Object detection device and method

Abstract

To provide an object detection and an object detection method which can detect an object more precisely.SOLUTION: An object detection device installed on a movable body for detecting an object existing in the periphery of the movable body as an object, comprises: object candidate detection parts 1, 3 which detect a candidate of the object as an object candidate; a travelable area detection part 2 which detects a travelable area of the movable body as a travelable area; and an object specification part 44 which specifies the object from the object candidate detected by the object candidate detection parts 1, 3. The object specification part 44 specifies the object from the object candidate which excludes the object candidate which overlaps with the travelable area detected by the travelable area detection part 2.SELECTED DRAWING: Figure 1

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 an object existing around the moving body as a target.

Background Art

[0002] In recent years, due to requirements for safety technology and requirements for autonomous driving technology, recognition of the surrounding situation has been desired. In particular, in the case of collision avoidance and lane change in a moving body such as a vehicle or a robot, recognition of the surrounding situation is important. For this reason, a moving body is often equipped with an object detection device (target detection device) for detecting an object in order to recognize the surrounding situation (see, for example, Patent Document 1).

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] For the detection of this target, an imaging device or a radar device is often used, and these are often used in combination. By the way, in such a case, due to distortion (aberration) of the imaging optical system (imaging optical system) in the imaging device, a target candidate may be generated at a position different from the actual position of the object, or when using a plurality of radar devices, a plurality of target candidates may be generated for one object, and there is a risk of detecting the object as a wrong target.

[0005] The present invention has been made in view of the above circumstances, and an object thereof is to provide a target detection device and a target detection method capable of detecting a target with higher accuracy.

Means for Solving the Problems

[0006] As a result of various studies, the present inventor has found that the above object can be achieved by the following present invention. That is, a target detection device according to one aspect of the present invention is a device mounted on a moving body for detecting an object existing around the moving body as a target, including a target candidate detection unit for detecting a candidate of the target as a target candidate, a travelable area detection unit for detecting an area where the moving body can travel as a travelable area, and a target identification unit for identifying the target from the target candidates detected by the target candidate detection. The target identification unit identifies the target from the target candidates excluding the target candidates overlapping with the travelable area detected by the travelable area detection unit. Preferably, in the above target detection device, the travelable area detection unit includes an imaging unit for generating an image and a travelable area extraction unit for extracting the travelable area based on the image generated by the imaging unit.

[0007] It is considered that no object exists in the travelable area and thus no target is detected. The present invention has been made by focusing on this point. Since the above target detection device identifies the target from the target candidates excluding the target candidates overlapping with the travelable area detected by the travelable area detection unit, it identifies the target from the target candidates in which the target candidates not corresponding to the object are removed as ghosts (virtual targets), so that the target can be detected with higher accuracy.

[0008] In another aspect, in the above target detection device, when a target candidate representative point representing the position of the target candidate overlaps with the travelable area, the target identification unit determines that the target candidate overlaps with the travelable area detected by the travelable area detection unit.

[0009] When the target candidate representative point representing the position of the target candidate overlaps with the travelable area, it is considered that there is a high probability that the target candidate is located in the travelable area where no object exists, and this target candidate is highly likely to be a ghost (virtual target). Since the above target detection device determines that the target candidate overlaps with the travelable area detected by the travelable area detection unit when the target candidate representative point overlaps with the travelable area, the ghost target candidate can be accurately removed, so that the target can be detected with higher accuracy.

[0010] In another aspect, in the above-described target detection device, the target candidate detection unit includes an imaging unit that generates an image, and a target candidate extraction unit that extracts the target candidate based on the image generated by the imaging unit. The target identification unit extracts a first one end and a first other end that face each other in the horizontal direction in the target candidate from the image generated by the imaging unit, and sets points that are at the same distance from the second one end or the second other end in the vertical direction at each of the extracted first one end and the first other end as a first one end point and a first other end point, and sets the target candidate representative point between the set first one end point and the first other end point. Preferably, in the above-described target detection device, the target identification unit sets the target candidate representative point at the central position between the set first one end point and the first other end point in order to make the target candidate representative point the center of gravity point of the target candidate. Preferably, in the above-described target detection device, the drivable area detection unit includes an imaging unit that generates an image, and a drivable area extraction unit that extracts the drivable area based on the image generated by the imaging unit. The drivable area extraction unit further extracts the target candidate based on the image generated by the imaging unit, and the imaging unit and the target candidate extraction unit in the target candidate detection unit, and the imaging unit and the drivable area extraction unit in the drivable area detection unit are used in common.

[0011] According to this, a target detection device that sets a target candidate representative point from an image can be provided.

[0012] In another aspect, in the above-described target detection device, the target identification unit determines whether a target candidate overlapping the drivable area detected by the drivable area detection unit based on the overlapping area between the target candidate and the drivable area. Preferably, in the above-described target detection device, when the ratio of the overlapping area between the target candidate and the drivable area to the total area of the target candidate is equal to or greater than a predetermined threshold, the target identification unit determines that the target candidate overlaps the drivable area detected by the drivable area detection unit. Preferably, in the above-described target detection device, when a target candidate representative point representing the position of the target candidate overlaps the drivable area, and the ratio of the overlapping area between the target candidate and the drivable area to the total area of the target candidate is equal to or greater than a predetermined threshold, the target identification unit determines that the target candidate overlaps the drivable area detected by the drivable area detection unit.

