oncoming vehicle Ob4 can be correctly identified, a rectangular frame AR4 is generated in the forward direction from the rear end BL4, and the position of the detected target (the position of the rectangular frame AR4) coincides with the actual position of the oncoming vehicle Ob4, and the target can be correctly detected. However, as shown in FIG. 5, when identifying the rear end of a preceding vehicle Ob3, there is a risk that the front end BL1 is mistaken for the rear end, and if such an error is made, a rectangular frame AR3 is generated in the forward direction from the front end BL1, and the position of the detected target (the position of the rectangular frame AR3) differs from the actual position of the preceding vehicle Ob3.
[0056] On the other hand, the above-mentioned target detection device D and target detection method extract the first one end and the first other end of the target based on an image, and set a target representative point between them to generate a predetermined frame, so that there is no risk of mistaking the front end BL1 as shown in Figure 5 for the rear end, and the position of the target can be detected with greater accuracy.
[0057] The line segment connecting the first one end point and the first other end point may be a diagonal line of the target, and in this case, the first one end point and the first other end point are considered to be both end points of the diagonal line of the target. Since the target detection device D and the target detection method set the target representative point at the center position between the first one end point and the first other end point, the target representative point can be set at the center position (center position) of the target, and therefore the position of the target can be detected with higher accuracy.
[0058] The target detection device D and the target detection method described above can detect the planar shape of the target by generating a rectangle having the first one end point, the first other end point, the nearest point, and the virtual point as vertices as the specified frame.
[0059] In order to express the present invention, the present invention has been described adequately and sufficiently through the embodiments with reference to the drawings in the above description, but it should be recognized that those skilled in the art can easily change and / or improve the above-mentioned embodiments. Therefore, unless the changes or improvements made by those skilled in the art are at a level that departs from the scope of the claims described in the claims, the changes or improvements are interpreted as being included in the scope of the claims. [Explanation of symbols]
[0060] D Target detection device 1. Imaging unit 2 Coordinate detection section 3. Control Processing Section 4 Storage section 21 Measuring part 22(33) Representative point setting section 23(34) Coordinate processing section 31 Control Unit 32 Edge extraction part 35 Target area generation unit 321 Pretreatment section 322 Extraction part
Claims
1. A target detection device that is mounted on a moving body and detects an object existing around the moving body as a target, an imaging unit that generates an image; an end extraction unit that extracts a first one end and a first other end that are opposed to each other in a horizontal direction of a target from the image generated by the imaging unit; a coordinate detection unit that sets, as a first one end point and a first other end point, points at the first one end point and the first other end point that are at the same distance from a second one end point or a second other end point in a vertical direction at each of the first one end point and the first other end point extracted by the end extraction unit, sets a target representative point that represents a position of the target between the set first one end point and the first other end point, and determines coordinates of the target representative point in a predetermined coordinate system; a target region generating unit that generates a predetermined frame having a center at the target representative point set by the coordinate detection unit as a target existence region, the coordinate detection unit includes a measurement unit that measures a distance from the aircraft to the target; the coordinate detection unit uses the measurement unit to determine a point of the target that is closest to the aircraft as a closest point between the first one end and the first other end, and sets a point that is point-symmetric to the closest point with respect to the target representative point as a virtual point; the target region generating unit generates, as the predetermined frame, a rectangle having the first one end point, the first other end point, the closest point, and the virtual point set by the coordinate detecting unit as vertices; Target detection device.
2. the coordinate detection unit sets the target representative point at a center position between the set first one end point and the set first other end point. The target detection device according to claim 1 .
3. A target detection method for use in a moving body, which detects an object existing around the moving body as a target, comprising: an imaging step for generating an image; an end extraction step of extracting a first one end and a first other end of the target, the first one end and the first other end being opposed to each other in a horizontal direction, from the image generated in the imaging step; a coordinate detection step of setting, as a first one end point and a first other end point, points at the first one end point and the first other end point that are at the same distance from a second one end point or a second other end point in a vertical direction, respectively, at the first one end point and the first other end point that are extracted in the end extraction step, setting a target representative point that represents a position of the target between the set first one end point and the first other end point, and determining coordinates of the target representative point in a predetermined coordinate system; a target region generating step of generating a predetermined frame having a center at the target representative point set in the coordinate detecting step as a region in which the target is present; the coordinate detection step includes a measurement step of measuring a distance from the aircraft to the target, The coordinate detection step includes determining a point of the target that is closest to the aircraft as a closest point using the measurement step between the first one end and the first other end, and setting a point that is point-symmetric to the closest point with respect to the target representative point as a symmetric point as a virtual point; the target region generating step generates, as the predetermined frame, a rectangle having the first one end point, the first other end point, the closest point, and the virtual point set in the coordinate detecting step as vertices; Target detection method.
Citation Information
Patent Citations
Image processor for vehicle
JP1994124340A
Travel environment recognition device
JP2004265432A
Vehicle detection method and vehicle detector
JP2005156199A
Improved object detection and motion state estimation for vehicle environment sensing systems
JP2019526781A
