JP6565650B2 - Object detection apparatus and object detection method - Google Patents

Description

  The present invention relates to an object detection device and the like.

  As a method of detecting an object from a photographed image by attaching a camera that reflects the outside world to a moving body such as a vehicle, there is the related art 1 using an optical flow. Prior Art 1 extracts feature points of an object from each captured image, associates each feature point of each captured image in time series order, calculates a positional deviation amount, and determines whether the object has a road surface pattern or a three-dimensional object. It is determined whether it is. In the following description, the misregistration amount is appropriately expressed as a flow amount.

  The flow amount of the three-dimensional object is larger than the flow amount of the road surface pattern. For this reason, in the related art 1, when the flow amount is equal to or greater than the threshold value, it is determined that the object is a three-dimensional object, and when the flow amount is less than the threshold value, it is determined that the object is a road surface pattern.

  FIG. 18 is a diagram for explaining the flow amount of the three-dimensional object and the flow amount of the road surface pattern. As shown in FIG. 18, a camera is attached to the vehicle 10, and the camera at time 1 before the vehicle 10 moves is referred to as a camera 11a, and the camera at time 2 after the vehicle 10 moves is referred to as a camera 11b. To do. A line segment passing through the camera 11a and the point of interest 50a of the three-dimensional object 50 is defined as a line segment 1. A line segment passing through the camera 11b and the point of interest 50a is defined as a line segment 2a. A line segment passing through the camera 11b and the point 55a of the road surface pattern 55 is defined as a line segment 2b.

  When the camera 11a moves to the camera 11b, the parallax change F1 of the virtual road surface pattern is the difference between the first intersection and the second intersection. The first intersection is an intersection between the road surface 5 and the line segment 1, and the second intersection is an intersection between the line segment 2 b and the road surface 5.

  The parallax change F2 of the virtual three-dimensional object is the difference between the first intersection and the third intersection. The third intersection is an intersection between the line segment 2 a and the road surface 5. Since the parallax change F2 is larger by 1A than the parallax change F1, it can be seen that the flow amount of the three-dimensional object is larger than the flow amount of the road surface pattern.

  Here, when identifying a three-dimensional object from the magnitude of the flow amount, if a three-dimensional object and a road surface are identified by only one feature point, it is affected by noise. For this reason, in the prior art 1, a cluster of feature points is extracted, and an object is determined as a three-dimensional object if there are a certain number or more of feature points having a flow amount greater than or equal to a threshold value in the vicinity region.

  However, in the related art 1, when the moving direction of the camera and the shape direction of the three-dimensional object are the same, the time series correspondence of the feature points is not uniquely determined, the flow amount cannot be calculated correctly, and the object is a three-dimensional object. It may not be possible to properly determine whether or not.

  FIG. 19 is a diagram for explaining the problem of the prior art 1. As shown in FIG. 19, when an object 60 is in the center of the image and the camera goes straight toward the object 60, the object 60 moves in the direction of the arrow. For example, it is assumed that the feature point 60a is detected from the object 60 at time 1 and the feature point 60b is detected from the object 60 at time 2. When the moving direction of the feature point matches the object shape direction, the correspondence of each feature point is not uniquely determined, and as a result, it cannot be determined whether or not the object is a three-dimensional object. For example, it is not uniquely determined whether the feature point 60a detected at time 1 corresponds to the feature point 60b detected at time 2 or the other feature points 60c and 60d.

  There is a prior art 2 that solves the problem shown in FIG. FIG. 20 is a diagram for explaining the related art 2. Prior art 2 extracts the end points of the object as feature points and associates the feature points. If it is an end point, even if the moving direction of the feature point coincides with the object shape direction, the correspondence of each feature point is uniquely determined. Prior art 2 extracts an outline of an object and identifies the extracted corner as an end point. In the example illustrated in FIG. 20, it is assumed that feature points 61 a and 61 b are detected from the object 60 at time 1 and feature points 61 c and 61 d are detected from the object 60 at time 2. In this case, the feature point 61a and the feature point 61c can be associated with each other, and the feature point 61b and the feature point 61d can be associated with each other.

  Here, the direction of the contour line extracted by the conventional technique 2 is a vertical direction including a slant. As a characteristic of the three-dimensional object, the contour line appears in the vertical direction, so that the vertical contour line is dominant, and even if there is a horizontal contour line, it is often different from the moving direction of the camera, and the correspondence is easily determined uniquely. In the following description, the vertical contour line is referred to as a contour line.

  FIG. 21 is a diagram for explaining a process of extracting a contour line. In the prior art 2, for each feature point 62 obtained from the image, the feature point 62 adjacent in the vertical direction is connected as the feature point of the same object, thereby extracting the outline 63.

JP 2014-104853 A International Publication No. 2014/017318

  However, in the above-described conventional technology, there is a problem in that it is not possible to determine whether an object is a pattern on a road surface or a stationary solid object if a part of a contour line of an object specified from an image is missing.

  FIG. 22 and FIG. 23 are diagrams for explaining the problems of the prior art 2. When the contour line is actually extracted, when the distance between the camera and the object is long, blurring occurs near the boundary between the background and the contour line due to insufficient resolution, and the contour line is missing. As shown in FIG. 22, at time 1, a contour line 65a is detected from the object 60. However, a lack 66 occurs due to insufficient resolution. On the other hand, when the object 65 approaches the camera at time 2, the object 65 is enlarged, and thus the contour 65 b is detected from the object 65.

  Due to the presence of the chip 66 shown in FIG. 22, the flow amount of the object cannot be detected correctly, and it cannot be determined whether the object is a pattern on the road surface or a three-dimensional object.

  As shown in FIG. 23, the contour line actually detected at time 1 is the contour line 65 a, but the appropriate end point is the end point 67. Here, assuming that the contour line is the contour line 65a, the flow amount of the object at time 1 and time 2 is the flow amount 2A. On the other hand, the appropriate flow amount based on the end point 67 is the flow amount 2B. For this reason, when chipping occurs, the flow amount 2A becomes larger than the appropriate flow amount 2B.

  When the flow amount 2A is larger than the appropriate flow amount 2B, it is similar to the flow amount of the three-dimensional object. Therefore, although the object 65 is a pattern on the road surface, the object 65 is erroneously a three-dimensional object. It may be judged.

  In one aspect, the present invention is an object detection capable of determining whether a target is a pattern on a road surface or a stationary solid object even when a part of the outline of the target specified from the image is missing. An object is to provide an apparatus and an object detection method.

