Information processing device

Abstract

To accurately detect a mobile body even when the mobile body is formed of a material which hardly reflects a laser beam.SOLUTION: An information processing device performs: time-sequentially acquiring first point group information based on three-dimensional point group information of a first area A1 of a mobile body road RW where a movement direction is fixed, second point group information based on three-dimensional point group information of a second area A2, and third point group information based on three-dimensional point group information of a third area A3 located on a downstream side of the second area A2; calculating a velocity VAM of a mobile body AM based on a temporal change of the first point group information A1; calculating a tip position FE of the mobile body AM at a first point in time T1 from the second point group information A2; calculating a rear end position RE of the mobile body AM at a second point in time T2 from the third point group information A3; and calculating a length LAM of the mobile body based on the velocity VAM of the mobile body, the tip position FE of the mobile body AM at the first point in time T1, and the rear end position RE of the mobile body AM at the second point in time T2.SELECTED DRAWING: Figure 4

Description

[Technical field]

[0001] The present invention relates to an information processing device. [Background technology]

[0002] LiDAR (Light Detection and Ranging, Laser Imaging Detection and Ranging) is used to obtain shape information of an object. This device is also called a laser radar or an optical distance measuring sensor. For example, Patent Document 1 discloses a method for estimating the width and length of a vehicle using point cloud information obtained by LiDAR.

[0003] LiDAR measures the distance to an object by irradiating a laser beam into a detection space and receiving the light reflected by the object. When the spread angle of the irradiated laser beam is large, the measured distance is shortened because the reflected component of the laser beam is reduced. On the other hand, when the spread angle of the irradiated laser beam is small, the measured distance is lengthened because the reflected component of the laser beam is increased (Patent Documents 2 and 3).

[0004] LiDAR also obtains shape information of objects in the detection space by scanning the emitted laser light with a MEMS mirror or the like. In order to expand the scanning angle of the MEMS mirror, it has been disclosed that a lens is combined with the laser light deflected by the MEMS mirror (Patent Document 4). A LiDAR type combined with a wide-angle lens has a wide scanning range, but the measurement distance is short because the spread angle of the laser light is large. On the other hand, a LiDAR type combined with a telephoto lens has a narrow scanning range, but the measurement distance is long because the spread angle of the laser light is small.

[0005] Depending on the measurement environment, multiple types of LiDAR as described above are often combined. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] JP 2017-146153 A [Patent Document 2] JP 2006-317304 A [Patent Document 3] International Publication No. 2012 / 117542 [Patent Document 4] Patent No. 6019866 Summary of the Invention [Problem to be solved by the invention]

[0007] LiDAR measures the distance to an object in each of multiple detection directions by irradiating a laser beam into a detection space and receiving the reflected light that is reflected by the object, thereby acquiring the shape of the object. Therefore, if the object is an object that does not easily reflect laser beam, such as a black vehicle, the accuracy of the distance measured by LiDAR may decrease.

[0008] One example of a problem to be solved by the present invention is to detect a moving object with high accuracy even if the object is made of a material that does not easily reflect laser light. [Means for solving the problem]

[0009] In order to solve the above problem, the information processing device of the invention described in claim 1 has a point cloud acquisition unit that acquires, in a time series, first point cloud information based on three-dimensional point cloud information of a first region of a moving body path whose moving direction is determined, and second point cloud information based on three-dimensional point cloud information of a second region downstream from the first region, a moving body speed calculation unit that calculates the speed of the moving body by estimating movement of the tip position of the moving body based on the change over time of the first point cloud information, and a tip position detection unit that estimates the tip position of the moving body from a point cloud within a range determined by the tip position of the moving body estimated from the speed of the moving body in the second point cloud information, and calculates the tip position of the moving body at a first time based on the change over time of the estimated tip position.

[0010] The information processing method of the invention described in claim 6 includes a point cloud acquisition step of acquiring, in a time series, first point cloud information based on three-dimensional point cloud information of a first region of a moving body path having a determined moving direction, and second point cloud information based on three-dimensional point cloud information of a second region downstream of the first region; a moving body speed calculation step of calculating the speed of the moving body by estimating movement of the tip position of the moving body based on the change over time of the first point cloud information; and a tip position detection step of estimating the tip position of the moving body from a point cloud within a range determined by the tip position of the moving body estimated from the speed of the moving body in the second point cloud information, and calculating the tip position of the moving body at a first time based on the change over time of the estimated tip position.

[0011] An information processing program according to the seventh aspect of the present invention causes a computer to execute the information processing method according to the sixth aspect.

[0012] A storage medium according to an eighth aspect of the present invention stores the information processing program according to the seventh aspect.

[0013] LiDAR also obtains the shape of an object by measuring the distance to the object in each of a number of detection directions by irradiating a laser beam into a detection space and receiving the reflected light that is reflected by the object. Therefore, if the object is an object that does not easily reflect laser beam, such as a black vehicle, the accuracy of the distance measured by LiDAR decreases, and the overall shape of the object may not be obtained. One example of a problem that the other invention of the present application aims to solve is how to measure the length of a moving object with high accuracy even if the object is made of a material that does not easily reflect laser light. An information processing device of another invention of the present application has a point cloud acquisition unit that acquires, in a time series, first point cloud information based on three-dimensional point cloud information of a first region of a moving body path having a determined movement direction, second point cloud information based on three-dimensional point cloud information of a second region of the moving body path, and third point cloud information based on three-dimensional point cloud information of a third region downstream of the second region, a moving body speed calculation unit that calculates the speed of a moving body based on a time change of the first point cloud information, a leading edge position calculation unit that calculates the leading edge position of the moving body at a first time from the second point cloud information, a trailing edge position calculation unit that calculates the trailing edge position of the moving body at a second time from the third point cloud information, and a moving body length calculation unit that calculates a length of the moving body based on the calculated speed of the moving body, the calculated leading edge position of the moving body at the first time, and the calculated trailing edge position of the moving body at the second time.

[0014] Another information processing method of the present invention includes a point cloud acquisition step of acquiring, in time series, first point cloud information based on three-dimensional point cloud information of a first region of a moving body path having a determined moving direction, second point cloud information based on three-dimensional point cloud information of a second region of the moving body path, and third point cloud information based on three-dimensional point cloud information of a third region downstream of the second region; a moving body speed calculation step of calculating the speed of the moving body based on a time change of the first point cloud information; a leading end position calculation step of calculating the leading end position of the moving body at a first time from the second point cloud information; and a rear end position calculation step of calculating the rear end position of the moving body at a second time from the third point cloud information. and a moving body length calculation process for calculating the length of the moving body based on the calculated speed of the moving body, the calculated leading end position of the moving body at the first time, and the calculated trailing end position of the moving body at the second time.

[0015] An information processing program according to another aspect of the present invention causes a computer to execute the information processing method according to the other aspect of the present invention.

