JP5470886B2 - Object detection device - Google Patents

Description

  The present invention relates to an object detection device that detects an object such as a vehicle.

  Conventionally, as an apparatus for detecting an object such as a vehicle, for example, as described in Japanese Patent Application Laid-Open No. 2000-206241, a plurality of reflection points detected by a radar are grouped to extract a line segment constituting the contour of the object. Things are known. This apparatus is intended to recognize a point where two extracted line segments are substantially orthogonal as an end point of an object.

JP 2000-206241 A

  However, in the above-described apparatus, if two line segments cannot be separated and extracted, different sides of the object are recognized as one side and an appropriate object cannot be detected. .

  Accordingly, the present invention has been made to solve such problems, and provides an object detection device that can improve the object detection accuracy by detecting the contour of an object adapted to the direction of the object. With the goal.

That is, the object detection device according to the present invention is an object detection device that detects an object, and detects a direction of the object based on a radar detection unit that detects a detection point of the object by a radar and a captured image of the object. A plurality of templates having different shapes, a template having an L shape is selected when the direction of the object is diagonal, and a linear shape is selected when the direction of the object is not diagonal. Contour detection means for detecting a contour of the object by applying a template to a point sequence of detection points detected by the radar detection means.

  According to the present invention, by applying a template selected according to the direction of an object to a point sequence of radar detection points to detect the contour of the object, the contour of the object is used using a template having a shape adapted to the direction of the object. Can be detected. For this reason, it is possible to appropriately detect the contour of the object and increase the detection accuracy of the object.

Further , by selecting a template having a shape including an L shape when the direction of the object is oblique and selecting a linear template when the direction of the object is not oblique, the object corresponding to the direction of the object is used. Can be detected. For this reason, it becomes possible to detect the outline of an object appropriately, and to improve the detection accuracy of an object.

  The object detection apparatus according to the present invention preferably further comprises object detection means for detecting the center position and orientation of the object based on the position of the contour of the object detected by the contour detection means.

  In the object detection device according to the present invention, it is preferable that the object is another vehicle and is mounted on the own vehicle to detect the other vehicle.

  According to the present invention, by applying a template selected according to the orientation of an object to a point sequence of radar detection points to detect the contour of the object, the contour of the object is used using a template having a shape adapted to the orientation of the object. Can be detected. For this reason, it is possible to appropriately detect the contour of the object and increase the detection accuracy of the object.

1 is a schematic configuration diagram of an object detection device according to an embodiment of the present invention. It is a flowchart which shows the object detection process in the object detection apparatus of FIG. It is explanatory drawing of the object detection process of FIG. It is explanatory drawing of the object detection process of FIG. It is explanatory drawing of the object detection process of FIG. It is explanatory drawing of the object detection process of FIG.

  Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the description of the drawings, the same elements are denoted by the same reference numerals, and redundant description is omitted. In this embodiment, the object detection apparatus and the object detection method according to the present invention are applied to a collision prevention apparatus.

  FIG. 1 is a diagram illustrating a block configuration of a collision prevention apparatus 10 according to the embodiment.

  As shown in FIG. 1, the collision prevention device 10 of the present embodiment is a device that prevents a collision with another vehicle that is mounted on a vehicle and travels forward, and includes an object detection device 1 and a braking unit 5. Configured.

  The object detection device 1 specifies an end portion of another vehicle that travels in front of the host vehicle (including an oblique front) and detects the center position and orientation of the other vehicle. For this reason, the object detection apparatus 1 includes a LIDAR (Light Detection And Ranging) 2, a camera 3, and an ECU (Electronic Control Unit) 4.

  The LIDAR 2 functions as a radar detection unit that detects a detection point of another vehicle using a radar. For example, a laser radar is used. This LIDAR 2 repeatedly detects position information of other vehicles traveling in front of the host vehicle at a constant cycle. The LIDAR 2 is attached to the front part of the vehicle, and repeatedly irradiates the laser beam forward at a constant period and detects the reflected light, thereby detecting the vehicle side surface (front surface, side surface, rear surface) of the other vehicle. Detect location information.

  The LIDAR 2 emits laser light while changing the direction in the vertical and horizontal directions at every predetermined angle (for example, 0.8 degrees). For this reason, in LIDAR2, it is possible to detect one or a plurality of vehicles traveling within a predetermined range with respect to the own vehicle, and a plurality of detection points are obtained for each other vehicle. In the present embodiment, the point sequence of detection points detected by LIDAR 2 is referred to as a LIDAR point sequence. And LIDAR2 transmits the detected LIDAR point sequence to ECU4 at any time.