  In the first plan, the object detection apparatus includes a specifying unit, a calculation unit, and a determination unit. The specifying unit specifies the first contour line of the object shown in the image captured at the first timing. Further, the specifying unit specifies a second contour line of the object that appears in an image captured at a second timing different from the first timing. The calculation unit uses the moving direction of the vehicle, the moving speed, and the time difference between the first timing and the second timing, and when the object is a road surface pattern, the first contour line is the second timing. The pattern predicted contour line to be photographed is calculated. The calculation unit calculates a three-dimensional object predicted contour that the first contour should be photographed at the second timing when the target is a three-dimensional object. When the length of the pattern predicted contour is closer to the length of the second contour than the length of the three-dimensional predicted contour among the pattern predicted contour and the three-dimensional predicted contour, Judged as a pattern. The determination unit determines that the target object is a three-dimensional object when the length of the three-dimensional object predicted contour line is closer to the length of the second contour line than the length of the pattern predicted contour line.

  Even when a part of the contour line of the target specified from the image is missing, it can be determined whether the target is a pattern on the road surface or a stationary solid object.

FIG. 1 is a diagram for explaining the characteristics when the object is a road surface and when there is no outline in the outline. FIG. 2 is a diagram for explaining the characteristics when the object is a three-dimensional object and when there is no outline in the outline. FIG. 3 is a functional block diagram illustrating the configuration of the object detection apparatus according to the first embodiment. FIG. 4 is a diagram illustrating an example of the data structure of the contour line DB according to the first embodiment. FIG. 5 is a flowchart illustrating a process in which the specifying unit according to the first embodiment specifies the contour line. FIG. 6 is a flowchart illustrating a process in which the contour line matching unit according to the first embodiment matches the contour line. FIG. 7 is a flowchart illustrating the similarity determination process executed by the contour line matching unit according to the first embodiment. FIG. 8 is a flowchart illustrating a process in which the calculation unit according to the first embodiment calculates a predicted contour line. FIG. 9 is a flowchart illustrating the process of the determination unit according to the first embodiment. FIG. 10 is a flowchart illustrating the processing procedure of the object detection apparatus according to the first embodiment. FIG. 11 is a functional block diagram illustrating the configuration of the object detection apparatus according to the second embodiment. FIG. 12 is a diagram (1) for explaining the process of the calculation unit according to the second embodiment. FIG. 13 is a diagram (2) for explaining the process of the calculation unit according to the second embodiment. FIG. 14 is a flowchart illustrating a process in which the calculation unit according to the second embodiment calculates a predicted contour line. FIG. 15 is a flowchart illustrating the process of the determination unit according to the second embodiment. FIG. 16 is a flowchart illustrating the processing procedure of the object detection apparatus according to the second embodiment. FIG. 17 is a diagram illustrating an example of a computer that executes an object detection program. FIG. 18 is a diagram for explaining the flow amount of the three-dimensional object and the flow amount of the road surface pattern. FIG. 19 is a diagram for explaining the problem of the prior art 1. FIG. 20 is a diagram for explaining the related art 2. FIG. 21 is a diagram for explaining a process of extracting a contour line. FIG. 22 is a diagram (1) for explaining the problems of the conventional technique 2. FIG. 23 is a diagram (2) for explaining the problems of the conventional technique 2.

  Embodiments of an object detection apparatus and an object detection method disclosed in the present application will be described below in detail with reference to the drawings. Note that the present invention is not limited to the embodiments.

  The moving amount of the vehicle is used to indicate at which position in the image taken at the current time the contour line of the object identified from the image taken at the past time by the camera of the moving vehicle. Thus, it can be predicted. At this time, there are the following characteristics.

  Characteristic: A predicted pattern contour indicating the contour of the object at the current time when the object is assumed to be a road surface pattern, and a solid indicating the contour of the object at the current time when the object is assumed to be a three-dimensional object. When comparing with the predicted object contour line, the predicted pattern contour line is always longer than the predicted solid object contour line.

  Focusing on the above characteristics, the magnitude relationship among the actual contour length at the current time, the pattern predicted contour length, and the three-dimensional object predicted contour length is as follows according to the conditions.

(1) When the object is a “road surface”, the size relationship between the actual contour length at the current time, the pattern predicted contour length, and the three-dimensional object predicted contour length depends on whether or not the contour is missing. It becomes as follows.
(1-1) When the outline specified at the past time is not missing: the length of the three-dimensional object predicted outline <the length of the pattern predicted outline = the actual outline length at the current time (1-2 ) If the outline specified at the past time is missing: the length of the three-dimensional object prediction outline <the length of the pattern prediction outline <the length of the actual outline at the current time

  FIG. 1 is a diagram for explaining the characteristics when the object is a road surface and when there is no outline in the outline. (1-1) in FIG. 1 will be described. The contour line 70 is a contour line having no chip specified at a past time. The contour line 71 is an actual contour line at the current time. The contour line 72 a is a three-dimensional object predicted contour line predicted based on the contour line 70. The contour line 72 b is a pattern predicted contour line predicted based on the contour line 70. As shown in FIG. 1, the length of the contour line 71 is closer to the length of the contour line 72b than the length of the contour line 72a.

  (1-2) in FIG. 1 will be described. The contour line 75 is a contour line having a defect identified at a past time. The contour line 76 is an actual contour line at the current time. The contour line 77 a is a three-dimensional object predicted contour line predicted based on the contour line 75. The contour line 77 b is a pattern predicted contour line predicted based on the contour line 75. As shown in FIG. 1, the length of the contour line 76 is closer to the length of the contour line 77b than the length of the contour line 77a.

  The length of the three-dimensional object predicted contour line is shorter than the length of the pattern predicted contour line. If there is no missing outline at the past time, the length of the predicted pattern outline is equal to the actual outline length at the current time. If the outline is missing at the past time, the length of the predicted pattern outline is shorter than the actual outline length at the current time. The magnitude relationship with the length of the line does not change. For this reason, when the object is a road surface regardless of the presence or absence of a chip, the length of the pattern predicted contour is longer than the length of the three-dimensional predicted contour. Get closer.

(2) When the object is a “three-dimensional object”, the relationship between the actual contour length at the current time, the pattern predicted contour length, and the three-dimensional object predicted contour length depending on whether the contour line is missing or not. Is as follows.
(2-1) When the outline specified at the past time is not missing: the length of the three-dimensional object prediction outline = the actual outline length at the current time <the length of the pattern prediction outline (2-2 ) If there is a missing outline specified in the past time: Length of the three-dimensional object prediction outline <Length of the actual outline at the current time <Length of the pattern prediction outline

  FIG. 2 is a diagram for explaining the characteristics when the object is a three-dimensional object and when there is no outline in the outline. (2-1) in FIG. 2 will be described. The contour line 80 is a contour line having no chip specified at a past time. The contour line 81 is an actual contour line at the current time. The contour line 82 a is a three-dimensional object predicted contour line predicted based on the contour line 80. The contour line 82 b is a pattern predicted contour line predicted based on the contour line 80. As shown in FIG. 2, the length of the contour line 81 is closer to the length of the contour line 82a than the length of the contour line 82b.

  (2-2) in FIG. 2 will be described. The contour line 85 is a contour line having a defect identified at a past time. The contour line 86 is an actual contour line at the current time. The contour line 87 a is a three-dimensional object predicted contour line predicted based on the contour line 85. The contour line 87 b is a pattern predicted contour line predicted based on the contour line 85. As shown in FIG. 2, the length of the contour line 86 is closer to the length of the contour line 87a than the length of the contour line 87b.