[0016] A storage medium according to another aspect of the present invention stores the information processing program according to the other aspect of the present invention. [Brief description of the drawings]

[0017] [Figure 1] FIG. 1 illustrates a LiDAR system 1000 according to an embodiment of the present invention. [Diagram 2] 1 is a diagram showing a relationship between the LiDAR 100 and a first area A1, a second area A2, and a third area A3. FIG. [Diagram 3] 1 is a diagram showing the relationship between the field of view of the LiDAR 100 and a first area A1, a second area A2, and a third area. [Figure 4] FIG. 2 illustrates a control unit 200 according to an embodiment of the present invention. [Diagram 5] 1 is a diagram showing an example of a vehicle AM ​​traveling in a first area A1 and a point cloud corresponding to the vehicle AM ​​measured at this time. [Figure 6] 1 is a diagram showing an example of a vehicle AM ​​traveling in a second area A2 and a point cloud corresponding to the vehicle AM ​​measured at this time. [Figure 7] 13 is a diagram showing an example of a vehicle AM ​​traveling in a third area A3 and a point cloud corresponding to the vehicle AM ​​measured at this time. FIG. [Figure 8] FIG. 11 is a diagram for explaining an example of calculation of a vehicle length LAM. [Figure 9] FIG. 4 is a diagram showing an example of a processing operation in a control unit 120 according to an embodiment of the present invention. [Figure 10] 2 is a diagram illustrating a vehicle speed calculation unit 220 of a control unit 200 according to an embodiment of the present invention. [Figure 11] FIG. 13 is a diagram showing an example in which two vehicles are present in a first area A1. [Figure 12] FIG. 2 is a diagram showing an example of a tip portion region AT. [Figure 13] FIG. 13 is a diagram showing an example of movement of an estimated tip position. [Figure 14] FIG. 4 is a diagram showing an example of a processing operation in a vehicle speed calculation unit 220 according to an embodiment of the present invention. [Figure 15] A figure showing an example of the relationship between a point group estimated by the segmentation unit 221 to correspond to the vehicle AM ​​and the width WAT of the tip portion region AT. [Figure 16] FIG. 13 is a diagram showing an example in which a point cloud corresponding to one vehicle is separated into two segments. [Figure 17] 13 is a diagram showing the relationship between a point group and a continuity determination region AC. FIG. [Figure 18] 11 is a diagram showing an example of the relationship between the movement of the point cloud and a continuity determination region AC. FIG. [Figure 19] FIG. 11 is a diagram showing an example of a processing operation in a continuity determination unit 225 according to an embodiment of the present invention. [Figure 20] FIG. 4 is a diagram showing an example of a processing operation in a vehicle speed calculation unit 220 according to an embodiment of the present invention. [Figure 21] FIG. 1 is a diagram illustrating scanning by LiDAR. [Figure 22] FIG. 13 is a diagram showing an example of a rule associating the position of an object on the roadway RW in the first area A1 with the time difference between the scanning time (time stamp) indicated by the time information and the time when the object is scanned. [Diagram 23] 13 is a diagram showing the relationship between the point group, the predicted tip position yFP, and the tip position detection region VFD. FIG. [Figure 24] 11 is a diagram showing the relationship between a point cloud and a tip position detection field VFD. FIG. [Diagram 25] FIG. 11 is a diagram showing an example of a processing operation in a tip position detection unit 230 according to an embodiment of the present invention. [Figure 26] 13 is a diagram showing the relationship between the point group, the predicted rear end position yRP, and the rear end position detection region VRD. FIG. [Figure 27] 11 is a diagram showing the relationship between a point group and a rear end position detection region VRD. FIG. [Figure 28] 11 is a diagram showing an example of a processing operation in a trailing end position detection unit 240 according to an embodiment of the present invention. FIG. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0018] An information processing device according to one embodiment of the present invention has a point cloud acquisition unit that acquires, in a time series, first point cloud information based on three-dimensional point cloud information of a first region of a moving body path having a determined moving direction, second point cloud information based on three-dimensional point cloud information of a second region of the moving body path, and third point cloud information based on three-dimensional point cloud information of a third region downstream of the second region, a moving body speed calculation unit that calculates the speed of a moving body based on a time change of the first point cloud information, a leading edge position calculation unit that calculates a leading edge position of the moving body at a first time from the second point cloud information, a trailing edge position calculation unit that calculates a trailing edge position of the moving body at a second time from the third point cloud information, and a moving body length calculation unit that calculates a length of the moving body based on the calculated speed of the moving body, the calculated leading edge position of the moving body at the first time, and the calculated trailing edge position of the moving body at the second time. Therefore, in this embodiment, in order to obtain the length of the moving body, it is possible to use two pieces of point cloud information, the second point cloud information showing the front surface of the moving body while moving and the third point cloud information showing the rear surface of the moving body while moving. As a result, in this embodiment, even in the case of a moving body that does not easily reflect laser light, such as a black vehicle, it is possible to use point clouds corresponding to parts that easily reflect laser light, such as a license plate or bumper arranged on the front or rear surface of the moving body, to detect the leading edge position and the rear edge position of the moving body, and it is possible to reliably obtain the leading edge position and the rear edge position of the moving body. Therefore, in this embodiment, even if a material that does not easily reflect laser light is used, it is possible to measure the length of the moving body with high accuracy.

[0019] The LiDAR system according to an embodiment of the present invention includes the information processing device and a LiDAR that generates three-dimensional point cloud information of the first area, the second area, and the third area. Therefore, in this embodiment, it is possible to generate point cloud information for calculating the length of a moving object.

[0020] The LiDAR may include a first LiDAR that generates three-dimensional point cloud information of the first region, a second LiDAR that generates three-dimensional point cloud information of the second region, and a third LiDAR that generates three-dimensional point cloud information of the third region. In this way, it is possible to generate point cloud information for calculating the length of a moving object by combining LiDARs with narrow fields of view.

[0021] The LiDAR includes a first LiDAR that generates three-dimensional point cloud information of the first region and the second region, and a second LiDAR that generates three-dimensional point cloud information of the third region. In this way, it is possible to generate point cloud information for calculating the length of a moving object by combining LiDARs with narrow fields of view while reducing the number of LiDARs.

[0022] Moreover, an information processing method according to one embodiment of the present invention includes a point cloud acquisition process for acquiring, in a time series, first point cloud information based on three-dimensional point cloud information of a first region of a moving body path having a determined moving direction, second point cloud information based on three-dimensional point cloud information of a second region of the moving body path, and third point cloud information based on three-dimensional point cloud information of a third region downstream of the second region; a moving body speed calculation process for calculating the speed of the moving body based on a time change of the first point cloud information; a leading edge position calculation process for calculating the leading edge position of the moving body at a first time from the second point cloud information; a trailing edge position calculation process for calculating the trailing edge position of the moving body at a second time from the third point cloud information; and a moving body length calculation process for calculating a length of the moving body based on the calculated speed of the moving body, the calculated leading edge position of the moving body at the first time, and the calculated trailing edge position of the moving body at the second time. Therefore, in this embodiment, in order to obtain the length of the moving body, it is possible to use two pieces of point cloud information, the second point cloud information showing the front surface of the moving body while moving and the third point cloud information showing the rear surface of the moving body while moving. As a result, in this embodiment, even in the case of a moving body that does not easily reflect laser light, such as a black vehicle, it is possible to use point clouds corresponding to parts that easily reflect laser light, such as a license plate or bumper arranged on the front or rear surface of the moving body, to detect the leading edge position and the rear edge position of the moving body, and it is possible to reliably obtain the leading edge position and the rear edge position of the moving body. Therefore, in this embodiment, even if a material that does not easily reflect laser light is used, it is possible to measure the length of the moving body with high accuracy.

[0023] Moreover, an information processing program according to an embodiment of the present invention causes a computer to execute the above-mentioned information processing method, and thus it is possible to use the computer to measure the length of a moving object having a low surface reflectance with high accuracy.

[0024] Moreover, a computer-readable storage medium according to an embodiment of the present invention stores the above-mentioned information processing program, which makes it possible to distribute the above-mentioned information processing program as a standalone program in addition to being incorporated into a device, and makes it easy to perform version upgrades, etc. EXAMPLES

[0025] <LiDARシステム1000> 1 is a diagram showing a LiDAR system 1000 according to an embodiment of the present invention. The LiDAR system 1000 includes a LiDAR 100 and a control unit 200. In this embodiment, the control unit 200 calculates the length of a vehicle based on three-dimensional point cloud information generated by the LiDAR 100.

[0026] The LiDAR 100 generates point cloud information used to calculate the length of a vehicle (mobile body) AM traveling on the near side lane of the roadway (mobile body path) RW. For this purpose, the LiDAR 100 is placed on the side of the roadway RW, for example, as shown in Fig. 2, so as to be able to measure three areas of the roadway RW, namely, a first area A1, a second area A2, and a third area A3. The LiDAR 100 may be installed above the roadway RW, or may be configured to measure the three areas from above.