  The camera 3 is an image capturing unit that captures an image of the front of the host vehicle and an image of another vehicle traveling in front of the host vehicle, and is attached in the vicinity of LIDAR 2 at the front of the vehicle. The camera 3 transmits the captured image to the ECU 4 as needed.

  The ECU 4 is a control unit that controls the entire apparatus, and is configured mainly by a computer including a CPU, a ROM, and a RAM, for example. The ECU 4 is connected to the LIDAR 2 and the camera 3, acquires a LIDAR point sequence transmitted from the LIDAR 2, and acquires a captured image transmitted from the camera 3.

  The ECU 4 functions as an object direction detection unit that detects an object to be detected, that is, a direction of another vehicle. For example, a captured image of the camera 3 is input, and the orientation of another vehicle is detected using the vehicle orientation detection template for the captured image.

  In addition, the ECU 4 functions as a contour detection unit that detects a contour of an object by applying a contour detection template to a point sequence of detection points detected by the LIDAR 2. For example, the ECU 4 has a plurality of templates having different shapes such as an L-shaped template or a linear one as a contour detection template, and selects a contour detection template according to the direction of another vehicle. . At this time, the ECU 4 functions as a template selection unit. Then, the contour of the other vehicle is detected by applying the selected contour detection template to the point sequence of the detection points.

  Note that the outline detection template including the L-shape may be composed of, for example, two orthogonal straight lines or a rectangular shape. At this time, the L shape includes a substantially L shape orthogonal to approximately 90 degrees.

  When selecting the contour detection template, the ECU 4 selects a template including an L shape when the direction of the other vehicle is oblique, and when the direction of the other vehicle is not oblique (front direction, rear direction or Select a linear template for landscape orientation.

  The braking unit 5 controls a brake (not shown) to perform a braking control that gives a braking force to the host vehicle. That is, the braking unit 5 determines whether or not the host vehicle collides with the other vehicle based on the center position and orientation of the other vehicle. And if the braking part 5 judges that it collides, it will give the braking force to a brake, and will decelerate the own vehicle.

  Next, operations of the collision prevention device 10 and the object detection device 1 will be described.

  2 is a flowchart illustrating the operation of the ECU, FIG. 3 is a diagram illustrating detection points detected by the LIDAR, and FIG. 4 is a diagram illustrating a captured image captured by the camera. The following control is repeatedly executed by the ECU 4, for example.

  First, in step S1, an LIDAR point sequence acquisition process is performed. That is, in LIDAR2, one or a plurality of other vehicles are detected by a plurality of detection points which are LIDAR point sequences, and the LIDAR point sequence is acquired by transmitting the LIDAR point sequence from LIDAR2. FIG. 3 shows an example in which another vehicle 21 is traveling ahead of the host vehicle 20. In this case, the other vehicle 21 is detected by the LIDAR point sequence P configured by a plurality of detection points p. Note that the position information of each detection point detected by the LIDAR 2 is represented by a coordinate axis (coordinate system of the camera 3) with the lens center of the camera 3 as the origin.

  And it transfers to step S2 and the grouping process of a LIDAR point sequence is performed. In this grouping process, the LIDAR point sequence acquired in step S1 is grouped corresponding to each other vehicle. In the following description, the process of each step is performed for each LIDAR point sequence grouped in step S2.

  Next, it transfers to step S3 and a point sequence conversion process is performed. The point sequence conversion processing is processing for converting a LIDAR point sequence into a point sequence on a horizontal plane. For example, the coordinate value of the LIDAR point sequence acquired as three-dimensional information is converted into a coordinate value of a two-dimensional xy coordinate system as seen from directly above. By this point sequence conversion process, the coordinate value of the point sequence corresponding to the contour of the other vehicle can be acquired.

  Next, it transfers to step S4 and the acquisition process of the captured image imaged with the camera 3 is performed. When one or more other vehicles are traveling in front of the host vehicle, a captured image including an image of one or more other vehicles is acquired.