  If the object is a three-dimensional object, the length of the three-dimensional object predicted contour is closer to the actual contour length at the current time than the length of the pattern predicted contour. However, if the outline of the past time is largely missing and the outline of the current time is not missing, the situation does not occur practically. Therefore, the above (2-2) is defined as having little missing. .

  The object detection apparatus according to the first embodiment determines whether the object is a three-dimensional object based on the contents described in (1-1), (1-2), (2-1), and (2-2). It is determined whether it is a pattern of as follows. The object detection device compares the length of the three-dimensional object prediction contour, the length of the pattern prediction contour, and the actual contour length of the current time. If it is closer to the length of the pattern predicted contour line than the length of, it is determined that the object is a road surface. On the other hand, the object detection device determines that the object is a three-dimensional object when the actual length of the contour line is closer to the length of the three-dimensional object predicted contour line than the length of the pattern predicted contour line.

  Next, an example of the configuration of the object detection apparatus according to the first embodiment will be described. FIG. 3 is a functional block diagram illustrating the configuration of the object detection apparatus according to the first embodiment. As illustrated in FIG. 3, the object detection apparatus 100 includes a camera 110, a movement amount detection unit 120, a contour DB (database) 130, a video input unit 140, a specification unit 150, and a contour matching unit 160. A calculation unit 170, a determination unit 180, and an output unit 190.

  The camera 110 is a camera that is installed on a moving body of a vehicle and captures an image in the imaging direction. The camera 110 outputs the captured video data to the video input unit 140. For example, video data is information in which a plurality of image frames are continuous in chronological order.

  The movement amount detection unit 120 detects the movement amount of the vehicle and outputs information on the detected movement amount to the contour line matching unit 160 and the calculation unit 170. The moving amount of the vehicle includes the moving direction and moving speed of the vehicle. The movement amount detection unit 120 detects the movement amount using various sensors installed in the vehicle.

  The contour DB 130 is a storage unit that stores information on contour lines detected from each image frame. FIG. 4 is a diagram illustrating an example of the data structure of the contour line DB according to the first embodiment. As shown in FIG. 4, the contour DB 130 associates contour numbers, start point coordinates, end point coordinates, and frame identification numbers.

  The contour line number is a number that uniquely identifies the contour line. The starting point coordinates are the starting point coordinates of the contour line. The end point coordinates are the end point coordinates of the contour line. The frame identification number is a number that uniquely identifies an image frame. In the example shown in the record in the first row in FIG. 4, the start point coordinates of the contour line with the contour line identification number “101” are “x1, y1” and the end point coordinates are “x2, y2”. It indicates that it exists in the image frame with the identification number “F01”.

  Returning to the description of FIG. The video input unit 140 is a processing unit that receives input of video data from the camera 110. The video input unit 140 outputs the received video data to the specifying unit 150. When the video data is analog video data, the video input unit 140 converts the video data into digital video data, and outputs the digital video data to the specifying unit 150. Further, when the video data is color video data, the video input unit 140 converts the video data into monochrome video data, and outputs the monochrome video data to the specifying unit 150.

  The identifying unit 150 is a processing unit that identifies the first contour line of the object that appears in the image frame at the first timing based on the video data, and identifies the second contour line that appears in the image frame at the second timing. is there. For example, the first timing corresponds to a past time that is a predetermined time before the current time. The second timing corresponds to the current time.

  The specifying unit 150 sequentially executes a process of extracting feature points from the image frame and a process of specifying a contour line.

  A process in which the specifying unit 150 extracts feature points from the image frame will be described. The specifying unit 150 performs edge detection on each image frame of the video data acquired from the video input unit 140, and generates a feature point image. The specifying unit 150 operates a general differential operator such as Sobel, compares the edge strength obtained by the differential processing with a predetermined edge strength threshold, and extracts an edge that is equal to or greater than the edge strength threshold. The specifying unit 150 compares the edge strengths of two adjacent horizontal pixels with respect to the extracted edge, and performs a thinning process to obtain an edge point when the peak value is reached. The identifying unit 150 extracts the extracted edge points as feature points.

  A process in which the specifying unit 150 specifies the contour line of the object included in the image frame will be described. The specifying unit 150 is a processing unit that specifies a contour line by connecting feature points extracted by the above processing. The identifying unit 150 registers the identified contour information in the contour DB 130. For example, the specifying unit 150 assigns a contour line number to the contour line, associates the contour line number, the start point coordinates, the end point coordinates, and the frame identification number, and outputs them to the contour line collation unit 160.

  FIG. 5 is a flowchart illustrating a process in which the specifying unit according to the first embodiment specifies the contour line. As shown in FIG. 5, the specifying unit 150 selects an unprocessed feature point A (step S10). The identifying unit 150 determines whether there is a feature point B that can be connected to an adjacent pixel on one line of the feature point A (step S11).

  When there is no feature point B that can be connected to an adjacent pixel on one line of the feature point A (No in step S11), the specifying unit 150 proceeds to step S19. On the other hand, when there is a feature point B that can be connected to an adjacent pixel on one line of the feature point A (step S11, Yes), the specifying unit 150 determines whether there are a plurality of feature points B. (Step S12).

  If there are not a plurality of feature points B (No at Step S12), the specifying unit 150 proceeds to Step S14. On the other hand, when there are a plurality of feature points B (step S12, Yes), the specifying unit 150 selects an optimum point from the plurality of feature points B (step S13). In step S <b> 13, the specifying unit 150 selects, as the optimum point, the feature point B whose edge strength and inclination are most similar to the edge strength and inclination of the feature point A.

  If the contour number of the selected feature point B is not unregistered (No at Step S14), the specifying unit 150 proceeds to Step S19. On the other hand, when the contour line number of the selected feature point B is not registered (step S14, Yes), the specifying unit 150 determines that the feature point C that can be connected to the feature point B on the same line as the feature point A. It is determined whether or not it exists (step S15).

  If there is no feature point C that can be connected to the feature point B on the same line as the feature point B (step S15, No), the specifying unit 150 assigns the same contour line number to the feature points A and B. (Step S16), the process proceeds to Step S19.

  On the other hand, when there is a feature point C that can be connected to the feature point B on the same line as the feature point B (step S15, Yes), the specifying unit 150 determines that the feature point C is a feature point rather than the feature point A. It is determined whether or not the feature point connected to B is appropriate (step S17). If the feature point C is not more appropriate than the feature point A (No at Step S17), the specifying unit 150 proceeds to Step S16. For example, the specifying unit 150 determines that the feature point C is more than the feature point A when the strength and inclination of the feature point B are more similar to the strength and inclination of the feature point A than the feature point C. Judge that it is not appropriate.