[0027] The LiDAR 100 measures three areas, the first area A1, the area A2, and the area A3, of the roadway RW, and generates three-dimensional point cloud information of the first area A1, the second area A2, and the third area A3 in time series. That is, the LiDAR 100 measures three areas, the first area A1, the second area A2, and the third area A3, of the roadway RW at a predetermined frame rate, and generates three-dimensional point cloud information of the first area A1, the second area A2, and the third area A3 at a predetermined frame rate. Therefore, each of the three-dimensional point cloud information of the first area A1, the second area A2, and the third area A3 is composed of a plurality of frames generated at a predetermined frame rate.

[0028] In each frame, each point is expressed by Cartesian coordinates (x, y, z). At this time, as shown in Fig. 2, the position of the LiDAR 100 is the origin O, the width direction of the roadway RW is the x-axis direction, the driving direction of the lane on the near side of the roadway RW is the y-axis direction, and the direction perpendicular to the width direction and driving direction of the roadway RW is the z-axis direction.

[0029] As shown in Fig. 2, the first area A1 is the most upstream of the three areas in the driving direction of the lane on the near side of the roadway RW, the second area A2 is downstream of the first area A1, and the third area A3 is downstream of the second area A2. In the example shown in Fig. 2, the first area A1 and the second area A2 do not overlap, but the first area A1 and the second area A2 may overlap. Also, in the example shown in Fig. 2, the second area A2 and the third area A3 do not overlap, but the second area A2 and the third area A3 may overlap.

[0030] 3 is a diagram showing the relationship between the field of view of the LiDAR 100 and the first area A1, the second area A2, and the third area. The LiDAR 100 may be configured with one LiDAR having a field of view that includes all of the first area A1, the second area A2, and the third area. The LiDAR 100 may also be configured with three LiDARs 130, for example, as shown in FIG. 3(A), including a first LiDAR 110 having a field of view FV1 that includes the first area A1, a second LiDAR 120 having a field of view FV2 that includes the second area A2, and a third LiDAR having a field of view FV3 that includes the third area A3. 3B, the LiDAR 100 may be configured with two LiDARs, a fourth LiDAR 140 having a field of view FV4 including both the first area A1 and the second area A2, and a fifth LiDAR 150 having a field of view FV5 including the third area A3. When the LiDAR 100 is configured with a plurality of LiDARs, the position where any one of the plurality of LiDARs is disposed may be set as the origin O.

[0031] 2, the first area A1 and the second area A2 are located upstream of the LiDAR 100, and the third area A3 is located downstream of the LiDAR 100. For this reason, the LiDAR 100 can measure the front surface (front surface, bonnet, windshield, etc.) and the side surface (the side where the LiDAR 100 is arranged) of the vehicle AM ​​traveling in the first area A1 and the second area A2, and can also measure the rear surface and the side surface (the side where the LiDAR 100 is arranged) of the vehicle AM ​​traveling in the third area A3.

[0032] 4 is a diagram showing a control unit 200 according to an embodiment of the present invention. The control unit 200 has a point cloud information acquisition unit 210, a vehicle speed calculation unit (mobile object speed calculation unit) 220, a front end position detection unit 230, a rear end position detection unit 240, and a vehicle length calculation unit 250. The control unit 200 is configured by, for example, a computer having a CPU (Central Processing Unit) and the like.

[0033] The point cloud information acquisition unit 210 acquires first point cloud information based on three-dimensional point cloud information of the first area A1 generated by LiDAR 100, second point cloud information based on three-dimensional information of the second area A2 generated by LiDAR 100, and third point cloud information based on three-dimensional information of the third area A3 generated by LiDAR 100.

[0034] In this embodiment, the control unit 200 calculates the length of the vehicle. Therefore, it is sufficient that the control unit 200 can acquire information on at least the length direction of the vehicle, that is, the traveling direction (y-axis direction) of the vehicle. Therefore, the first point cloud information may be the three-dimensional point cloud information of the first area A1 generated by the LiDAR itself, or may be one-dimensional point cloud information (y coordinate value of each point) of this three-dimensional point cloud information, or may be two-dimensional point cloud information (x coordinate value, y coordinate value of each point). Similarly, the second point cloud information (third point cloud information) may be the three-dimensional point cloud information of the second area A2 (third area A3) generated by the LiDAR itself, or may be one-dimensional point cloud information (y coordinate value of each point) of this three-dimensional point cloud information, or may be two-dimensional point cloud information (x coordinate value, y coordinate value of each point).

[0035] In this embodiment, the control unit 200 calculates the length LAM of the vehicle AM ​​traveling on the roadway RW based on the three pieces of point cloud information acquired by the point cloud information acquisition unit 210. Specifically, the control unit 200 calculates the speed VAM of the vehicle AM ​​based on information (first point cloud information) when the vehicle AM ​​is traveling in the first area A1, calculates the front end position FE of the vehicle AM ​​at the first time T1 based on information (second point cloud information) when the vehicle AM ​​is traveling in the second area A2, calculates the rear end position RE of the vehicle AM ​​at the second time T2 based on information (third point cloud information) when the vehicle AM ​​is traveling in the third area A3, and calculates the length LAM of the vehicle AM ​​from the calculated speed VAM of the vehicle AM, the front end position FE of the vehicle AM ​​at the first time T1, and the rear end position RE of the vehicle AM ​​at the second time T2.

[0036] The vehicle speed calculation unit 220 calculates the speed VAM of the vehicle AM ​​based on the time change of the first point cloud information. FIG. 5(A) is a diagram showing the vehicle AM ​​traveling in the first area A1, and FIG. 5(B) is a diagram showing the point cloud corresponding to the vehicle AM ​​measured at this time. The first area A1 is located upstream of the LiDAR 100 as shown in FIGS. 2 and 5. Therefore, as shown in FIG. 5(A), the front surface (front surface, bonnet, windshield, etc.) and the side surface (the side where the LiDAR 100 is arranged) of the vehicle AM ​​traveling in the first area A1 are visible from the LiDAR 100, and the first point cloud information includes, for example, point cloud data corresponding to the front surface and the side surface of the vehicle AM ​​as shown in FIG. 5(B). Therefore, the vehicle speed calculation unit 220 calculates the speed VAM of the vehicle AM ​​by detecting the movement of the tip position FE of the vehicle AM ​​using this point cloud, for example. In addition, since the first area A1 is wide in the direction in which the vehicle moves, point cloud information is acquired for a relatively long period of time. For this reason, the vehicle speed calculation unit 220 may improve the calculation accuracy of the speed VAM by using a Kalman filter or the like for the time series data of multiple frames.

[0037] The tip position detection unit 230 calculates the tip position FE of the vehicle AM ​​at the first time T1 based on the second point cloud information. FIG. 6(A) is a diagram showing the vehicle AM ​​traveling in the second area A2, and FIG. 6(B) is a diagram showing the point cloud corresponding to the vehicle AM ​​measured at this time. The second area A2 is located upstream of the LiDAR 100 as shown in FIG. 2. Therefore, as shown in FIG. 6(A), the front surface (front surface, bonnet, windshield, etc.) and the side surface (the side where the LiDAR 100 is arranged) of the vehicle AM ​​traveling in the second area A2 are visible from the LiDAR 100, and as shown in FIG. 6(B), for example, point cloud data corresponding to the front surface and the side surface of the vehicle AM ​​is acquired. Therefore, the tip position detection unit 230 obtains the tip position FE of the vehicle AM ​​at the first time T1 by detecting a point corresponding to the tip position FE of the vehicle AM ​​in the frame at the first time T1 from the second point cloud information, for example. In addition, since the second area A2 is narrow in the direction in which the vehicle moves, point cloud information is acquired for a relatively short period. For this reason, the tip position detection unit 230 may improve the accuracy of the tip position FE by calculating the average value of the time information (T11, T12, T13) and tip positions (FE1, FE2, FE3) of the time series data of multiple frames (e.g., three frames).