  And it transfers to step S5 and a point sequence coordinate conversion process is performed. The point sequence coordinate conversion process is a process for converting the coordinate value of the LIDAR point sequence into a coordinate value on a planar image viewed from the camera 3. For example, the coordinate value of the LIDAR point sequence is converted into a two-dimensional coordinate value in the uv coordinate system. The point sequence information on the captured image can be acquired by this point sequence coordinate conversion process.

  Next, the process proceeds to step S6, and rectangular area extraction processing is performed. The rectangular area extraction process is a process of extracting a rectangular area on the captured image so as to include the point sequence acquired in step S5. For example, a plurality of rectangular areas including the point sequence acquired in step S5 are set, and those close to the outer shape of the other vehicle are selected and extracted as rectangular areas.

  Specifically, as shown in FIG. 4, a rectangular area A is extracted so as to include the LIDAR point sequence P and conform to the outer shape of the vehicle. At this time, the rectangular area A is extracted so that the left and right end portions of the rectangular area A match the left and right end portions of the image of the other vehicle.

  In FIG. 4, the image G of the front of the other vehicle is shown. However, in actuality, the other vehicle may be an image inclined to the back, side, or diagonally. In this case, the rectangular area A is extracted so that the left and right end portions of the rectangular area A match the left and right end portions of the image of the other vehicle.

  In addition, you may perform the search of the image of other vehicles by matching with the predetermined vehicle image template prepared previously, for example. Moreover, you may perform the search of the image of another vehicle by scanning all or one part of a captured image, without using a LIDAR point sequence.

  Next, the process proceeds to step S7, and the direction detection process of the other vehicle is performed. The direction detection process of the other vehicle is a process of detecting the direction of the other vehicle based on the image of the rectangular area extracted in step S6. For example, the direction of another vehicle is detected using a template of a vehicle image recorded in advance.

  Next, the process proceeds to step S8, and coordinate conversion processing of the positions of the left and right ends of the rectangular area is performed. In the coordinate conversion processing of the left and right end positions, first, the left and right positions of the left and right ends (a1 and a2 in FIG. 4) of the rectangular area extracted in step S6 are obtained. The left and right end positions are coordinate-converted to corresponding positions on the horizontal plane on the same coordinate axis as the detection point detected by LIDAR 2 and with the lens center of the camera 3 as the origin.

  FIG. 5 is a diagram showing a positional relationship on the horizontal plane. As shown in FIG. 5, in step S <b> 8, coordinate conversion is performed on the assumption that the captured image G is arranged between the host vehicle 20 and the other vehicle 21 and at a predetermined position determined by the magnification setting of the camera 3. . In FIG. 5, the captured image G is shown in a three-dimensionally inclined manner for easy understanding.

  Next, the process proceeds to step S9, and the calculation processing of the azimuth straight lines β1 and β2 at the left and right ends of the rectangular area where the coordinate conversion is performed in step S8 is performed. As shown in FIG. 5, in the calculation processing of the azimuth straight lines β1 and β2, the lens center O of the camera 3 that is the origin of the coordinate axes and the positions a11 and a12 of the left and right ends of the rectangular area in which the coordinate conversion is performed in step S6. A straight line passing through is calculated. The calculated straight lines are set as azimuth straight lines β1 and β2. Therefore, the azimuth straight lines β1 and β2 are straight lines passing through the left and right end portions of the other vehicle 21 when viewed from the lens center O of the camera 3 that is the origin of the coordinate axes.

  Next, the process proceeds to step S10, and template selection processing is performed. The template selection process is a process for selecting a template used for detecting the contour of another vehicle. For example. As a template for contour detection, a template having a shape including an L shape and a template having a linear shape are set. A template having an L shape is selected when the direction of the other vehicle is oblique, and a linear template is selected when the direction of the other vehicle is not oblique. When the direction of the other vehicle is the front direction, the back direction, or the side direction, the linear template is selected when the direction of the other vehicle is not oblique.

  Next, it transfers to step S11 and an outline detection process is performed. The contour detection processing is processing for detecting a contour by applying a contour detection template to a point sequence of another vehicle, and calculating an intersection between the contour detection template and the azimuth line.

  For example, as shown in FIG. 5, when the other vehicle is facing the front (not obliquely), a linear template α is used as the contour detection template, and the template α is applied to the point sequence P as a template. The position of α is set. Then, intersections X1 and X2 of the template α and the azimuth lines β1 and β2 at the position are calculated.