[0013] According to this, it is possible to provide a target detection device that determines whether a target candidate is a ghost based on the overlapping condition between the target candidate and the drivable area.

[0014] In another aspect, in the above-described target detection device, the target candidate detection unit includes an imaging unit that generates an image, a target candidate extraction unit that extracts the target candidate based on the image generated by the imaging unit, a radar unit that transmits while scanning a predetermined detection wave and detects a detection point that reflects the detection wave at its position by receiving the reflected wave of the detection wave, and a second target candidate extraction unit that extracts the target candidate based on each detection point detected by the radar unit. When the target identification unit integrates the target candidate extracted by the target candidate extraction unit and the target candidate extracted by the second target candidate extraction unit, the target identification unit identifies the target from the target candidates excluding the target candidates overlapping the drivable area detected by the drivable area detection unit. Preferably, in the above-described target detection device, when the target identification unit integrates the target candidate extracted by the target candidate extraction unit and the target candidate extracted by the second target candidate extraction unit, the target identification unit does not exclude the target candidate extracted by the target candidate extraction unit, and identifies the target from the target candidates excluding the target candidates overlapping the drivable area detected by the drivable area detection unit among the target candidates extracted by the second target candidate extraction unit.

[0015] According to this, it is possible to provide an object detection device that removes ghost object candidates and detects an object by so-called sensor fusion.

[0016] An object detection method according to another aspect of the present invention is a method used for a moving body to detect an object existing around the moving body as an object, including an object candidate detection step of detecting a candidate of the object as an object candidate, a travelable area detection step of detecting an area where the moving body can travel as a travelable area, and an object identification step of identifying the object from the object candidates detected in the object candidate detection step, wherein the object identification step identifies the object from the object candidates excluding the object candidates that overlap with the travelable area detected in the travelable area detection step.

[0017] Such an object detection method identifies the object from the object candidates excluding the object candidates that overlap with the travelable area detected in the travelable area detection step. Therefore, since the object is identified from the object candidates obtained by removing the object candidates not corresponding to the object as ghosts (false objects), the object can be detected with higher accuracy.

Advantages of the Invention

[0018] The object detection device and the object detection method according to the present invention can detect an object with higher accuracy.

Brief Description of the Drawings

[0019]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Embodiments for Carrying Out the Invention

[0020] 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 each figure, components denoted by the same reference numerals are the same components, and the description thereof will be omitted as appropriate. In this specification, when referring to components generically, reference numerals without suffixes are used, and when referring to individual components, reference numerals with suffixes are used.

[0021] The target detection device in the embodiment is mounted on a moving body and detects an object existing around the moving body (an object visible from the moving body within a predetermined range) as a target. This target detection device includes a target candidate detection unit that detects candidates for the target as target candidates, a travelable area detection unit that detects the area where the moving body can travel as a travelable area, and a target identification unit that identifies the target from the target candidates detected by the target candidate detection. Then, the target identification unit identifies the target from the target candidates excluding those that overlap with the travelable area detected by the travelable area detection unit. The moving body is a device capable of changing its own position, such as a vehicle or a robot. The vehicle is, for example, a parts transporter in a factory and an automobile. Hereinafter, the case where a target detection device is mounted on a vehicle, which is an example of a moving body, will be described more specifically.

[0022] FIG. 1 is a block diagram showing an electrical schematic configuration of a vehicle equipped with an object detection device according to an embodiment. FIG. 2 is a diagram for explaining the arrangement positions of the imaging unit and the first to third radar units in the object detection device. FIG. 3 is a diagram for explaining a drivable area detected (extracted) by the object detection device. FIG. 4 is a diagram for explaining a method of removing object candidates in the object detection device. FIG. 5 is a diagram for explaining a method of setting representative points of object candidates in the object detection device. FIG. 5A is a plan view schematically showing the situation of the host vehicle and other vehicles traveling on a one-lane road on one side. FIG. 5B is a diagram schematically showing an image generated by imaging another vehicle traveling forward from the host vehicle shown in FIG. 5A.

[0023] A vehicle CV equipped with an object detection device according to an embodiment includes, for example, as shown in FIG. 1, a first object candidate detection unit 1, a drivable area detection unit 2, a second object candidate detection unit 3, a control processing unit 4, and a storage unit 5. For traveling, the vehicle further includes a vehicle speed sensor 61 that measures the speed of the host vehicle VC (the vehicle is appropriately referred to as the "host vehicle" when distinguishing it from other vehicles such as a preceding vehicle), an acceleration sensor 62 that measures the acceleration of the host vehicle VC, a steering angle sensor 63 that measures the steering angle, an accelerator sensor 64 that measures the depression amount (stroke amount) of the accelerator pedal, a brake sensor 65 that measures the depression amount (stroke amount) of the brake pedal, an engine control unit 66 that controls a power source such as an internal combustion engine or a motor, a brake control unit 67 that controls the brake, and a steering control unit 68 that controls the steering angle of the steered wheels. Each of the sensors 61 to 65 is connected to the control processing unit 4, measures its measurement target according to the control of the control processing unit 4, and outputs the measurement result to the control processing unit 4. Each of the control units 66 to 68 is connected to the control processing unit 4 and controls its control target according to the control of the control processing unit 4. In this embodiment, as shown in FIG. 1, the first object candidate detection unit 1 and the drivable area detection unit 2 are shared. Of course, the first object candidate detection unit 1 and the drivable area detection unit 2 may be configured separately.