[0038] The rear end position detection unit 240 calculates the rear end position RE of the vehicle at the second time based on the third point cloud information. FIG. 7(A) is a diagram showing the vehicle AM ​​traveling in the third area A3, and FIG. 7(B) is a diagram showing the point cloud corresponding to the vehicle AM ​​measured at this time. The third area A3 is located downstream of the LiDAR 100 as shown in FIG. 2. Therefore, as shown in FIG. 7(A), the rear surface and the side surface of the vehicle AM ​​traveling in the third area A3 are visible from the LiDAR 100, and the third point cloud information includes point cloud data corresponding to the rear surface and the side surface of the vehicle AM, as shown in FIG. 7(B), for example. Therefore, the rear end position detection unit 240 obtains the front end position RE of the vehicle AM ​​at the second time T2 by detecting a point corresponding to the rear end position RE of the vehicle AM ​​in the frame at the second time T2 of the third point cloud information, for example. In addition, since the third area A3 is narrow in the direction in which the vehicle moves, point cloud information is acquired for a relatively short period of time. For this reason, the rear end position detection unit 240 may improve the accuracy of the rear end position RE by calculating the average value of the time information (T21, T22, T23) and rear end positions (RE1, RE2, RE3) of the time series data of multiple frames (e.g., three frames).

[0039] The vehicle length calculation unit 250 calculates the length LAM of the vehicle AM ​​based on the speed VAM of the vehicle AM, the leading edge position FE of the vehicle AM ​​at the first time T1, and the rear edge position RE of the vehicle AM ​​at the second time T2. For example, when the y coordinate of the leading edge position FE at the first time T1 calculated by the leading edge position detection unit 230 is yFET1 (FIG. 8(A)), if the vehicle AM ​​is traveling at the speed VAM of the vehicle AM ​​calculated by the vehicle speed calculation unit 220 between the first time T1 and the second time T2, it is estimated that the y coordinate of the leading edge position FE of the vehicle AM ​​at the second time T2 is yFET2=yFET1+VAM(T2-T1) (FIG. 8(B)). Therefore, when the y coordinate of the rear end position RE at the second time T2 calculated by the rear end position detection unit 240 is yRET2, the length LAM of the vehicle AM ​​can be calculated, for example, as LAM = yFET2 - yRET2 = yFET1 - yRET2 + VAM (T2 - T1).

[0040] In this way, in this embodiment, two pieces of point cloud information, the second point cloud information including the front surface of the moving vehicle and the third point cloud information including the rear surface of the moving vehicle, are used to obtain the length of the vehicle. As a result, in this embodiment, even in the case of a black vehicle that does not easily reflect laser light, it is possible to use point clouds corresponding to parts that easily reflect laser light, such as license plates and bumpers arranged on the front and rear surfaces of the vehicle, and it is possible to reliably obtain the leading and trailing end positions of the vehicle. Therefore, in this embodiment, it is possible to measure the length of a moving vehicle that does not easily reflect laser light, such as a black vehicle, with high accuracy.

[0041] 9 is a diagram showing an example of a processing operation in the control unit 120 according to an embodiment of the present invention. The point cloud information acquisition unit 210 acquires first point cloud information based on three-dimensional information of the first area A1 of the roadway RW, second point cloud information based on three-dimensional information of the second area A2, and third point cloud information based on three-dimensional information of the third area A3 (step S901). The speed calculation unit 210 calculates the speed VAM of the vehicle AM ​​based on the time change of the first point cloud information (step S902). The front end position detection unit 230 calculates the front end position FE of the vehicle AM ​​at the first time T1 based on the second point cloud information (step S903). The rear end position detection unit 240 calculates the rear end position RE of the vehicle AM ​​at the second time T2 based on the third point cloud information (step S904). The vehicle length calculation unit 250 calculates the length LAM of the vehicle AM ​​based on the speed VAM of the vehicle AM, the leading edge position FE of the vehicle AM ​​at the first time T1, and the trailing edge position RE of the vehicle AM ​​at the second time T2 (step S905).

[0042] <Calculation of vehicle speed (VAM)> 10 is a diagram showing the vehicle speed calculation unit 220 of the control unit 200 according to an embodiment of the present invention. The vehicle speed calculation unit 220 has a segmentation unit 221, a representative point setting unit 222, a tip position movement estimation unit 223, and a speed calculation unit 224.

[0043] The segmentation unit 221 estimates a point cloud corresponding to the same vehicle based on the distance between two points in each frame of the first point cloud information.

[0044] Each frame of the first point cloud information may include multiple vehicles. Fig. 11 is a diagram showing an example in which there are two vehicles in the first area A1. When two vehicles (vehicle AM1 and vehicle AM2) are traveling in the first area A1 as shown in Fig. 11(A), the first point cloud information includes two point cloud clusters as shown in Fig. 11(B).

[0045] Therefore, the segmentation unit 221 divides the points of each frame of the first point cloud information into point clouds (segments) each of which is estimated to correspond to the same vehicle. In the example shown in Fig. 11(B), the segmentation unit 221 divides the points of the frame into two segments: a segment S1 corresponding to a vehicle AM1 and a segment S2 corresponding to a vehicle AM2.

[0046] At this time, the segmentation unit 221 divides the points of the frame into a plurality of segments based on the distance between the two points. Specifically, the segmentation unit 221 divides the points of the frame so that two points that are less than a first distance (for example, 1 m or 1.5 m) in each frame are included in the same segment. The distance between two points (P1(x1, y1, z1), P2(x2, y2, z2)) is calculated as the three-dimensional distance (((x1-x2) 2 +(y1-y2) 2 +(z1-z2) 2 ) 1 / 2 ) or the two-dimensional distance (((x1-x2) 2 +(y1-y2) 2 ) 1 / 2 ) or one-dimensional distance (|y1-y2|).

[0047] The representative point setting unit 222 extracts a point cloud of the tip portion region from each of the point clouds (segments) estimated to correspond to the same vehicle in each frame of the first point cloud information, and sets a representative point based on the extracted point clouds.

[0048] It is highly likely that the tip end point (the point with the largest y coordinate value) of each segment obtained by dividing the points by the segmentation unit 221 indicates the tip end position of the vehicle corresponding to that segment. Therefore, for example, it is possible to detect this tip end point of the segment in each frame and obtain the speed of the vehicle corresponding to this segment based on the time change of this detected tip end point, that is, the movement of the tip end point.

[0049] However, there is a possibility that the distance measurement accuracy of this leading edge point is poor. Therefore, in this embodiment, the representative point setting unit 222 sets a representative point for each segment that represents that segment, and calculates the vehicle speed corresponding to each segment based on the movement of this representative point instead of the movement of the leading edge point.

[0050] At this time, in this embodiment, the representative point setting unit 222 sets, for each segment, an area (tip portion area AT) that spreads backward from the tip point PT, which is the tipmost point, as shown in Fig. 12. That is, when the y coordinate of the tip point is yPT, the representative point setting unit 222 sets an area determined by yPT-WAT≦y≦yPT as the tip portion area AT. At this time, the tip portion area AT has a width WAT in the y-axis direction. The width WAT may be, for example, 10 cm to 1 m.

[0051] Then, the representative point setting unit 222 averages the y coordinates of the points in the tip partial region AT and sets this average value as the representative point. That is, in this embodiment, the representative point is set only by the y coordinate values ​​of the points in the tip partial region AT, and the representative point has only a y coordinate value. For example, when the tip partial region AT includes n points: P1 (x1, y1, z1), P2 (x2, y2, z2), P3 (x3, y3, z3), ..., Pn (xn, yn, zn), the y coordinate of the representative point is (y1 + y2 + y3 + ... + yn) / n.

[0052] The tip position movement estimating unit 223 calculates an estimated tip position, which is an estimate of the tip position of the vehicle, based on the representative point set by the representative point setting unit 222. At this time, the y coordinate value of the representative point may simply be set as the estimated tip position, which is an estimate of the tip position of the vehicle. The tip position movement estimating unit 223 may also calculate the movement of the estimated tip position, which is an estimate of the tip position of the vehicle, by applying a Kalman filter to the y coordinate value of the representative point. That is, the tip position movement estimating unit 223 may estimate the tip position of the vehicle (that is, calculate the estimated tip position) by using a Kalman filter, taking the y coordinate value of the representative point as the observed value of the tip position of the vehicle in each frame. At this time, in this embodiment, the representative point has only a y coordinate value, so the estimated tip position is also a position determined only by the y coordinate value.