  When the feature point C is more appropriate than the feature point A (Yes in step S17), the specifying unit 150 assigns the same contour line number to the feature points B and C (step S18).

  If the processing has not been completed for all feature points (No at Step S19), the specifying unit 150 proceeds to Step S10. When the processing has been completed for all feature points (Yes in step S19), the specifying unit 150 ends the processing.

  Returning to the description of FIG. The contour line matching unit 160 compares the contour line of the current time specified by the specifying unit 150 with the contour line obtained by moving the past contour line stored in the contour line DB 130 according to the amount of movement of the vehicle. It is a process part which collates whether there exists any outline. If there is an identical contour line, the contour line matching unit 160 registers the contour line information at the current time in the contour line DB 130 and deletes the past contour line information that has been collated from the contour line DB 130. . On the other hand, when the same contour line does not exist and the contour line is lost, the contour line matching unit 160 deletes the lost contour line information from the contour DB 130.

  The contour line collation unit 160 outputs information on the contour line at the current time when the collation is successful and information on the contour line at the past time to the calculation unit 170.

  FIG. 6 is a flowchart illustrating a process in which the contour line matching unit according to the first embodiment matches the contour line. As shown in FIG. 6, the contour line matching unit 160 determines whether or not there are two or more image frames corresponding to the contour line (step S20). If it is not two frames or more (step S20, No), the contour line matching unit 160 proceeds to step S30. On the other hand, if there are two or more frames (step S20, Yes), the contour line matching unit 160 proceeds to step S21.

  The contour line matching unit 160 calculates the moving direction of the time series line segment (step S21). For example, the contour line matching unit 160 calculates the movement amount of the time-series line segment from the movement amount (translation amount, rotation amount) of the vehicle and the installation parameters of the camera 110 using a geometric condition called epipolar constraint.

  The contour line matching unit 160 determines whether or not there is a line segment in the movement direction (step S22). For example, the contour line matching unit 160 determines that there is a line segment in the movement direction when the contour line at the current time exists in the movement direction of the time-series line segment based on the contour line at the past time. .

  When there is a line segment in the movement direction (step S22, Yes), the contour line matching unit 160 proceeds to step S23. When there is no line segment in the movement direction (No at Step S22), the contour line matching unit 160 proceeds to Step S26.

  The contour line matching unit 160 calculates the similarity between the contour line at the past time and the contour line at the current time (step S23). If the similarity is not greater than or equal to the threshold (No at Step S24), the contour line matching unit 160 proceeds to Step S26. On the other hand, when the similarity is equal to or greater than the threshold (Yes in step S24), the contour matching unit 160 determines that the contour at the past time and the contour at the current time are the same line segment. Information is updated (step S25). In step S <b> 25, the contour line matching unit 160 registers the contour line identification number, the start point coordinates, the end point coordinates, and the frame identification number of the current contour line in the contour line DB 130. The contour line matching unit 160 removes the contour line identification number, the start point coordinates, the end point coordinates, and the frame identification number of the past contour lines from the contour line DB 130.

  The process proceeds to step S26. The contour line matching unit 160 calculates the number of lost times (step S26). For example, in step S <b> 26, the contour matching unit 160 determines that the past contour is lost if the contour at the current time does not exist in the movement direction or the similarity is not equal to or greater than the threshold. A predetermined value is added to.

  If the number of lost times is not greater than or equal to the threshold (No at Step S27), the contour line matching unit 160 proceeds to Step S29. On the other hand, if the number of lost times is equal to or greater than the threshold (Yes at Step S27), the contour line matching unit 160 removes the corresponding contour line from the contour line DB 130 (Step S28), and proceeds to Step S29.

  If the loop processing has not ended (No at Step S29), the contour line matching unit 160 proceeds to Step S21. On the other hand, when the loop processing is completed (Yes at Step S29), the contour line matching unit 160 newly registers information on the contour line at the unsupported current time in the contour DB 130 (Step S30).

  Next, the processing procedure of the similarity determination process in which the contour line matching unit 160 described in steps S23 and S24 in FIG. 6 calculates the similarity and determines whether the similarity is equal to or greater than a threshold value will be described. . FIG. 7 is a flowchart illustrating the similarity determination process executed by the contour line matching unit according to the first embodiment. As shown in FIG. 7, the contour line matching unit 160 calculates the overlapping position of the contour line of the past time moved in the movement direction and the contour line of the current time (step S40).

  The contour line matching unit 160 adjusts the scale of the contour line at the past time moved (step S41), and calculates the contour line at the past time and the center of gravity of the contour line at the current time (step S42). The contour line matching unit 160 calculates the amount of horizontal displacement of each contour line for the current y coordinate (step S43).

  If the processing has not been completed for all the y coordinates (No in step S44), the contour line matching unit 160 proceeds to step S43. When the processing has been completed for all y coordinates (step S44, Yes), the contour line matching unit 160 determines whether or not the total amount of deviation is within a threshold value (step S45).

  If the total amount of deviation is within the threshold (Yes in step S45), the contour matching unit 160 determines that the similarity between the corresponding contour at the past time and the contour at the current time is greater than or equal to the threshold. It is determined that there is (step S46). On the other hand, when the total value of the deviation amounts is not within the threshold (No in step S45), it is determined that the similarity is less than the threshold (step S47).

  Returning to the description of FIG. The calculation unit 170 calculates the pattern prediction contour and the three-dimensional object prediction contour described with reference to FIGS. 1 and 2 and the like based on the contour information of the past time and the information on the movement amount of the vehicle. Is a processing unit. The information regarding the moving amount includes, for example, the moving direction, moving speed, and interframe time of the vehicle. The interframe time corresponds to the time from the past time to the current time. The calculation unit 170 outputs the contour information of the past time, the pattern prediction contour information, and the three-dimensional object prediction contour information to the determination unit 180.

  An example of processing in which the calculation unit 170 calculates the pattern predicted contour line based on the contour line at the past time will be described. Note that the contour line of the past time is the contour line of the past time that has been successfully matched with the contour line of the current time.

  The calculation unit 170 assumes that the object is a road surface pattern, and calculates the three-dimensional coordinates of the lower end and the upper end of the contour line at a past time. For example, the calculation unit 170 uses the image frame of the contour line of the past time and the same contour line included in the next image frame of this image frame, based on the principle of the stereo image, The three-dimensional coordinates of the lower end and the upper end of the contour line are calculated.

  The calculating unit 170 calculates a moving distance by multiplying the moving speed of the vehicle and the time between frames, and moves the lower end and the upper end of the contour line of the past time according to the moving distance and the moving direction. Then, the lower end and the upper end of the contour line at the current time are calculated. The calculation unit 170 calculates a line segment connecting the calculated lower and upper ends as a pattern predicted contour line.

  An example of processing in which the calculation unit 170 calculates a three-dimensional object predicted contour line based on the contour line at a past time will be described. The calculation unit 170 assumes that the road surface is a three-dimensional object, and calculates the three-dimensional coordinates of the lower end and the upper end of the contour line at the past time. The process of calculating the three-dimensional coordinates of the lower end of the contour line is the same as the process of calculating the pattern predicted contour line.

