JP2011095914A - System, method and program for tracking vehicle - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a vehicle tracking system, method and program, allowing improvement of accuracy of vehicle tracking processing by reducing false detection when tracking a vehicle by use of a sensor detecting the vehicle on a road. <P>SOLUTION: When one of respective unit areas on the road detected by the sensor 11 is in a specific physical state indicating possibility of presence of the vehicle, a movement track calculation means 12 calculates a movement track as a temporal change. A vehicle track analysis means 14 analyzes a travel route of the vehicle by use of the movement track and a ratio of the false detection of the vehicle of each unit area registered in a false detection ratio registration means 13. <P>COPYRIGHT: (C)2011,JPO&amp;INPIT

Description

  The present invention relates to a vehicle tracking system, a vehicle tracking method, and a vehicle tracking program for tracking a vehicle. The present invention particularly relates to a vehicle tracking system, a vehicle tracking method, and a vehicle tracking program that can suppress erroneous tracking of a vehicle.

  As a keyword for automobile development in recent years, “safety and security” is widely used. With the aging of Japanese society, it is inevitable that the elderly population of drivers who drive vehicles will increase in the future. Therefore, it is strongly desired for the automobile industry to support safe and reliable driving for elderly drivers who have weak judgment, attention, or physical strength.

  “Safety / safety” for automobiles means avoiding traffic accidents and minimizing the damage in the event of a traffic accident. With the advancement of technology such as higher performance of automobiles and revisions to the Road Traffic Law, the number of traffic fatalities has been steadily decreasing from 9,640 in 1997 to 6,352 in 2006. Yes. On the other hand, the number of injuries caused by traffic accidents is almost flat, and further improvements are required for the safety of automobiles.

  By the way, as a technique necessary for avoiding a collision between an automobile and a person or an object, there is an automobile surrounding situation recognition technique. For example, a front vehicle collision avoidance system using millimeter wave radar, a night road image clarifying system by displaying a far infrared image on a car navigation screen, and a lane departure warning system by recognizing a white line using a front camera It has been put into practical use as a seed technology. Each of these vehicle surrounding situation recognition technologies completes processing with one vehicle.

  On the other hand, vehicle-vehicle cooperation technology and road-vehicle cooperation technology exist as approaches for realizing higher-level situation recognition. With vehicle-to-vehicle cooperation technology, each vehicle traveling on the road communicates wirelessly with each other and provides the surrounding situation of the own vehicle to other vehicles, so it is possible to acquire the surrounding situation that is difficult to grasp only with the own vehicle become. By realizing vehicle-to-vehicle cooperation, for example, it is expected to avoid a collision accident at the time of encounter due to another vehicle popping out from the blind spot.

  In road-vehicle cooperation technology, the running state of each vehicle is monitored using a sensor installed on the side of the road, and the running information is distributed to each vehicle. By realizing road-vehicle cooperation, for example, it becomes possible to detect and notify a vehicle at the end of a traffic jam. In road-vehicle cooperation technology, a camera may be used as one of various sensors. By detecting individual vehicles from the road image acquired by the camera and tracking the travel route, it is possible to construct a collision prevention warning system for a right turn vehicle and a straight vehicle in an intersection.

  In a service using a road-vehicle cooperation framework, if there is an error in vehicle detection or tracking, a system that is inconvenient for the user, such as sending a “stop” signal to the vehicle when it can start. End up. Therefore, it is necessary to improve the accuracy of vehicle tracking by identifying a vehicle false detection signal output from the sensor and performing post-processing such as rejection.

  For example, it is assumed that a vehicle constructed in advance using machine learning such as a neural network is used for collation with respect to an image acquired from a sensor as a camera to detect vehicles in the image. At this time, if a texture such as a shadow or a pedestrian crossing in the image has a feature similar to that of a vehicle, a false detection signal is repeatedly generated at the corresponding location. In the following description, such erroneous detection will be referred to as steady erroneous detection, and the vehicle motion trajectory calculated therefrom will be referred to as erroneous detection trajectory.

  FIG. 31 shows an example of regular erroneous detection. FIGS. 31A to 31E show five images taken at the same place on the road taken by a camera (not shown). Here, time elapses in order from FIG. 31 (A) to FIG. 31 (E).

  In each image showing a road, a pedestrian crossing mark 951 is recorded at intervals in the horizontal direction. It is assumed that the vehicle 952 travels across the pedestrian crossing mark 951. Although it depends on the position of the sun, the pixel value as the image density of the windshield of the vehicle 952 is similar to the pixel value of the pedestrian cross mark 951.

  Therefore, the camera that should sequentially detect the traveling position 953 of the vehicle 952 erroneously detects the position 955 indicating the presence of the region 954 including a part of the mark 951 of the pedestrian crossing as the position of the vehicle. Since the pedestrian crossing mark 951 is drawn at a predetermined position on the road, the vehicle tracking device (not shown) that receives the camera information recognizes that the vehicle is stopped on the pedestrian crossing.