[0024] The first object candidate detection unit 1 is a device that detects candidates for the object (first object candidates) as objects existing around a vehicle VC, which is an example of a moving body. The first object candidate detection unit 1 includes, for example, in this embodiment, an imaging unit 11 and a first object candidate extraction unit 12(42) in order to detect the first object candidates based on an image. The imaging unit 11 is mounted on the vehicle VC, connected to the control processing unit 4, and is a device that images the surroundings of the object detection device within a predetermined imaging range according to the control of the control processing unit 4 and generates an image thereof. The imaging unit 11 outputs the generated image to the control processing unit 4. The imaging unit 11 is mounted on the vehicle VC so as to image, for example, the front (travel direction, progress direction). For example, in this embodiment, as shown in FIG. 2, the imaging unit 11 is disposed within the vehicle VC on the ceiling surface (inner roof surface) near the windshield so as to image the front of the host vehicle VC within a predetermined imaging range (angle of view) R1 including the road surface, with its imaging direction (optical axis direction) directed obliquely downward. The imaging unit 11 is, for example, a digital camera including an imaging optical system that forms an optical image of a subject (object) on a predetermined imaging surface, an area image sensor that is disposed with its light receiving surface coinciding with the imaging surface and converts the optical image of the subject into an electrical signal, and an image processing unit that generates image data representing the image of the subject by performing image processing on the output of the area image sensor. The digital camera may be a visible camera or an infrared camera. The first object candidate extraction unit 12(42) extracts the first object candidates based on the image generated by the imaging unit 11 and is functionally configured in the control processing unit 4 as will be described later. The first object candidate extraction unit 12(42) will be further described in the following description of the next travelable region detection unit 2.

[0025] The travelable area detection unit 2 is a device that detects the travelable area of the vehicle VC, which is an example of the moving body, as the travelable area. In the present embodiment, for example, the travelable area detection unit 2 includes an imaging unit 11 and a travelable area extraction unit 22(42) in order to detect the travelable area based on an image. As described above, the travelable area detection unit 2 is also used as the first target candidate detection unit 1. Therefore, the imaging unit 11 of the travelable area detection unit 2 is also the imaging unit 11 of the first target candidate detection unit 1, and the travelable area extraction unit 22(42) of the travelable area detection unit 2 is also the first target candidate extraction unit 12(42) of the first target candidate detection unit 1. For this reason, the first target candidate extraction unit (travelable area extraction unit 22(42)) 12(42) extracts the first target candidate and the travelable area based on the image generated by the imaging unit 11. Such a first target candidate extraction unit (travelable area extraction unit 22(42)) 12(42) includes, for example, a machine learning model (segmentation model) that machine-learns to divide an image into image regions for each type of subject (object), and from each image region divided by the machine learning model, an image region whose type is a vehicle is extracted as the first target candidate, an image region (road surface region) whose type is a road surface is selected, a boundary line (vehicle-road surface boundary line) between the selected image regions of the first target candidates and the road surface region is obtained, and an extraction processing unit that extracts the road surface region on the front side (imaging unit 11 side) of the obtained vehicle-road surface boundary line as the travelable area. In the above case, at least a vehicle and a road surface are included in the types of the objects. In one example, when the image shown in FIG. 3 is input to the machine learning model, each image region POb1 to POb3 of the first to third preceding vehicles is extracted as three first target candidates, the road surface region PRD is selected, the vehicle-road surface boundary line DL is obtained from these three image regions POb1 to POb3 of the first target candidates and the road surface region, and the road surface region on the front side of the obtained vehicle-road surface boundary line DL is extracted as the travelable area FS where there is no obstacle on the road surface that hinders the travel of the host vehicle VC.Here, when there is no boundary line between the image region POb of the first object candidate and the road surface region PRD, such as between the image region POb2 of the first object candidate and the image region POb3 of the first object candidate, for example, the boundary line between the image region POb2 of the first object candidate and the road surface region PRD may be extended to the boundary line between the image region POb3 of the first object candidate and the road surface region PRD to obtain the vehicle-road surface boundary line DL. Alternatively, for example, the boundary line between the road surface and the background may be set as the vehicle-road surface boundary line DL between the image region POb2 of the first object candidate and the image region POb3 of the first object candidate. Alternatively, for example, the vehicle-road surface boundary line DL may be formed by sequentially connecting the lower end lines of the three image regions POb1 to POb3 of the first object candidates without using (without selecting) the road surface region. Further, the left and right boundary lines of the road surface region may be detected by detecting the image region of the boundary object that separates the road surface such as the curb and the outside thereof. In this case, the front-side region surrounded by one boundary line LN1 of the left and right boundary lines of the road surface region, the vehicle-road surface boundary line DL, and the other boundary line LN2 of the left and right boundary lines is defined as the drivable region FS.