  After calculating the lower end of the contour line at the past time, the calculation unit 170 calculates the upper end three-dimensional coordinate with a restriction that the depth of the lower end and the upper end of the three-dimensional coordinate are the same. For example, the x-axis value and the y-axis value of the three-dimensional coordinates of the lower end and the upper end are the same value. The calculation unit 170 may calculate the z-axis value at the upper end based on the principle of the stereo image, or from the geometrical relationship based on the camera setting parameter and the coordinate position on the image frame, z An axis value may be calculated.

  The calculating unit 170 calculates a moving distance by multiplying the moving speed of the vehicle and the time between frames, and moves the lower end and the upper end of the contour line of the past time according to the moving distance and the moving direction. Then, the lower end and the upper end of the contour line at the current time are calculated. The calculation unit 170 calculates a line segment connecting the calculated lower and upper ends as a three-dimensional object predicted contour line.

  FIG. 8 is a flowchart illustrating a process in which the calculation unit according to the first embodiment calculates a predicted contour line. The predicted contour corresponds to a pattern predicted contour and a three-dimensional object predicted contour. As illustrated in FIG. 8, the calculation unit 170 determines whether or not the selected contour line at the past time has succeeded in contour line collation (step S50). If the contour matching is not successful (No at Step S50), the calculating unit 170 proceeds to Step S57.

  On the other hand, when the selected contour line at the past time has succeeded in contour matching (step S50, Yes), the calculation unit 170 determines that the lower end of the contour line at the past time is the road surface. Dimensional coordinates are calculated (step S51). The calculation unit 170 calculates the three-dimensional coordinates when the upper end of the contour line at the past time is taken as the road surface (step S52).

  The calculation unit 170 calculates the three-dimensional coordinates when the upper end of the contour line at the past time is a three-dimensional object (step S53). The calculation unit 170 calculates the three-dimensional coordinates of the current time from the three-dimensional coordinates of the past time (step S54). The calculation unit 170 calculates the image position from the three-dimensional coordinates of the current time (step S55).

  The calculation unit 170 sets a line segment connecting the upper and lower image positions as a predicted contour line (step S56). The calculation unit 170 determines whether or not to end the loop processing of the contour line in the contour line DB 130 (step S57). If the calculation unit 170 does not end the loop process (No at Step S57), the calculation unit 170 proceeds to Step S50. If the calculation unit 170 ends the loop process (step S57, Yes), the calculation unit 170 ends the process.

  Returning to the description of FIG. The determination unit 180 compares the length of the contour line at the current time, the length of the pattern predicted contour line, and the length of the three-dimensional object predicted contour line, and determines whether the contour line at the current time is a road surface pattern or It is a processing part which determines whether it is a thing. The determination unit 180 outputs the determination result to the output unit 190.

  The determination unit 180 determines that the contour of the current time is a road surface pattern when the length of the contour of the current time is closer to the length of the predicted pattern contour than the length of the three-dimensional object predicted contour. To do. On the other hand, when the length of the contour line at the current time is closer to the length of the three-dimensional object predicted contour line than the length of the pattern predicted contour line, the determination unit 180 determines that the contour line at the current time is a three-dimensional object. judge.

  Specifically, the determination unit 180 calculates a first difference value between the length of the contour line at the current time and the length of the pattern predicted contour line. The determination unit 180 calculates a second difference value between the length of the contour line at the current time and the length of the three-dimensional object predicted contour line. When the first difference value is smaller than the second difference value, the determination unit 180 determines that the contour line at the current time is a road surface pattern. The determination unit 180 determines that the contour line at the current time is a three-dimensional object when the second difference value is smaller than the first difference value.

  FIG. 9 is a flowchart illustrating the process of the determination unit according to the first embodiment. As illustrated in FIG. 9, when the outline of the past time has not been successfully collated (No in Step S60), the determination unit 180 proceeds to Step S64. On the other hand, when the outline of the past time has been successfully collated (Yes at Step S60), the determination unit 180 proceeds to Step S61.

  When the length of the three-dimensional object predicted contour line is closer to the current time contour length than the pattern predicted contour length (step S61, Yes), the determination unit 180 determines that the contour line is the contour of the three-dimensional object. It determines with it being a line (step S62), and transfers to step S64.

  When the length of the three-dimensional object predicted contour is not closer to the length of the contour at the current time than the length of the predicted pattern contour (step S61, No), the determination unit 180 determines that the contour is a road surface pattern. (Step S63), and the process proceeds to step S64.

  If the determination unit 180 does not end the contour loop processing in the contour DB 130 (step S64, No), the determination unit 180 proceeds to step S60. The determination unit 180 ends the processing when the loop processing of the contour line in the contour DB 130 is ended (step S64, Yes).

  Returning to the description of FIG. The output unit 190 is a device that acquires a determination result from the determination unit 180 and outputs information of a contour line determined to be a three-dimensional object out of the contour lines at the current time to an external device.

  Next, a processing procedure of the object detection apparatus 100 according to the first embodiment will be described. FIG. 10 is a flowchart illustrating the processing procedure of the object detection apparatus according to the first embodiment. As illustrated in FIG. 10, the object detection apparatus 100 acquires video data from the camera 110 (step S101). The specifying unit 150 of the object detection apparatus 100 extracts feature points (step S102) and extracts contour lines (step S103).

  The calculation unit 170 of the object detection apparatus 100 creates a pattern predicted contour line and a three-dimensional object predicted contour line (step S104). The determination unit 180 of the object detection apparatus 100 determines whether the contour line is a three-dimensional object or a road surface pattern (step S105).

  The determination unit 180 outputs the determination result (step S106). When the object detection apparatus 100 continues the process (step S107, Yes), the object detection apparatus 100 proceeds to step S101. If the object detection apparatus 100 does not continue the process (No at Step S107), the object detection apparatus 100 ends the process.

  Next, the effect of the object detection apparatus 100 according to the first embodiment will be described. The object detection apparatus 100 compares the length of the three-dimensional object predicted contour, the length of the pattern predicted contour, and the actual contour length at the current time. When the length of the predicted pattern contour is closer to the length of the line, it is determined that the object is a road surface. On the other hand, the object detection apparatus 100 determines that the object is a three-dimensional object when the actual length of the contour line is closer to the length of the three-dimensional object predicted contour line than the length of the pattern predicted contour line. As a result, even if a part of the outline of the object specified from the image is missing, it is possible to determine whether the target is a pattern on the road surface or a stationary solid object.

  FIG. 11 is a functional block diagram illustrating the configuration of the object detection apparatus according to the second embodiment. As shown in FIG. 11, the object detection apparatus 200 includes a camera 210, a movement amount detection unit 220, a contour DB (database) 230, a video input unit 240, a specification unit 250, and a contour matching unit 260. A calculation unit 270, a determination unit 280, and an output unit 290.