[0053] As described above, in this embodiment, in each frame, the segmentation unit 221 divides the points of the frame into a group of points (segments) corresponding to the same vehicle, and for each segment, the representative point setting unit 222 sets a representative point, and for each segment, the tip position movement estimation unit 223 calculates an estimated tip position based on the representative point set. The estimated tip position (i.e., y coordinate value) calculated in this manner moves as the frame progresses, that is, as time passes, as shown in FIG. 13. FIG. 13 shows an example of a change in the estimated tip position corresponding to the vehicle when the tip of the vehicle enters the first area A1 in frame F1 and leaves the first area A1 after frame F6. In this embodiment, the movement of the vehicle corresponding to this estimated tip position is observed by observing how this estimated tip position moves.

[0054] When the estimated tip position reaches a speed calculation area AV set in the first area A1, the speed calculation unit 224 calculates the vehicle speed corresponding to this estimated tip position based on the movement of the estimated tip position. The speed calculation area AV is set as an area determined by yAVL≦y≦yAVH, for example. In other words, when the value of the estimated tip position exceeds yAVL, the speed calculation unit 224 calculates the vehicle speed corresponding to this estimated tip position based on the movement of the estimated tip position.

[0055] The vehicle speed may be calculated based on the movement of the estimated tip position between two frames, or the movement between multiple frames. In the example shown in FIG. 13, the estimated tip position reaches the speed calculation area AV in frame F7, so the vehicle speed is calculated at the stage when the estimated tip position for frame F7 is calculated. In the example shown in FIG. 13, the vehicle speed may be calculated based on the movement of the estimated tip position between the frame (frame F7) when the speed calculation area AV is reached and the frame immediately before that (frame F6), or the vehicle speed may be calculated based on the movement of the estimated tip position between the frames (frames F1-F7) from when the tip of the vehicle enters vehicle A1 until it leaves vehicle A1. Alternatively, since the aforementioned Kalman filter process calculates the moving speed at the same time as estimating the tip position, the speed value calculated by the Kalman filter may be used.

[0056] FIG. 14 is a diagram showing an example of a processing operation in the vehicle speed calculation unit 220 according to this embodiment. In FIG. 14, it is assumed that the first point cloud information is composed of M1 frames Fn (n=1, 2, 3, . . . , M1). The vehicle speed calculation unit 220 sets n=1 (step S1401). The segmentation unit 221 divides the points of the frame Fn into point clouds (segments) estimated to correspond to the same vehicle (step S1402). The representative point setting unit 222 sets a representative point for each of the point clouds (segments) estimated to correspond to the same vehicle (step S1403). The tip position movement estimation unit 223 calculates an estimated tip position, which is an estimated value of the tip position of the vehicle corresponding to the segment, based on the set representative point for each of the point clouds (segments) estimated to correspond to the same vehicle (step S1404).

[0057] If there is a segment in frame Fn whose estimated end position reaches the speed calculation region AV (step S1405, YES), the speed calculation unit 224 calculates the vehicle speed corresponding to the segment whose estimated end position reaches the speed calculation region AV (step S1406) and checks whether n=M1 (step S1407). If n=M1 (step S507, YES), the vehicle speed calculation unit 220 ends the process. If n=M1 is not true (step S1407, NO), the vehicle speed calculation unit 220 sets n=n+1 (step S1408) and returns to step S1402.

[0058] If there is no segment in which the estimated front end position reaches the speed calculation area AV in the frame Fn (step S1405, NO), the vehicle speed calculation unit 220 checks whether n=M1 or not (step S1407).

[0059] <Size of tip area> In the above embodiment, the representative point setting unit 222 sets the representative point based on the point cloud of the tip partial region AT. As shown in FIG. 15, the point cloud (FIG. 15(A)) estimated by the segmentation unit 221 to correspond to the vehicle AM ​​when the vehicle AM ​​is far from the LiDAR 100 has a smaller number of points than the point cloud (FIG. 15(B)) estimated by the segmentation unit 221 to correspond to the vehicle AM ​​when the vehicle AM ​​is near the LiDAR 100. For this reason, if a small width WTA1 is used as the y-direction width WAT of the tip partial region AT when the vehicle AM ​​is far from the LiDAR 100, the number of points included in the tip partial region AT will be reduced, and the distance measurement accuracy tends to be worse for distant points than for nearby points, so that the influence of noise may be large. Therefore, the y-direction width WAT of the tip partial region AT when the vehicle AM ​​is far from the LiDAR 100 needs to be somewhat large, as in the width WTA2 of FIG. 15(A). However, even when the vehicle AM ​​is in the vicinity of the LiDAR 100, if the width WAT2 used when the vehicle AM ​​is far from the LiDAR 100 is used as the width WAT in the y direction of the tip portion area AT, the tip portion area AT includes not only points on the front side of the vehicle AM ​​but also points on the side of the vehicle AM. In such a case, the representative point, which is the average of the y coordinate values ​​of the points included in the tip portion area AT, is smaller than the y coordinate value of the actual tip position. For this reason, the estimated tip position estimated based on this representative value ends up being located further back than the actual tip position. As a result, when the tip position is estimated using this representative point and the speed VAM of the vehicle AM ​​is calculated, the calculated speed VAM ends up being smaller than the actual speed of the vehicle AM.

[0060] Therefore, the representative point setting unit 222 according to an embodiment of the present invention sets the width WAT in the y direction of the tip partial region AT to be smaller as the vehicle AM ​​approaches the LiDAR 100. At this time, the representative point setting unit 222 sets the width WAT in the y direction of the tip partial region AT based on, for example, the distance between the LiDAR 100 and the vehicle AM. That is, the representative point setting unit 222 sets the width WAT in the y direction of the tip partial region AT to be smaller as, for example, the distance between the LiDAR 100 and the vehicle AM ​​becomes smaller.

[0061] The representative point setting unit 222 may, for example, calculate the distance between the LiDAR 100 and the vehicle AM ​​in each frame Fn by using an estimated tip position for a segment corresponding to the vehicle AM ​​in the previous frame Fn-1. That is, the representative point setting unit 222 may, for example, set the width WAT in the y direction of the tip partial region AT in each frame Fn to be smaller as the distance DAM between the estimated tip position in the previous frame Fn-1 and the origin O becomes smaller.

[0062] At this time, the width WAT may change continuously or discontinuously with respect to the distance DAM. For example, a plurality of thresholds may be prepared for the distance DAM, and the width WAT may change each time the distance DAM falls below the threshold. In particular, two thresholds (a first threshold (e.g., 25 m) and a second threshold (e.g., 17 m)) may be prepared, and when the distance DAM is equal to or greater than the first threshold, the width WAT may be set to the first width (e.g., 60 cm), when the distance DAM is smaller than the first threshold and equal to or greater than the second threshold, the width WAT may be set to the second width (e.g., 30 cm), and when the distance DAM is smaller than the second threshold, the width WAT may be set to the third width (e.g., 15 cm).

[0063] <Point Cloud Continuity> The segmentation unit 221 divides the point cloud of each frame so that two points that are equal to or smaller than the first distance are included in the same segment. At this time, the first distance needs to be set to a small value so that points corresponding to different vehicles do not belong to the same segment. However, if the first distance is too small, points that actually correspond to the same vehicle may be separated into two different segments. For example, in the case of a semi-trailer TR in which a tractor TR1 pulls a trailer TR2 as shown in FIG. 16(A), the point clouds of the front and side of each of the tractor TR1 and the trailer TR2 are captured as shown in FIG. 16(B), so that one semi-tractor may be separated into two segments (a segment corresponding to the tractor TR1 and a segment corresponding to the trailer TR2).