[0026] Note that in the above description, a machine learning model was used. However, the image region of the vehicle may be obtained as the first object candidate by extracting the contour line of the vehicle from the image, for example, by using an edge filter that detects edges or by performing pattern matching of the vehicle. Then, the left and right lane regions may be extracted by extracting the white line from the image, the lower end line of the image region of the vehicle may be obtained as the vehicle-road surface boundary line, and the front-side region surrounded by the left lane region, the vehicle-road surface boundary line, and the right lane region may be defined as the drivable region.

[0027] The second target candidate detection unit 3 is a device that detects candidates for the target as target candidates (second target candidates) in order to detect an object existing around the vehicle VC, which is an example of the moving body, in the same manner as the first target candidate detection unit 1. However, the second target candidates are detected by a detection method different from that of the first target candidate detection unit 1. The first target candidate detection unit 1 detected the first target candidates based on an image, but the second target candidate detection unit 3 detects the second target candidates, for example, by receiving a reflected wave caused by a predetermined detection wave. In an image, it is difficult to distinguish between an object that is far and large from the host vehicle VC and an object that is close and small to the host vehicle VC. However, by using such different detection methods from each other, an object can be detected more reliably. The second target candidate detection unit 3 is connected to, for example, the control processing unit 4, and in accordance with the control of the control processing unit 4, transmits while scanning a predetermined detection wave, and receives the reflected wave of the detection wave to detect the detection point that reflected the detection wave at its position (distance and direction). It includes a radar unit 31 and a second target candidate extraction unit 32 (43) that extracts second target candidates based on each detection point detected by the radar unit 31.

[0028] The radar unit 31 may be, for example, a single device arranged to scan forward at the central position in the vehicle width direction. However, in this embodiment, in order to more reliably detect an object ahead, for example, as shown in FIG. 2, it includes three first to third radar units 31-1 to 31-3. The first radar unit 31-1 is provided on one side in the vehicle width direction (for example, the right front corner) at the front part of the vehicle VC, and detects an object around the vehicle VC within a predetermined measurement range R2 centered on the traveling direction of the vehicle VC. The second radar unit 31-2 is provided at a substantially central position in the vehicle width direction at the front part of the vehicle VC, and detects an object around the vehicle VC within a predetermined measurement range R2 centered on the traveling direction of the vehicle VC. The third radar unit 31-3 is provided on the other side (for example, the left front corner) at the front part of the vehicle VC, and detects an object around the vehicle VC within a predetermined measurement range R2 centered on the traveling direction of the vehicle VC.

[0029] These first to third radar units 31-1 to 31-3 each include, for example, a transmission unit that transmits a transmission wave in the millimeter wave band while scanning a measurement range R2, a reception unit that receives a reflected wave obtained by reflecting the transmission wave from 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. It is a so-called radar device. The signal processing unit determines the direction of the object from the transmission direction of the transmission wave in the scanning of the measurement range R2 for each direction (each point in the measurement range R2) by scanning, and based on the time difference between the transmission timing of the transmission wave and the reception timing of the reflected wave, determines the distance to the object (TOF (Time-Of-Flight) method). Note that the first to third radar units 31-1 to 31-3 are not limited to such a configuration and may be radar devices of an appropriate type. For example, each of the first to third radar units 31-1 to 31-3 may be a lidar device that uses laser light instead of the millimeter wave band. Also, each of the first to third radar units 31-1 to 31-3 may include a plurality of reception antennas and may be a device that determines the direction of the target object from the phase difference of the reflected waves received by the plurality of reception antennas.

[0030] Since the object has a predetermined size, one or more reflected waves can be obtained from each position (each location) corresponding to the spatial resolution of the radar unit 31 (31-1 to 31-3) in one object. Therefore, each of these first to third radar units 31-1 to 31-3 detects one or more detection points from one object. Each of these first to third radar units 31-1 to 31-3 outputs the one or more detection points (their relative positions (relative directions, relative distances)) thus detected to the control processing unit 4.

[0031] As will be described later, the second target candidate extraction unit 32(43) is functionally configured in the control processing unit 4, and obtains second target candidates based on one or more detection points detected by the first to third radar units 31-1 to 31-3. More specifically, the second target candidate extraction unit 32(43) obtains clusters (groups) of detection points as second target candidates, for example, by clustering one or more detection points detected by the first to third radar units 31-1 to 31-3.