  In FIG. 11, the description regarding the camera 210, the movement amount detection unit 220, the contour line DB 230, and the video input unit 240 is the description regarding the camera 110, the movement amount detection unit 120, the contour line DB 130, and the video input unit 140 shown in FIG. 3. It is the same. Moreover, the description regarding the specific | specification part 250 and the outline collation part 26 is the same as the description regarding the specific part 150 and the outline collation part 160 which were shown in FIG.

  The calculation unit 270 is a processing unit that calculates a pattern predicted contour line and a three-dimensional object predicted contour line based on information on the contour line of the past time and information on the movement amount of the vehicle. The process in which the calculation unit 270 calculates the pattern predicted contour and the three-dimensional object predicted contour is the same as the process in which the calculation unit 170 described in the first embodiment calculates the pattern predicted contour and the three-dimensional object predicted contour. .

  In addition to the above processing, the calculation unit 270 predicts the luminance value of the pattern predicted contour line and the luminance value of the three-dimensional object predicted contour line based on the luminance value of the contour line at the past time.

  The calculating unit 270 outputs the contour information of the past time, the pattern predicted contour information, and the three-dimensional object predicted contour information to the determination unit 280.

  An example of processing in which the calculation unit 270 predicts the luminance value of the pattern predicted contour line based on the luminance value of the contour line at the past time will be described. FIG. 12 is a diagram (1) for explaining the process of the calculation unit according to the second embodiment. An outline 90 shown in FIG. 12 indicates an outline at a past time, and an outline 94 indicates a pattern prediction outline. For convenience of explanation, it is assumed that the contour line 90 includes pixels 90a to 90e, and the contour line 94 includes pixels 94a to 94m. In each of the pixels 90a to 90e of the outline 90, a luminance value based on an image frame at a past time is set. The luminance values of the pixels 94a to 94m of the contour line 94 are blank in the initial state.

  The calculation unit 270 distributes the luminance values of the pixels 90a to 90e of the contour line 90 to the pixels 94a to 94m of the contour line 94 at equal intervals. In the example shown in FIG. 12, the calculation unit 270 sets the luminance value of the pixel 90a to the luminance value of the pixel 94a, sets the luminance value of the pixel 90b to the luminance value of the pixel 94d, and sets the luminance value of the pixel 90c. , The luminance value of the pixel 94g is set. The calculation unit 270 sets the luminance value of the pixel 90d to the luminance value of the pixel 94j, and sets the luminance value of the pixel 90e to the luminance value of the pixel 94m.

  The calculation unit 270 calculates the luminance value of the blank pixel among the pixels 94a to 94m of the contour line 94 by interpolation based on the luminance values of the preceding and succeeding pixels. For example, the calculation unit 270 calculates the luminance values of the pixels 94b and 94c by interpolation between the luminance value of the pixel 94a and the luminance value of the pixel 94d.

  Next, an example of processing in which the calculation unit 270 predicts the luminance value of the three-dimensional object predicted contour line based on the luminance value of the contour line at the past time will be described. FIG. 13 is a diagram (2) for explaining the process of the calculation unit according to the second embodiment. An outline 90 illustrated in FIG. 13 indicates an outline at a past time, and an outline 95 indicates a three-dimensional object prediction outline. For convenience of explanation, it is assumed that the contour line 90 includes pixels 90a to 90e, and the contour line 95 includes pixels 95a to 95i. In each of the pixels 90a to 90e of the outline 90, a luminance value based on an image frame at a past time is set. The luminance values of the pixels 95a to 95i of the contour line 95 are blank in the initial state.

  The calculation unit 270 distributes the luminance values of the pixels 90a to 90e of the contour line 90 to the pixels 95a to 95i of the contour line 95 at equal intervals. In the example shown in FIG. 13, the calculation unit 270 sets the luminance value of the pixel 90a to the luminance value of the pixel 95a, sets the luminance value of the pixel 90b to the luminance value of the pixel 95c, and sets the luminance value of the pixel 90c. , The luminance value of the pixel 95e is set. The calculation unit 270 sets the luminance value of the pixel 90d to the luminance value of the pixel 95g, and sets the luminance value of the pixel 90e to the luminance value of the pixel 95i.

  Of the pixels 95a to 95i of the contour line 95, the calculation unit 270 calculates the luminance value of the blank pixel by interpolation based on the luminance values of the previous and subsequent pixels. For example, the calculation unit 270 calculates the luminance value of the pixel 95b by interpolation between the luminance value of the pixel 95a and the luminance value of the pixel 95c.

  The calculation unit 270 calculates the luminance value of the pattern predicted contour line and the luminance value of the three-dimensional object predicted contour line by executing the processing described with reference to FIGS. In general, since the pattern predicted contour is longer than the three-dimensional object predicted contour, the pattern predicted contour is more blurred than the three-dimensional object predicted contour.

  FIG. 14 is a flowchart illustrating a process in which the calculation unit according to the second embodiment calculates a predicted contour line. As illustrated in FIG. 14, the calculation unit 270 determines whether or not the selected contour line at the past time has succeeded in contour line collation (step S <b> 70). If the contour matching is not successful (step S70, No), the calculation unit 270 proceeds to step S76.

  On the other hand, when the selected contour line at the past time has succeeded in the contour matching (Yes in step S70), the calculation unit 270 calculates the brightness value of the contour line included in the image frame at the past time. Recording is performed (step S71).

  The calculation unit 270 calculates the three-dimensional coordinates of the contour line when the contour line at the past time is a solid object / road surface (step S72). The calculation unit 270 calculates the three-dimensional coordinates of the lower end and the upper end of the current time from the three-dimensional coordinates of the contour line of the past time (step S73).

  The calculating unit 270 calculates the luminance values of the pattern predicted contour and the three-dimensional object predicted contour from the three-dimensional coordinates at the lower end and the upper end of the current time (step S74). The calculation unit 270 calculates the luminance value of the pattern predicted contour line and the luminance value of the three-dimensional object predicted contour line (step S75).

  The calculating unit 270 determines whether or not to end the contour loop processing of the contour DB 230 (step S76). When the loop process is not finished (No at Step S76), the calculation unit 270 proceeds to Step S70. The calculation unit 270 ends the process when the loop process ends (Yes in step S76).

  Returning to the description of FIG. The determination unit 280 compares the contour line of the current time with the pattern prediction contour line, calculates the first overlap rate and the first luminance difference, and calculates the first score from the first overlap rate and the first luminance difference. calculate. The determination unit 280 compares the contour line of the current time with the three-dimensional object predicted contour line, calculates the second overlap rate and the second luminance difference, and calculates the second score from the second overlap rate and the second luminance difference. Is calculated. The determination unit 280 determines whether the contour line at the current time is a solid object or a road surface pattern based on the first score and the second score. The determination unit 280 outputs the determination result to the output unit 290.