[0064] Therefore, the vehicle speed calculation unit 220 according to an embodiment of the present invention further includes a continuity determination unit 225. The continuity determination unit 225 determines the continuity of the points of the first point cloud information based on a temporal change of the point cloud included in a continuity determination area AC set in the first area A1. The continuity determination area AC is set as an area determined by yACL≦y≦yACH, for example.

[0065] For example, the continuity determination unit 225 may store a continuity state indicating the continuity of the points of the first point cloud information in the continuity determination area AC. For example, as shown in FIG. 17(A), when there is no point in the continuity determination area AC, the continuity determination unit 225 stores the continuity state of the frame as "OFF", and as shown in FIG. 17(B), when there is a point in the continuity determination area AC, the continuity state of the frame is stored as "ON". That is, as shown in FIG. 18, when three points (PC1, PC2, PC3) are moving, the continuity states of the frames shown in FIG. 18(A) and 18(G) are stored as "OFF", and the continuity states of the frames shown in FIG. 18(B) to 18(F) are stored as "ON".

[0066] Therefore, the continuity state of the frame (FIG. 18(A)) one frame before the frame (FIG. 18(B)) in which the tip point PC1 of the three points reaches the continuity judgment region AC is "OFF". On the other hand, the continuity state of the frame (FIG. 18(B)) one frame before the frame (FIG. 18(C)) in which the non-tip point PC2 of the three points reaches the continuity judgment region AC is "ON". Therefore, by checking the continuity state of the frame one frame before the frame in which each point reaches the continuity judgment region AC, it is possible to check whether the point is the tip point of the continuous points. Therefore, the continuity determination unit 225 can determine whether each point of the first point cloud information is continuous with other points.

[0067] In this embodiment, when the continuity determination unit 225 determines that the points of two point clouds (segments) estimated by the segmentation unit 221 to correspond to different moving bodies are continuous, the tip position estimation unit 223 determines that the two point clouds (segments) are point clouds that correspond to the same vehicle. For example, when the estimated tip position reaches the continuity determination area AC, the tip position estimation unit 223 checks the continuity state of the frame immediately preceding the frame at which the estimated tip position reached the continuity determination area AC, and if the continuity state is "ON", determines that the segment corresponding to the estimated tip position corresponds to the same vehicle as the segment ahead of the segment.

[0068] Then, the speed calculation unit 224 calculates only one speed for two segments determined by the tip position estimation unit 223 to correspond to the same vehicle. In other words, when the estimated tip position reaches the speed calculation area AV, if the segment corresponding to this estimated tip position is the frontmost segment of the multiple segments determined by the tip position estimation unit 223 to correspond to the same vehicle, the speed calculation unit 224 calculates the vehicle speed. On the other hand, when the estimated tip position reaches the speed calculation area AV, if the segment corresponding to this estimated tip position is not the frontmost segment of the multiple segments determined by the tip position estimation unit 223 to correspond to the same vehicle, the speed calculation unit 224 does not calculate the vehicle speed for that segment. In this way, even if one vehicle is separated into two segments by the segmentation unit 221, it is possible to prevent the vehicle speed from being calculated for each segment. Alternatively, since the above-mentioned Kalman filter process calculates the moving speed at the same time as estimating the tip position, the speed calculation unit 224 may select the moving speed calculated by the Kalman filter.

[0069] At this time, it is advisable to set the speed calculation region AV and the continuity region AC so that the lower limit yAVL of the speed calculation region AV and the lower limit yACL of the continuity judgment region AC are the same. By doing so, when the estimated end position becomes equal to or greater than yAVL, the estimated end position also becomes equal to or greater than yACL, and it becomes possible to synchronize the timing for judging the continuity of the estimated end position and the timing for calculating the vehicle speed corresponding to this estimated end position.

[0070] The width yACH-yACL of the continuity determination area AC in the y-axis direction is set to be shorter than the vehicle distance taken by a vehicle traveling on the roadway RW. In this way, two vehicles will not enter the continuity determination area AC. For this reason, in this embodiment, all points that are included in the continuity determination area AC correspond to one vehicle. As a result, in this embodiment, the continuity determination unit 225 checks the temporal change in the point group included in the continuity determination area AC, and as long as the state in which points are included in the continuity determination area AC continues, it is possible to determine that one vehicle is passing through the continuity determination area AC, and that the points that entered the continuity determination area AC during this period are one point. A traveling vehicle generally has a vehicle distance of 10 m or more. Therefore, it is preferable to set the width yACH-yACL of the continuity determination area AC in the y-axis direction to a value smaller than 10 m, for example, 3 m.

[0071] The number of points corresponding to the vehicle AM ​​measured when the vehicle AM ​​is near the LiDAR 100 is greater than the number of points corresponding to the vehicle AM ​​measured when the vehicle AM ​​is far from the LiDAR 100. Therefore, it is preferable to position the continuity determination area AC near the LiDAR 100. In this way, the distance between points corresponding to the same vehicle AM ​​is reduced, making it possible to more accurately determine the continuity of the point cloud.

[0072] FIG. 19 is a diagram showing an example of the processing operation in the continuity determination unit 225 according to this embodiment. In FIG. 19, it is assumed that the first point cloud information is composed of M1 frames Fn (n=1, 2, 3, . . . , M1). The continuity determination unit 225 sets n=1 (step S1901). The continuity determination unit 225 checks whether or not a point exists in the continuity determination area AC (step S1902). If a point exists in the continuity determination area AC (step S1902, YES), the continuity state of the frame Fn is stored as "ON" (step S1903), and it is checked whether or not n=M1 (step S1904). If n=M1 (step S1904, YES), the continuity determination unit 225 ends the processing. If n=M1 is not satisfied (step S1904, NO), the vehicle speed calculation unit 220 sets n=n+1 (step S1905), and the process returns to step S1902.

[0073] If no point exists in the continuity determination area AC (step S1902, NO), the continuity state of frame Fn is stored as "OFF" (step S1906), and it is confirmed whether n=M1 (step S1904).

[0074] FIG. 20 is a diagram showing an example of a processing operation in the vehicle speed calculation unit 220 according to this embodiment. In FIG. 20, it is assumed that the first point cloud information is composed of M1 frames Fn (n=1, 2, 3, . . . , M1). The vehicle speed calculation unit 220 sets n=1 (step S2001). The segmentation unit 221 divides the points of the frame Fn into segments of a point cloud estimated to correspond to the same vehicle (step S2002). The representative point setting unit 222 sets a representative point for each segment of the point cloud estimated to correspond to the same vehicle (step S2003). The tip position movement estimation unit 223 calculates an estimated tip position, which is an estimated value of the tip position of the vehicle corresponding to the segment, based on the set representative point for each segment of the point cloud estimated to correspond to the same vehicle (step S2004).

[0075] If there is a segment in frame Fn whose estimated tip position has reached the speed calculation area AV (step S2005, YES), the tip position movement estimation unit 223 checks whether the continuity determination unit 225 has determined that the point of the segment corresponding to the estimated tip position is continuous with the point of the segment ahead of this segment (step S2006). If it is determined that they are not continuous (step S2006, NO), the speed calculation unit 224 calculates the speed of the vehicle corresponding to this segment (step S2009). If it is determined that they are continuous (step S2006, YES), the tip position movement estimation unit 223 determines that this segment corresponds to the same vehicle as the vehicle corresponding to the preceding segment, and does not perform vehicle speed calculation by the speed calculation unit 224.

[0076] If there is no segment in which the estimated front end position reaches the speed calculation area AV in the frame Fn (step S2005, NO), the vehicle speed calculation unit 220 checks whether n=M1 or not (step S2007).

[0077] The vehicle speed calculation unit 220 checks whether n=M1 (step S2007). If n=M1 (step S2007, YES), the vehicle speed calculation unit 220 ends the process. If n=M1 is not true (step S2007, NO), the vehicle speed calculation unit 220 sets n=n+1 (step S2008) and returns to step S2002.