[0032] The memory unit 5 is a circuit connected to the control processing unit 4 and stores various predetermined programs and various predetermined data according to the control of the control processing unit 4. The various predetermined programs include, for example, a control processing program. The control processing program includes a control program for controlling each of the parts 1 to 3, 5, 61 to 68 of the vehicle VC equipped with the target detection device according to the functions of the respective parts, a first target candidate extraction program for extracting a first target candidate based on the image generated by the imaging unit 11, a drivable area extraction program for extracting a drivable area based on the image generated by the imaging unit 11, a second target candidate extraction program for extracting a second target candidate based on one or more detection points detected by the radar unit 31, and a target identification program for identifying a target from the first and second target candidates extracted by the first and second target candidate extraction programs, etc. In the present embodiment, the first target candidate extraction program and the drivable area extraction program are combined into one program, and include the machine learning model (segmentation model) and an extraction processing program for extracting the first target candidate and the drivable area based on each image area divided by the machine learning model. The various predetermined data include, for example, data necessary for executing these programs, such as the extracted first and second target candidates and the drivable area. Such a memory unit 5 includes, for example, a ROM (Read Only Memory) which is a non-volatile memory element, an EEPROM (Electrically Erasable Programmable Read Only Memory) which is a rewritable non-volatile memory element, etc. And the memory unit 5 includes a RAM (Random Access Memory) etc. which serves as a working memory of the so-called control processing unit 4 for storing data etc. generated during the execution of the predetermined program.

[0033] The control processing unit 4 is a circuit that controls each of the parts 1 to 3, 5, 61 to 68 of the vehicle VC equipped with the target detection device according to the functions of the respective parts, and detects an object existing around the vehicle VC as a target. The control processing unit 4 is configured to include, for example, a CPU (Central Processing Unit) and its peripheral circuits. When the control processing program is executed, the control processing unit 4 functionally includes a control unit 41, a first target candidate extraction unit (travelable area extraction unit 22) 42 (12), a second target candidate extraction unit 43 (32), and a target identification unit 44.

[0034] The control unit 41 controls each of the parts 1 to 3, 5, 61 to 68 of the vehicle VC equipped with the target detection device according to the functions of the respective parts, and is in charge of the overall control of the vehicle VC.

[0035] As described above, the first target candidate extraction unit (travelable area extraction unit 22) 42 (12) extracts a first target candidate and a travelable area based on the image generated by the imaging unit 11.

[0036] As described above, the second target candidate extraction unit 43 obtains a second target candidate based on one or a plurality of detection points detected by the first to third radar units 31-1 to 31-3.

[0037] The target identification unit 44 identifies a target from the first and second target candidates extracted by the first and second target candidate extraction units 42(12) and 43(32). Here, since it is considered that no object exists in the drivable area and thus no target is detected normally, when identifying the target, the target identification unit 44 uses the drivable area detected by the drivable area detection unit 2. In this embodiment, the target identification unit 44 identifies the target from the first and second target candidates excluding the target candidates that overlap with the drivable area extracted by the first target candidate extraction unit (drivable area extraction unit 22) 42(12). It is also possible to remove the target candidates that overlap with the drivable area from the first target candidates, but in this embodiment, since the accuracy of the machine learning model is relatively high and for the simplification of information processing, the target candidates that overlap with the drivable area are removed only from the second target candidates. That is, when integrating (sensor fusion) the first target candidate extracted by the first target candidate extraction unit 42(12) and the second target candidate extracted by the second target candidate extraction unit 43(32), the target identification unit 44 does not exclude the first target candidate extracted by the first target candidate extraction unit 42(12), but from the second target candidates extracted by the second target candidate extraction unit 43(32), excluding the target candidates that overlap with the drivable area detected by the drivable area detection unit 2, the target is identified from the remaining target candidates. More specifically, when the target candidate representative point representing the position of the target candidate overlaps with the drivable area, it is considered that the probability that the target candidate is located in the drivable area where no object exists is high, and this target candidate is highly likely to be a ghost (false target). Therefore, when the target candidate representative point representing the position of the target candidate, in this embodiment, the second target candidate representative point representing the position of the second target candidate overlaps with the drivable area, the target identification unit 44 determines that it is a target candidate that overlaps with the drivable area detected by the drivable area detection unit 2. By identifying the target in this way, the target detection device in this embodiment is preferably used for detecting targets on highways or motorways where it is difficult to assume a sudden emergence from the side.

[0038] For example, as shown in FIG. 4, the target identification unit 44 forms a two-dimensional virtual plane of an XY orthogonal coordinate system with the center position of the own vehicle VC as the coordinate origin, the traveling direction as the X-axis, and the vehicle width direction as the Y-axis. On this two-dimensional virtual plane, the drivable area FS detected by the drivable area detection unit 2 is developed, and the second target candidates detected by the second target candidate detection unit 3 are arranged. In the example shown in FIG. 4, three second target candidates GI1, GI2, and RI are arranged. In the example shown in FIG. 4, the second target candidate representative point PMri of the second target candidate RI does not overlap with the drivable area FS, and the second target candidate representative points PMgi1 and PMgi2 of the second target candidates GI1 and GI2 overlap with the drivable area FS. Therefore, the target identification unit 44 removes the second target candidates GI1 and GI2 and retains the second target candidate RI.

[0039] Such a second target candidate representative point PM may be obtained, for example, by representing the second target candidate with a rectangle that encloses the cluster of detection points clustered by the second target candidate extraction unit 32(43) with the minimum size, and setting the center of gravity point (the intersection of two diagonals) of the rectangle as the second target candidate representative point. However, in the present embodiment, the second target candidate representative point is obtained as follows. Note that, when the second target candidate is a vehicle, the size of the vehicle can be assumed in advance for the rectangle that encloses it with the minimum size. Therefore, after obtaining the second target candidate representative point as described above, it may be changed (resized) to a rectangle of a predetermined size centered on this second target candidate representative point.