  A process in which the determination unit 280 calculates the first score will be described. The determination unit 280 adjusts the overlapping position of the contour line of the current time and the pattern predicted contour line within the perturbation range, and specifies the position where the overlapping ratio is the highest. The determination unit 280 sets the identified highest duplication rate as the first duplication rate. The determination unit 280 compares the first duplication rate and the duplication score table to calculate a score for the first duplication rate. The duplication score table is a table for converting the duplication rate into a score, and a larger score is given as the duplication rate is larger.

  The determination unit 280 fixes the contour line at the current time at which the duplication rate is the highest and the pattern prediction contour line, and calculates a difference value between the current time contour line and the pattern prediction contour line for each pixel. Then, the first luminance difference is calculated by summing the difference values. The determination unit 280 compares the first luminance difference with the luminance difference score table, and calculates a first luminance difference score. The luminance difference score table is a table for converting a luminance difference into a score, and a larger score is given as the luminance difference is smaller.

  The determination unit 280 calculates the first score by summing the score of the first overlap rate obtained as described above and the score of the first luminance difference.

  A process in which the determination unit 280 calculates the second score will be described. The determination unit 280 adjusts the overlapping position of the contour line of the current time and the three-dimensional object predicted contour line within the perturbation range, and specifies the position where the overlapping ratio is the highest. The determination unit 280 sets the identified highest duplication rate as the second duplication rate. The determination unit 280 compares the second duplication rate and the duplication score table to calculate a score for the second duplication rate.

  The determination unit 280 fixes the contour line of the current time at which the duplication rate is the highest and the three-dimensional object predicted contour line, and sets the difference value of the luminance value between the current time contour line and the three-dimensional object predicted contour line for each pixel. The second luminance difference is calculated by calculating and summing each difference value. The determination unit 280 compares the second luminance difference and the luminance difference score table to calculate a second luminance difference score.

  The determination unit 280 calculates the second score by summing the score of the second overlap ratio obtained above and the score of the second luminance difference.

  When the first score is greater than the second score, the determination unit 280 determines that the contour line at the current time is a road surface pattern. On the other hand, when the first score is not greater than the second score, the determination unit 280 determines that the contour line at the current time is a three-dimensional object.

  FIG. 15 is a flowchart illustrating the process of the determination unit according to the second embodiment. As illustrated in FIG. 15, when the outline of the past time has not been successfully collated (No in step S80), the determination unit 280 proceeds to step S88. On the other hand, the determination part 280 transfers to step S81, when the outline of the past time has succeeded in collation (step S80, Yes).

  The determination unit 280 compares the contour line at the current time with the pattern predicted contour line, and calculates the first overlap rate and the first luminance difference (step S81). The determination unit 280 calculates a first score based on the first overlap rate and the first luminance difference (step S82).

  The determination unit 280 compares the contour line at the current time with the three-dimensional object predicted contour line, and calculates the second overlap rate and the second luminance difference (step S83). The determination unit 280 calculates a second score based on the second overlap rate and the second luminance difference (Step S84).

  The determination unit 280 determines whether or not the first score is greater than the second score (step S85). If the first score is greater than the second score (Yes at Step S85), the determination unit 280 determines that the contour line at the current time is a road surface pattern (Step S86), and proceeds to Step S88.

  On the other hand, when the first score is not greater than the second score (No at Step S85), the determination unit 280 determines that the contour line at the current time is a three-dimensional object (Step S87), and proceeds to Step S88. To do.

  If the determination unit 280 has not finished the loop processing of the contour line in the contour DB 230 (step S88, No), the determination unit 280 proceeds to step S80. On the other hand, when the loop processing of the contour line in the contour DB 230 is completed (Yes in step S88), the determination unit 280 ends the processing.

  Next, a processing procedure of the object detection apparatus 200 according to the second embodiment will be described. FIG. 16 is a flowchart illustrating the processing procedure of the object detection apparatus according to the second embodiment. As shown in FIG. 16, the object detection apparatus 200 acquires video data from the camera 210 (step S201). The specifying unit 250 of the object detection apparatus 200 extracts feature points (step S202) and extracts contour lines (step S203).

  The calculation unit 270 of the object detection device 200 creates a pattern predicted contour and a three-dimensional object predicted contour (step S204). The determination unit 280 of the object detection apparatus 200 determines whether the contour line is a three-dimensional object or a road surface pattern based on the first score and the second score (step S205).

  The determination unit 280 outputs a determination result (step S206). When the object detection apparatus 200 continues the process (step S207, Yes), the object detection apparatus 200 proceeds to step S201. If the object detection apparatus 200 does not continue the process (No at Step S207), the object detection apparatus 200 ends the process.

  Next, effects of the object detection apparatus 200 according to the second embodiment will be described. The object detection apparatus 200 calculates the first score of the pattern predicted contour based on the first overlap rate and the first luminance difference, and calculates the three-dimensional object predicted contour based on the second overlap rate and the second luminance difference. A second score is calculated, and it is determined whether the object is a three-dimensional object or a road surface pattern based on the first score and the second score. In addition to the processing of the first embodiment, the object detection apparatus 200 of the second embodiment further uses the difference in luminance value of each contour line to perform the determination, and thus more accurately determines whether the object is a three-dimensional object or the road surface. Whether it is a pattern can be determined.

  Next, an example of a computer that executes an object detection program that realizes the same functions as those of the object detection apparatuses 100 and 200 described in the above embodiment will be described. FIG. 17 is a diagram illustrating an example of a computer that executes an object detection program.

  As illustrated in FIG. 17, the computer 300 includes a CPU 301 that executes various arithmetic processes, an input device 302 that receives input of data from a user, and a display 303. The computer 300 includes a reading device 304 that reads a program and the like from a storage medium, an interface device 305 that exchanges data with other computers via a network, and a camera 306. The computer 300 also includes a RAM 307 that temporarily stores various types of information and a hard disk device 308. The devices 301 to 308 are connected to the bus 309.

  The hard disk device 308 has a specific program 308a, a collation program 308b, a calculation program 308c, and a determination program 308d. The CPU 301 reads the specific program 308a, the collation program 308b, the calculation program 308c, and the determination program 308d and develops them in the RAM 307.

  The specific program 308a functions as a specific process 307a. The collation program 308b functions as a collation process 307b. The calculation program 308c functions as a calculation process 307c. The determination program 308d functions as a determination process 307d.

  The processing of the specific process 307a corresponds to the processing of the specifying units 150 and 250. The processing of the matching process 307b corresponds to the processing of the contour matching units 160 and 260. The processing of the calculation process 307c corresponds to the processing of the calculation units 170 and 270. The processing of the determination process 307d corresponds to the processing of the determination units 180 and 280.

  Note that the specific program 308a, the collation program 308b, the calculation program 308c, and the determination program 308d are not necessarily stored in the hard disk device 308 from the beginning. For example, each program is stored in a “portable physical medium” such as a flexible disk (FD), a CD-ROM, a DVD disk, a magneto-optical disk, and an IC card inserted into the computer 300. Then, the computer 300 may read and execute each of the programs 308a to 308d.