[0078] <Time correction for representative points> As shown in Fig. 21, in the case of a LiDAR that generates one frame of data by scanning from top to bottom (or bottom to top), there is a difference between the time when each scanning layer is actually scanned and the time of the frame timestamp. In the example shown in Fig. 21, scanning from left to right is performed 10 times to scan the entire range AS. For example, if the time when scanning starts is used as the timestamp of each frame, and the time it takes to scan the entire range AS is TS, the difference between the frame timestamp and the time when the nth layer is scanned in each frame is (n-1)TS / 10.

[0079] Therefore, when calculating the speed of the vehicle based on data generated by such LiDAR, it is possible to calculate a more accurate speed by taking this time difference into consideration. For example, when the time when scanning is started is used as the timestamp of each frame, the time when the nth layer is scanned is later than the timestamp of the frame. As a result, if the above-mentioned time difference is not taken into consideration, the calculated vehicle speed will be greater than the actual vehicle speed. However, in this embodiment, the representative point used when calculating the vehicle speed is the average value of the y coordinates of the points in the tip portion area AT. For this reason, there is no information on which scanning layer the representative point is, and it is not possible to perform time correction taking into account the above-mentioned time difference.

[0080] Therefore, the vehicle speed calculation unit 220 according to an embodiment of the present invention further includes a representative point scanning time estimation unit 226. The representative point scanning time estimation unit 226 corrects the scanning time (timestamp) indicated by the time information by using a rule that associates the position of an object on the roadway RW with the time difference between the scanning time (timestamp) indicated by the time information and the time when the object is scanned, and the position of the representative point, to estimate the time when the representative point is scanned. In this embodiment, the leading edge position movement estimation unit 223 calculates an estimated leading edge position, which is an estimated value of the leading edge position of the vehicle, by using the time when the representative point is estimated to be scanned by the representative point scanning time estimation unit 226. The speed calculation unit 224 also calculates the speed of the vehicle by using the time when the representative point is estimated to be scanned by the representative point scanning time estimation unit 226.

[0081] FIG. 22 is a diagram showing an example of a rule that associates the position of an object on the roadway RW in the first area A1 with the time difference between the scanning time (timestamp) indicated by the time information and the time when the object is scanned. In this diagram, the horizontal axis is the y coordinate, and the vertical axis is the time difference between the timestamp of the frame and the time when the object is scanned. That is, in FIG. 22, the time difference between the time when an object at y=ys1 is actually scanned and the frame is ts1. This rule is obtained in advance by running a vehicle on the roadway RW and measuring the vehicle with the LiDAR 100, or by placing multiple objects on the center line of the lane on the near side of the roadway RW and measuring the multiple objects with the LiDAR 100.

[0082] The representative point scanning time estimation unit 226 estimates the time when the representative point was scanned in each frame from the value of the representative point set by the representative point setting unit 222 and the above rule. For example, when the scanning start time is used as the timestamp of each frame, in the example shown in Fig. 22, if the representative point set by the representative point setting unit 222 in frame Fn whose timestamp is TFn is ys1, the time when the representative point was scanned is estimated to be TFn+ts1.

[0083] <Calculation of tip position> The leading edge position detection unit 230 calculates the leading edge position FE of the vehicle AM ​​at the first time T1 based on the second point cloud information for the vehicle AM ​​whose vehicle speed VAM has been calculated by the vehicle speed calculation unit 220. The vehicle AM, which had the vehicle speed VAM in the first area A1, is considered to reach the second area A2 downstream of the first area A1 at approximately the same vehicle speed as this vehicle speed VAM. Therefore, the leading edge position detection unit 230 detects the leading edge position FE of the vehicle in an area (leading edge position detection area VFD) of the second area A2 where the leading edge position FE of the vehicle AM ​​is predicted to exist based on this vehicle speed VAM.

[0084] First, the tip position detection unit 230 calculates the y coordinate value (predicted tip position yFP) of the position where the tip position FE of the vehicle AM ​​is predicted to exist for the vehicle AM ​​whose vehicle speed VAM has been calculated by the vehicle speed calculation unit 220. As described above, the speed calculation unit 224 of the vehicle speed calculation unit 220 calculates the speed VAM of the vehicle AM ​​when this estimated tip position reaches the speed calculation area AV. Then, the tip position detection unit 230 acquires, for example, the time tVAM when this estimated tip position reaches the speed calculation area AV, the value yVAM of the estimated tip position, and the vehicle speed VAM. In this way, the tip position detection unit 230 can obtain, for example, the predicted tip position yFP at time t by yFP=yVAM+VAM(t-tVAM) (FIG. 23).

[0085] Then, the leading edge position detection unit 230 sets an area including this predicted leading edge position yFP as the leading edge position detection area VFD. The leading edge position detection area VFD is set as an area determined by, for example, yVFDL≦y≦yVFDH. Since the leading edge position FE of the vehicle AM ​​may be located either ahead of or behind this predicted leading edge position yFP, the lower limit yVFDL and upper limit yVFDH of the y coordinate value of the leading edge position detection area VFD are set to sandwich the predicted leading edge position yFP therebetween (yVFDL <yFP<yVFDH)。

[0086] Then, when the upper limit yVFDH of the tip position detection area VFD is in the second area A2 (FIG. 24), the tip position detection unit 230 detects the tipmost point PF (the point with the largest y coordinate) from the point group in this tip position detection area VFD. At this time, the tip position detection unit 230 may detect the tip point PF in all frames when the upper limit yVFDH of the tip position detection area VFD is in the second area A2, or may detect the tip point PF in some of the frames when the upper limit yVFDH of the tip position detection area VFD is in the second area A2.

[0087] When the number of frames used to detect the leading edge point PF is M2, the leading edge position detection unit 230 detects M2 leading edge points PFn (1,...,M2). The leading edge position detection unit 230 calculates the average (yPFn+...+yPFM2) / M2 of the y coordinate values ​​yPFn of the M2 leading edge points PFn as the leading edge position FE. The leading edge position detection unit 230 also calculates the average (TF1+...+TFM2) / M2 of the timestamps TFn of the M2 frames Fn in which the leading edge point PFn was detected as the first time T1.

[0088] FIG. 25 is a diagram showing an example of the processing operation in the leading edge position detection unit 230 according to this embodiment. The leading edge position detection unit 230 calculates a predicted leading edge position yFP for the vehicle AM ​​for which the vehicle speed VAM is calculated by the vehicle speed calculation unit 220 (step S2501). The leading edge position detection unit 230 sets an area including this predicted leading edge position yFP as a leading edge position detection area VFD (step S2502). The leading edge position detection unit 230 detects the leading edge point PFn from the point cloud within the leading edge position detection area VFD in a frame Fn during which the upper limit yVFDH of the leading edge position detection area VFD is in the second area A2 (step S2503). The leading edge position detection unit 230 calculates the leading edge position FE of the vehicle AM ​​at the first time T1 based on the detected leading edge point PFn and the timestamp TFn of the frame Fn in which the leading edge point PFn is detected (step S2504).

[0089] <Calculation of rear end position> The rear-end position detection unit 240 calculates the rear-end position RE at the second time T2 based on the third point cloud information for the vehicle AM for which the vehicle speed VAM has been calculated by the vehicle speed calculation unit 220. The vehicle AM that had the vehicle speed VAM in the first region A1 is considered to reach the third region A3 downstream of the first region A1 at approximately the same vehicle speed as this vehicle speed VAM. Therefore, the rear-end position detection unit 240 detects the rear-end position RE of this vehicle in a region (rear-end position detection region VRD) where the rear-end position RE of the vehicle AM is predicted to exist based on this vehicle speed VAM.

[0090] Similar to the front-end position detection unit 230, the rear-end position detection unit 240 also calculates the y-coordinate value (predicted front-end position yPF) of the position where the front-end position FE of the vehicle AM for which the vehicle speed VAM has been calculated by the vehicle speed calculation unit 220 is predicted to exist. Then, the rear-end position detection unit 240 sets the y-coordinate value of the position that is LE (estimated moving body length) behind the position where the front-end position FE of this vehicle AM is predicted to exist as the position (predicted rear-end position yPR) where the rear-end position RE is predicted to exist. That is, the predicted rear-end position yPR is yPR = yPF - LE. Here, the estimated vehicle LE is an estimated value of the length of the vehicle AM, and it may be possible to use the length of a typical vehicle, or it may be calculated based on the first point cloud information as detailed below.