[0040] The target identification unit 44 extracts the first one end and the first other end that face each other in the horizontal direction in the target candidate, in this embodiment, the second target candidate, from the image generated by the imaging unit 11, and sets each point at the same distance from the second one end or the second other end in the vertical direction at each of the extracted first one end and the first other end as the first one end point and the first other end point. Between the set first one end point and the first other end point, the target candidate representative point, in this embodiment, the second target candidate representative point is set. Preferably, the target identification unit 44 sets the target candidate representative point at the central position between the set first one end point and the first other end point in order to make the target candidate representative point the center of gravity point of the target candidate. The distance from the second one end or the second other end is appropriately set in advance.

[0041] More specifically, the object identification unit 44 first generates a silhouette image obtained by filling the image area of the vehicle with a single color from the image generated by the imaging unit 11, and selects the silhouette image closest to the position of the second object candidate extracted by the second object candidate extraction unit 43 (32) as the silhouette image of the second object candidate. For such image processing, for example, known techniques for object detection used in collision detection can be utilized. Alternatively, the object identification unit 44 may be configured to extract edges by applying an edge filter to the image generated by the imaging unit 11, extract straight lines by applying a Hough transform of a straight line to the extracted edges, and generate a silhouette image obtained by filling a substantially rectangular shape formed by the extracted straight lines with a single color as the image area of the vehicle.

[0042] For example, as shown in FIG. 5A, the host vehicle VC, the first other vehicle Oba, and the second other vehicle Obb are traveling on a one-lane road on one side. The first other vehicle Oba is traveling in front of the host vehicle VC in the same lane as the lane in which the host vehicle VC is traveling. When the second other vehicle Obb is traveling diagonally in front of the host vehicle VC in the oncoming lane of the host vehicle VC, in the silhouette image generated by the object identification unit 44 from the image generated by the imaging unit 11 of the host vehicle VC, an image area POba of the first other vehicle Oba shown in FIG. 5B is formed. In FIG. 5B, for the sake of explanation, the image area POba of the first other vehicle Oba is not filled with a single color. In such a case, the object identification unit 44 first extracts both ends (the left end and the right end when viewing FIG. 5B in a plan view) EL1 and ER1 that face each other in the horizontal direction in the image area POba of the first other vehicle Oba as the first one end (left end) EL1 and the first other end (right end) ER1, respectively. Next, the object identification unit 44 sets points PL1 and PR2 at the same distance from the second one end (upper end) or the second other end (lower end) in the vertical direction at each of the extracted one end (left end) EL1 and the other end (right end) ER1 as the first one end point (left end point) PL1 and the first other end point (right end point) PR1, respectively. Then, the object identification unit 44 sets a second object candidate representative point PM1 at the central position between the set first one end point PL1 and the first other end point PR1 in this embodiment. Further, in this embodiment, the object identification unit 44 determines, between the first one end EL1 and the first other end EL2, the point of the first other vehicle Oba that is at the closest distance from the host vehicle VC (the point that is closest to the host vehicle VC in terms of distance among the detected points clustered as the second object candidates) as the closest point PC1, sets a point symmetric to the closest point PC1 with respect to the line segment connecting the first one end point PL1 and the first other end point PL2 as the virtual point PV1, and generates a rectangle having each of the first one end point PL1, the first other end point PL2, the closest point PC1, and the virtual point PV1 as vertices as a frame representing the second object candidate of the first other vehicle Oba.

[0043] Then, the target identification unit 44 identifies the target by integrating the first and second target candidates excluding the target candidates overlapping with the drivable area extracted by the first target candidate extraction unit (drivable area extraction unit 22) 42 (12). In the present embodiment, the target identification unit 44 identifies the target by integrating the first target candidate extracted by the first target candidate extraction unit 42 (12) and the second target candidate excluding the target candidates overlapping with the drivable area extracted by the first target candidate extraction unit (drivable area extraction unit 22) 42 (12). For this integration, known conventional means are used. For example, for each of the first and second target candidates, the distance between the representative points of the first and second target candidates is obtained, and the first and second target candidates that are closest to each other in terms of distance are grouped into one target, whereby the first and second target candidates are integrated.

[0044] Such a control processing unit 4 and a storage unit 5 can be configured by a computer called a so-called ECU (Electronic Control Unit).

[0045] In the above description, the target detection device and the target detection method implemented therein are configured by the first and second target candidate detection units 1 and 3, the drivable area detection unit 2, the control processing unit 4, and the storage unit 5.

[0046] Next, the operation of the present embodiment will be described. FIG. 6 is a flowchart showing the operation of the target detection device.

[0047] When the vehicle starts operating, the target detection device mounted on such a vehicle executes initialization of necessary parts and starts its operation. By executing the control processing program, the control processing unit 4 is functionally configured with a control unit 41, a first target candidate extraction unit (drivable area extraction unit 22) 42 (12), a second target candidate extraction unit 43 (32), and a target identification unit 44. Then, for example, until the ignition of the vehicle is turned off, each of the processes S1 to S13 shown in FIG. 6 below is repeatedly executed at a predetermined sampling interval set in advance as appropriate, and a target is detected.