  Further, as shown in FIG. 6, when the other vehicle is in an oblique direction, an L-shaped template αL is used as a contour detection template, and the position of the template α is determined by applying the template αL to the point sequence P. Is set. Then, the intersections X1 and X2 of the template αL and the azimuth lines β1 and β2 at the position are calculated, and the corner position Xc of the template αL is calculated. The position Xc may be calculated based on the position and tilt angle of the template αL.

  In this case, since the other vehicle 21 is oblique with respect to the host vehicle 20, the point sequence P has a shape having a corner. For this reason, the L-shaped template αL can be easily applied to the point sequence P, the fitting error is reduced, and appropriate contour detection can be performed.

  Next, it transfers to step S12 and the calculation process of the center position and direction of another vehicle is performed. In the calculation process of the center position and direction of the other vehicle, a rectangle having the intersection points x1 and x2 calculated in step S11 as end points is applied.

  In the object detection device 1, a rectangular template corresponding to the shape of the vehicle is prepared in advance. In step S12, the rectangular template is arranged along the template α or αL, and enlarged or reduced. Applied.

  At this time, when the other vehicle is not inclined, whether the other vehicle is front-facing, rearward-facing, or lateral-facing may be determined by estimating the direction based on the distance between the intersections X1 and X2 and applying a rectangular template. Then, the center position and orientation of the fitted rectangle are calculated as the center position and orientation of the other vehicle.

  When the center position and orientation of the other vehicle are calculated by the object detection device 1, braking control is performed by the braking unit 5. That is, the possibility of collision with each other vehicle is determined from the center position and orientation of the other vehicle calculated by the object detection device 1. As a result, if it is determined that there is a possibility of collision with any other vehicle, the braking force is applied to the host vehicle by controlling the brake, and contact with the other vehicle is prevented.

  As described above, according to the object detection apparatus 1 according to the present embodiment, the template selected according to the direction of the other vehicle is applied to the point sequence of the radar detection points to detect the contour of the other vehicle. The contour of another vehicle can be detected using a template having a shape adapted to the direction of the vehicle. For this reason, it is possible to appropriately detect the contour of the other vehicle and increase the detection accuracy of the other vehicle. For example, it is suppressed that the front and side surfaces of the other vehicle are recognized as straight lines and the contour of the other vehicle is erroneously detected, and the detection accuracy of the other vehicle is improved.

  Moreover, in the object detection apparatus 1 according to the present embodiment, a template having an L shape is selected when the direction of the other vehicle is oblique, and a linear template is selected when the direction of the other vehicle is not oblique. Thus, it is possible to detect the contour of the other vehicle using a template corresponding to the direction of the other vehicle. For this reason, it becomes possible to detect the outline of other vehicles appropriately, and to improve the detection accuracy of an object.

  The embodiment described above is an example of the object detection device according to the present invention, and the object detection device according to the present invention is not limited to the one described in the present embodiment. The object detection apparatus according to the present invention may be a modification of the object detection apparatus according to the embodiment or application to other objects without changing the gist described in each claim.

  For example, in the above-described embodiment, the case where LIDAR2 is used as the radar detection unit has been described. However, position information of other vehicles existing around the host vehicle may be detected using a device other than LIDAR2. For example, a millimeter wave radar may be used.

  In the above-described embodiment, the case where the other vehicle is set as the detection target object has been described. However, the present invention may be applied to detection of an object other than the other vehicle.

  DESCRIPTION OF SYMBOLS 1 ... Object detection apparatus, 2 ... LIDAR, 3 ... Camera, 4 ... ECU, 5 ... Brake part.

Claims (3)

In an object detection device that detects an object,
Radar detecting means for detecting a detection point of the object by radar;
Object direction detection means for detecting the direction of the object based on a captured image of the object;
A template selection unit that includes a plurality of templates having different shapes, selects a template having an L shape when the direction of the object is oblique, and selects a linear template when the direction of the object is not oblique. When,
Contour detection means for detecting the contour of the object by applying the template selected by the template selection means to a point sequence of detection points detected by the radar detection means;
An object detection apparatus comprising: An object detection means for detecting the center position and orientation of the object based on the position of the contour of the object detected by the contour detection means;
The object detection apparatus according to claim 1. The object is another vehicle and is mounted on the own vehicle to detect the other vehicle.
The object detection apparatus according to claim 1 or 2 .

Images