[0091] Then, the rear-end position detection unit 240 sets the region including this predicted rear-end position yRP as the rear-end position detection region VRD. The rear-end position detection region VRD is set as a region determined by, for example, yVRDL ≤ y ≤ yVRDH. Since the rear-end position RE of the vehicle AM may be located in front of or behind this predicted rear-end position yRP, the lower limit yVRDL and the upper limit yVRDH of the rear-end position detection region VRD are arranged so as to sandwich the predicted rear-end position yRP as shown in FIG. 26 (yVRDL < yRP < yVRDH). In the example shown in FIG. 26, the upper limit yVRDH of the rear-end position detection region VRD is set to the same position as the predicted front-end position yFP, but it may be set behind the predicted front-end position yFP.

[0092] When the lower limit yVRDL of the rear end position detection area VRD is in the third area A3 (FIG. 27), the rear end position detection unit 240 detects the rear end point PR (the point with the smallest y coordinate) from the group of points in this rear end position detection area VRD. At this time, the rear end position detection unit 240 may detect the rear end point PR in all of the frames when the lower limit yVRDL of the rear end position detection area VRD is in the third area A3, or may detect the rear end point PR in some of the frames when the lower limit yVRDL of the rear end position detection area VRD is in the third area A3.

[0093] When the number of frames used to detect the rear end point PR is M3, the rear end position detection unit 240 detects M3 rear end points PRn (1,...,M3).The rear end position detection unit 240 then calculates the average (yPRn+...+yPRM3) / M3 of the y coordinate values ​​yPRn of the M3 rear end points PRn as the rear end position RE.The rear end position detection unit 240 also calculates the average (TF1+...+TFM3) / M3 of the timestamps TFn of the M3 frames Fn in which the rear end point PRn was detected as the second time T2.

[0094] FIG. 28 is a diagram showing an example of the processing operation in the rear end position detection unit 240 according to this embodiment. The rear end position detection unit 240 calculates a predicted rear end position yRP for the vehicle AM ​​whose vehicle speed VAM is calculated by the vehicle speed calculation unit 220 (step S2801). The rear end position detection unit 240 sets an area including the predicted rear end position yRP as a rear end position detection area VRD (step S2802). The rear end position detection unit 240 detects the rearmost point PRn from the point cloud within the rear end position detection area VRD in the frame Fn during which the lower limit yVRDL of the rear end position detection area VRD is in the third area A3 (step S2803). The rear end position detection unit 240 calculates the rear end position RE of the vehicle AM ​​at the second time T2 based on the detected rear end point PRn and the timestamp TFn of the frame Fn in which the rear end point PRn is detected (step S2804).

[0095] <Estimated vehicle length> The estimated vehicle length LE may be, for example, the length of a typical vehicle. However, the vehicle length is significantly different between a two-wheeled vehicle and a semi-trailer. Therefore, the vehicle speed calculation unit 220 of the control unit 200 according to an embodiment of the present invention further includes an estimated vehicle length calculation unit 227 that calculates an estimated vehicle length LE, which is a value that estimates the vehicle length of the vehicle AM, based on the first point cloud information. Then, the rear end position detection unit 240 according to this embodiment calculates a predicted rear end position yPR using the estimated vehicle length LE calculated by the estimated vehicle length calculation unit 227.

[0096] The estimated vehicle length calculation unit 227 calculates the estimated vehicle length LE, which is an estimate of the vehicle length of the vehicle AM, based on the temporal change of the point cloud included in the continuity determination area AC. The estimated vehicle length calculation unit 227 determines that the point clouds of the same vehicle are continuous as long as points exist in the continuity determination area AC in consecutive frames, just like the continuity determination unit 225. In other words, the estimated vehicle length calculation unit 227 determines that the three points PC1, PC2, and PC3 in FIG. 18 are continuous.

[0097] Then, the estimated vehicle length calculation unit 227 calculates the estimated vehicle length LE by calculating the distance between the leading point and the trailing point of the consecutive points. In the example of FIG. 18, the distance between the points PC1 and PC3 is calculated. For example, if the leading point of the consecutive points is the y coordinate value yE of the leading point in the frame FE when the consecutive points enter the continuity determination area AC, and the time stamp of the frame FE is TFE, and the trailing point of the consecutive points is the y coordinate value yL of the trailing point in the frame FL when the consecutive points leave the continuity determination area AC, and the time stamp of the frame FL is TFL, the estimated vehicle length LE is calculated as LE=yE-yL+VAM(TFL-TFE) if the speed of the vehicle corresponding to the consecutive points is VAM. Alternatively, the estimated vehicle length LE may be calculated as LE=AC(TFL-TFE) from the width of AC and the time stamp.

[0098] The present invention has been described above in terms of preferred embodiments thereof. Although the present invention has been described herein with reference to specific examples, various modifications and variations can be made to these examples without departing from the spirit and scope of the present invention as set forth in the claims. [Explanation of symbols]

[0099] 1000 LiDAR System 100 LiDAR 200 Control section 210 Point cloud information acquisition unit 220 Vehicle speed calculation unit 221 Segmentation Department 222 Representative point setting section 223 Tip position movement estimation unit 224 Speed ​​calculation section 225 Continuity Determination Unit 226 Representative point scanning time estimation unit 227 Estimated vehicle length calculation section 230 Tip position detector 240 Rear end position detector 250 Vehicle length calculation unit

Claims

1. a point cloud acquisition unit that acquires, in time series, first point cloud information based on three-dimensional point cloud information of a first region of a moving body path whose moving direction is determined, and second point cloud information based on three-dimensional point cloud information of a second region downstream of the first region; a moving object speed calculation unit that calculates a speed of the moving object by estimating a movement of a tip position of the moving object based on a time change of the first point cloud information; and a tip position detection unit that estimates a tip position of the moving body from a point cloud within a range determined by a tip position of the moving body estimated from a speed of the moving body in the second point cloud information, and calculates the tip position of the moving body at a first time based on a change over time in the estimated tip position.

2. the point cloud acquisition unit acquires third point cloud information based on three-dimensional point cloud information of a third region located downstream of the second region; The information processing device according to claim 1, further comprising a rear end position detection unit that estimates a rear end position of the moving body from a point cloud within a range determined by an estimated moving body length, which is an estimated value of the length of the moving body, and estimates the rear end position of the moving body from the third point cloud information, and calculates the rear end position of the moving body at a second time based on a change over time in the estimated rear end position.

3. 3. The information processing device according to claim 2, further comprising an estimated moving body length calculation unit that calculates the estimated moving body length by determining continuity of points of the first point cloud information based on a temporal change in a point cloud included in a continuity determination area set within the first area.

4. 4. The information processing device according to claim 2, further comprising a moving body length calculation unit that calculates a length of the moving body based on the calculated speed of the moving body, the calculated leading end position of the moving body at the first time, and the calculated trailing end position of the moving body at the second time.

5. An information processing device according to any one of claims 2 to 4; A LiDAR system having a LiDAR that generates three-dimensional point cloud information of the first area, the second area, and the third area.

6. a point cloud acquisition step of acquiring, in time series, first point cloud information based on three-dimensional point cloud information of a first region of a moving body path whose moving direction is determined, and second point cloud information based on three-dimensional point cloud information of a second region downstream of the first region; a moving object speed calculation step of estimating a movement of a tip position of the moving object based on a time change of the first point cloud information, to calculate a speed of the moving object; and a tip position detection step of estimating a tip position of the moving body from a point cloud within a range determined by a tip position of the moving body estimated from a speed of the moving body in the second point cloud information, and calculating a tip position of the moving body at a first time based on a change over time in the estimated tip position.

7. An information processing program that causes a computer to execute the information processing method according to claim 6.

8. A storage medium storing the information processing program according to claim 7.

Images