[0048] In FIG. 6, the object detection device causes the imaging unit 11 to generate an image, causes the first to third radar units 31-1 to 31-3 to detect detection points, and stores these in the storage unit 5 by the control processing unit 4 (S1).

[0049] Next, the object detection device extracts a first object candidate and extracts a drivable area from the image generated in process S1 by the first object candidate extraction unit (drivable area extraction unit 22) 42 (12) of the control processing unit 4, and stores these in the storage unit 5 (S2).

[0050] Next, the object detection device extracts a second object candidate from the detection points detected in process S1 by the second object candidate extraction unit 43 (32) of the control processing unit 4, and stores this in the storage unit 5 (S3).

[0051] Next, the object detection device obtains a second object candidate representative point of the second object candidate extracted in process S3 by the object identification unit 44 of the control processing unit 4, and stores this in the storage unit 5 (S4).

[0052] Next, the object detection device generates a two-dimensional virtual space by the object identification unit 44 of the control processing unit 4 (S5).

[0053] Next, the object detection device arranges the drivable area extracted in process S2 and arranges the second object candidate extracted in process S3 in the two-dimensional virtual space generated in process S5 by the object identification unit 44 of the control processing unit 4 (S6).

[0054] Next, the object detection device determines whether or not the second object candidate extracted in process S3 overlaps the drivable area extracted in process S2 by the object identification unit 44 of the control processing unit 4 (S7). As a result of this determination, if there is an overlap (Yes), the object identification unit 44 then executes process S8. On the other hand, as a result of this determination, if there is no overlap (No), the object identification unit 44 then executes process S10.

[0055] In this process S8, the target detection device determines, by the target identification unit 44 of the control processing unit 4, whether the second target candidate representative point of the second target candidate is within the drivable area. As a result of this determination, if it is within the drivable area (Yes), the target identification unit 44 then executes process S9. On the other hand, as a result of the determination, if it is not within the drivable area (No), the target identification unit 44 then executes process S10.

[0056] In this process S9, the target detection device removes (erases from the storage unit 5) the second target candidate determined in process S8 by the target identification unit 44 of the control processing unit 4, and then executes process S11.

[0057] In the said process S10, the target detection device leaves the second target candidate determined in process S7 or process S8 by the target identification unit 44 of the control processing unit 4, and then executes process S11.

[0058] In this process S11, the target detection device determines, by the target identification unit 44 of the control processing unit 4, whether the above determination process has been executed for all the second target candidates. As a result of this determination, if it has been determined for all the second target candidates (Yes), the target identification unit 44 then executes process S12. On the other hand, as a result of the determination, if it has not been determined for all the second target candidates (No), the target identification unit 44 returns the process to process S7.

[0059] In this process S12, the target detection device integrates (sensor fusion) the first target candidate and the second target candidate remaining by the above process by the target identification unit 44 of the control processing unit 4.

[0060] Next, the target detection device outputs the target specified by the target specifying unit 44 of the control processing unit 4 through the integration of process S12, and ends this current process (S13). For example, when the target detected by the target detection device is used for collision avoidance, the target specifying unit 44 outputs the target specified by the integration of process S12 to a collision avoidance unit (not shown) that is functionally configured in the control processing unit 4 by executing a collision avoidance program that controls the engine control unit 66, the brake control unit 67, and the steering control unit 68 so as to avoid the host vehicle VC from the object represented by the target.

[0061] As described above, the target detection device mounted on the vehicle VC in the embodiment and the target detection method implemented thereon specify a target from target candidates excluding target candidates overlapping with the drivable area detected by the drivable area detection unit 2. Therefore, since the target is specified from the target candidates in which target candidates not corresponding to an object are removed as ghosts (virtual targets), the target can be detected with higher accuracy.

[0062] For example, when a plurality of radar units 31 (31-1 to 31-3) are provided as in the present embodiment, since target candidates are detected by each of the plurality of radar units 31 for one object, there is a possibility that the target candidates detected by each of the plurality of radar units 31 may not be integrated into one. However, since the target detection device and the target detection method in the present embodiment remove target candidates not corresponding to an object as ghosts (virtual targets), they can be appropriately integrated.

[0063] Since the target detection device and the target detection method determine that a target candidate is a target candidate overlapping with the drivable area detected by the drivable area detection unit 2 when the target candidate representative point overlaps with the drivable area, the ghost target candidates can be accurately removed, so that the target can be detected with higher accuracy.

[0064] According to the present embodiment, a target detection device and a target detection method for setting target candidate representative points from an image can be provided, and a target detection device and a target detection method for removing ghost target candidates and detecting a target by so-called sensor fusion can be provided.