  The following supplementary notes are further disclosed with respect to the embodiments including the above examples.

(Supplementary note 1) The first contour line of the object shown in the image taken at the first timing is specified, and the second object shown in the image taken at the second timing different from the first timing is specified. A specific part for specifying an outline,
The first contour line is photographed at the second timing when the object is a road surface pattern using the moving direction of the vehicle, the moving speed, and the time difference between the first timing and the second timing. A calculation unit that calculates a pattern prediction contour to be performed and a three-dimensional object prediction contour to be photographed at the second timing when the object is a three-dimensional object;
Of the predicted pattern contour and the three-dimensional object predicted contour, if the length of the pattern predicted contour is closer to the length of the second contour than the length of the three-dimensional predicted contour, A determination unit that determines a road surface pattern, and determines the target object as a three-dimensional object when the length of the three-dimensional object predicted contour line is closer to the length of the second contour line than the length of the pattern predicted contour line; An object detection apparatus comprising:

(Supplementary Note 2) The specifying unit further specifies the luminance value of the first contour line and the luminance value of the second contour line, and the calculation unit is configured to determine the luminance value of the first contour line based on the luminance value of the first contour line. The brightness value of the pattern predicted contour and the brightness value of the three-dimensional object predicted contour are further calculated, and the determination unit further uses the brightness value of the pattern predicted contour and the brightness value of the three-dimensional object predicted contour, The object detection apparatus according to appendix 1, wherein it is determined whether the object is a road surface pattern or a three-dimensional object.

(Additional remark 3) The said determination part is calculated from the overlap value of the said 2nd outline and the said pattern prediction outline, and the difference value of the luminance value of the said 2nd outline, and the luminance value of the said pattern prediction outline. 1 score is calculated, and the second rate is calculated based on the overlap ratio between the second contour line and the three-dimensional object predicted contour line, and the difference value between the luminance value of the second contour line and the luminance value of the three-dimensional object predicted contour line. The object detection apparatus according to appendix 2, wherein a score is calculated, and it is determined whether the target object is a road surface pattern or a three-dimensional object based on the first score and the second score. .

(Appendix 4) An object detection method executed by a computer,
Identify the first contour line of the object in the image taken at the first timing,
Identify the second contour line of the object in the image captured at a second timing different from the first timing;
The first contour line is photographed at the second timing when the object is a road surface pattern using the moving direction of the vehicle, the moving speed, and the time difference between the first timing and the second timing. A pattern predicted contour to be performed, and a three-dimensional object predicted contour to be photographed at the second timing when the object is a three-dimensional object;
Of the predicted pattern contour and the three-dimensional object predicted contour, if the length of the pattern predicted contour is closer to the length of the second contour than the length of the three-dimensional predicted contour, Judged as a road pattern,
An object that executes a process of determining a target object as a three-dimensional object when the length of the three-dimensional object predicted contour is closer to the length of the second contour than the length of the pattern predicted contour Detection method.

(Supplementary Note 5) The computer further specifies the luminance value of the first contour line and the luminance value of the second contour line, and the calculation process is performed based on the luminance value of the first contour line. The brightness value of the pattern predicted contour and the brightness value of the three-dimensional object predicted contour are further calculated, and the determination process further uses the brightness value of the pattern predicted contour and the brightness value of the three-dimensional object predicted contour. The object detection method according to appendix 4, wherein it is determined whether the object is a road surface pattern or a solid object.

(Additional remark 6) The said computer is 1st from the difference value of the overlapping rate of said 2nd outline and said pattern prediction outline, and the luminance value of said 2nd outline, and the luminance value of said pattern prediction outline. A score is calculated, and the second score is calculated based on the overlap ratio between the second contour line and the three-dimensional object predicted contour line, and the difference value between the luminance value of the second contour line and the luminance value of the three-dimensional object predicted contour line. The object detection method according to appendix 5, wherein the object is determined to be a road surface pattern or a three-dimensional object based on the first score and the second score.

100, 200 Object detection device 110, 210 Camera 120, 220 Movement amount detection unit 130, 230 Contour DB
140,240 Video input unit 150,250 Identification unit 160,260 Contour line matching unit 170,270 Calculation unit 180,280 Determination unit 190,290 Output unit

Claims (4)

The first contour line of the object captured in the image captured at the first timing is identified, and the second contour line of the object captured in the image captured at the second timing different from the first timing is identified. Specific part to do,
The first contour line is photographed at the second timing when the object is a road surface pattern using the moving direction of the vehicle, the moving speed, and the time difference between the first timing and the second timing. A calculation unit that calculates a pattern prediction contour to be performed and a three-dimensional object prediction contour to be photographed at the second timing when the object is a three-dimensional object;
Of the predicted pattern contour and the three-dimensional object predicted contour, if the length of the pattern predicted contour is closer to the length of the second contour than the length of the three-dimensional predicted contour, A determination unit that determines a road surface pattern, and determines the target object as a three-dimensional object when the length of the three-dimensional object predicted contour line is closer to the length of the second contour line than the length of the pattern predicted contour line; An object detection apparatus comprising:   The specifying unit further specifies a luminance value of the first contour line and a luminance value of the second contour line, and the calculation unit is configured to determine the pattern predicted contour line based on the luminance value of the first contour line. And the determination unit further uses the luminance value of the pattern prediction contour line and the luminance value of the three-dimensional object prediction contour line so that the object is a road surface. The object detection apparatus according to claim 1, wherein it is determined whether the pattern is a solid pattern or a three-dimensional object.   The determination unit calculates a first score from an overlap rate between the second contour line and the pattern predicted contour line, and a difference value between a luminance value of the second contour line and a luminance value of the pattern predicted contour line. Then, the second score is calculated from the overlap rate between the second contour line and the three-dimensional object predicted contour line, and the difference value between the luminance value of the second contour line and the luminance value of the three-dimensional object predicted contour line. 3. The object detection apparatus according to claim 2, wherein it is determined whether the target object is a road surface pattern or a three-dimensional object based on the first score and the second score. An object detection method executed by a computer,
Identify the first contour line of the object in the image taken at the first timing,
Identify the second contour line of the object in the image captured at a second timing different from the first timing;
The first contour line is photographed at the second timing when the object is a road surface pattern using the moving direction of the vehicle, the moving speed, and the time difference between the first timing and the second timing. A pattern predicted contour to be performed, and a three-dimensional object predicted contour to be photographed at the second timing when the object is a three-dimensional object;
Of the predicted pattern contour and the three-dimensional object predicted contour, if the length of the pattern predicted contour is closer to the length of the second contour than the length of the three-dimensional predicted contour, Judged as a road pattern,
An object that executes a process of determining a target object as a three-dimensional object when the length of the three-dimensional object predicted contour is closer to the length of the second contour than the length of the pattern predicted contour Detection method.

Images