[0065] In the above-described embodiment, the ghost target candidates are removed from the second target candidates detected by the second target candidate detection unit 3 including the first to third radar units 31-1 to 31-3. However, the ghost target candidates may be removed from the first target candidates detected by the first target candidate detection unit 1 including the imaging unit 11. For example, as a result of the image being distorted due to so-called aberrations of a lens (imaging optical system, imaging optical system), a deviation occurs between the position of an actual object and the position of a target based on the image. When integrating (sensor fusion), there is a possibility that the target of one object may not be integrated into one. Alternatively, when a plurality of imaging units are used, as in the case where a plurality of radar units are used, there is a possibility that the target candidates detected by each of the plurality of imaging units for one object may not be integrated into one. However, in the above cases, since the target candidates not corresponding to the object are removed as ghosts (false targets), the integration can be appropriately performed. Alternatively, the ghost target candidates may be removed from the first target candidates detected by the first target candidate detection unit 1, and the ghost target candidates may be removed from the second target candidates detected by the second target candidate detection unit 3.

[0066] Also, in the above-described embodiment, when the target candidate representative point overlaps with the drivable area, the target identification unit 44 determines the target candidate that overlaps with the drivable area detected by the drivable area detection unit 2. However, the target identification unit 44 may determine whether or not the target candidate overlaps with the drivable area detected by the drivable area detection unit 2 based on the overlapping area between the target candidate and the drivable area. For example, when the ratio of the overlapping area between the target candidate and the drivable area to the total area of the target candidate is equal to or greater than a predetermined threshold, the target identification unit 44 determines the target candidate that overlaps with the drivable area detected by the drivable area detection unit 2. The predetermined threshold is appropriately set in advance from a plurality of samples, and is set to any value within the range of, for example, 40[%] to 60[%]. According to this, it is possible to provide a target detection device that determines whether a target candidate is a ghost based on the degree of overlap between the target candidate and the drivable area. Alternatively, when the target candidate representative point representing the position of the target candidate overlaps with the drivable area, and the ratio of the overlapping area between the target candidate and the drivable area to the total area of the target candidate is equal to or greater than a predetermined threshold, the target identification unit 44 may determine the target candidate that overlaps with the drivable area detected by the drivable area detection unit.

[0067] In order to describe the present invention, the embodiments have been appropriately and fully described above with reference to the drawings. However, it should be recognized that those skilled in the art can easily make changes and / or improvements to the above-described embodiments. Therefore, as long as the changes or improvements made by those skilled in the art do not depart from the scope of the claims described in the claims, such changes or improvements are construed as being included in the scope of the claims of the claims.

Description of Reference Numerals

[0068] VC Vehicle 1 First Target Candidate Detection Unit 2 Drivable Area Detection Unit 3 Second Target Candidate Detection Unit 4 Control Processing Unit 5 Storage Unit 11(21) Imaging Unit 31 (31-1 to 31-3) Radar unit (First to third radar units) 42 (12, 22) First target candidate extraction unit (Drivable area extraction unit) 43 Second target candidate extraction unit 44 Target identification unit

Claims

1. A target detection device mounted on a moving body and detecting an object existing around the moving body as a target, comprising: a target candidate detection unit that detects a candidate of the target as a target candidate; a travelable area detection unit that detects an area where the moving body can travel as a travelable area; a target identification unit that identifies the target from the target candidates detected by the target candidate detection unit; wherein the target identification unit identifies the target from the target candidates excluding the target candidates that overlap with the travelable area detected by the travelable area detection unit; a target detection device.

2. When a target candidate representative point representing the position of the target candidate overlaps with the travelable area, the target identification unit determines that the target candidate overlaps with the travelable area detected by the travelable area detection unit. The target detection device according to Claim 1.

3. The target candidate detection unit includes an imaging unit that generates an image, and a target candidate extraction unit that extracts the target candidate based on the image generated by the imaging unit. The target identification unit extracts a first one end and a first other end that face each other in the horizontal direction in the target candidate from the image generated by the imaging unit, and sets points at the same distance from a second one end or a second other end in the vertical direction at each of the extracted first one end and the first other end as a first one end point and a first other end point, and sets the target candidate representative point between the set first one end point and the first other end point. The target detection device according to Claim 2.

4. The target identification unit determines whether the target candidate overlaps with the travelable area detected by the travelable area detection unit based on an overlapping area between the target candidate and the travelable area. The target detection device according to Claim 1.

5. The target candidate detection unit includes an imaging unit that generates an image, a target candidate extraction unit that extracts the target candidate based on the image generated by the imaging unit, a radar unit that transmits while scanning a predetermined detection wave and detects a detection point that reflects the detection wave at its position by receiving the reflected wave of the detection wave, and a second target candidate extraction unit that extracts the target candidate based on each detection point detected by the radar unit. When integrating the target candidate extracted by the target candidate extraction unit and the target candidate extracted by the second target candidate extraction unit, the target identification unit identifies the target from the target candidates excluding the target candidates that overlap with the travelable area detected by the travelable area detection unit. The object detection device according to any one of claims 1 to 4.

6. An object detection method used for a moving body to detect an object existing around the moving body as an object target, comprising: An object candidate detection step of detecting a candidate of the object target as an object candidate; A travelable area detection step of detecting an area where the moving body can travel as a travelable area; An object identification step of identifying the object target from the object candidates detected in the object candidate detection step; The object identification step identifies the object target from the object candidates excluding the object candidates overlapping the travelable area detected in the travelable area detection step. Object detection method.

Images