  If such erroneous recognition occurs randomly in time, it can be distinguished from a true vehicle detection signal by adopting a technique such as averaging the occurrence frequency. However, such a technique cannot be adopted if erroneous detection occurs regularly. In addition, when erroneous detection of the vehicle 952 occurs repeatedly at a specific place such as a pedestrian crossing, if these detection results are invalidated, the vehicle 952 stopped at the specific place cannot be detected.

  Therefore, a technique for recognizing an object or a situation on a road based on a signal that is constantly generated is a first related technique (for example, see Patent Document 1) and a second related technique (for example, see Patent Document 2) of the present invention. ).

  Among these, the first related technique deals with a technique for determining a stationary object on a road using an image, that is, an obstacle. The obstacle determination unit according to the first related technique uses six consecutive plane images among the plane images obtained by capturing 30 frames per second. Then, when the positions of the binarized objects in the first and sixth images in these image sequences match, this object is considered to be stationary.

  The second related technique of the present invention deals with a technique for detecting what is truly an obstacle to the traveling of the host vehicle. In the second related technique, a data group of detection points by a radar for an object on a road is divided by a traveling road dividing line. Then, clusters that are different between what is on the road and those that are outside, that is, grouped in recording units, are used to detect what is truly an obstacle to the running of the vehicle. Yes.

JP2007-156568 (paragraph 0043, FIG. 1) Japanese Patent Laid-Open No. 08-124080 (paragraphs 0009 and 0028, FIG. 5)

  Among these, in the first related technique, the movement is not observed on a plurality of images, and the shape of the object that is considered to be stationary does not match the registered shape of the vehicle, and is determined as an obstacle. For this reason, when the texture information is similar to the vehicle, it is excluded from the obstacle, and as a result of this, it is impossible to identify this type of false detection that occurs regularly.

  On the other hand, in the second related technique, the “moving body-likeness” is defined from the smoothness of the motion trajectory of the moving object, the length of the observation time, and the positional relationship between the sensor detection range and the detected object. Then, the object that is less likely to be a moving object is deleted from the moving object. However, according to the second related technique, the observation time becomes longer for the erroneous detection that occurs regularly. For this reason, there is a problem that the likelihood of a moving object is high with respect to erroneous detection that occurs regularly, and as a result, it is determined that the detection is erroneous and cannot be deleted.

  SUMMARY OF THE INVENTION Accordingly, an object of the present invention is to provide a vehicle tracking system and a vehicle tracking method capable of reducing false detections and improving the accuracy of vehicle tracking processing when tracking a vehicle using a sensor that detects a vehicle on a road. And providing a vehicle tracking program.

  In the present invention, (b) a sensor that sequentially detects the physical state on the road in each unit area obtained by dividing a predetermined range of the area on the road into a plurality of times, and (b) detected by this sensor. When the physical state of any of the unit areas on the road described above is a specific physical state indicating the possibility of a vehicle, the motion trajectory is calculated as the movement path of the temporal position of these unit areas. (C) an erroneous detection rate registration unit for registering a rate at which the specific physical state described above is steadily generated for each unit area, and (d) the above-described motion track calculation unit. Vehicle tracking analysis means for analyzing the travel route of the vehicle using the calculated motion trajectory and the vehicle false detection rate for each unit area registered in the error detection rate registration means. Both tracking device comprises.

  In the present invention, (a) a sensor detection step of sequentially detecting a physical state on the road in each unit area obtained by dividing a predetermined range of the predetermined area on the road with time using a sensor; B) When the physical state of any one of the unit areas detected by the sensor detection step is a specific physical state indicating the possibility of the presence of the vehicle, as a movement path of the temporal position of these unit areas (C) an erroneous detection rate registration step for registering a rate at which the above-described specific physical state is steadily generated for each unit area; The vehicle using the motion trajectory calculated in the motion trajectory calculation step and the vehicle false detection rate for each unit area registered in the error detection rate registration step. Vehicle Tracking analysis step and the vehicle tracking method of analyzing the path of travel is provided.

  Further, according to the present invention, as a vehicle tracking program, (a) a physical state on the road in each unit area obtained by dividing a predetermined range on the road into a plurality of parts is sequentially detected using a sensor. (B) when the physical state of any of the unit areas detected by this sensor detection process is a specific physical state indicating the possibility of the presence of the vehicle, the time of these unit areas A motion trajectory calculation process for calculating a motion trajectory as a movement path of a specific position, and (c) an erroneous detection rate registration process for registering a rate at which the above-described specific physical state constantly occurs for each unit region. And (d) erroneous detection of the vehicle for each unit region registered in the above-described motion trajectory calculated by the above-described motion trajectory calculation processing and the above-described erroneous detection rate registration processing. Using the ratio is characterized by executing a vehicle tracking analysis process for analyzing the path of travel of the vehicle.

  As described above, according to the present invention, since the vehicle motion trajectory is tracked by reflecting the erroneous detection result of the vehicle that occurs regularly, it is possible to prevent erroneous tracking.

It is a claim corresponding | compatible figure of the vehicle tracking apparatus of this invention. It is a claim corresponding | compatible figure of the vehicle tracking method of this invention. It is a claim corresponding | compatible figure of the vehicle tracking program of this invention. 1 is a system configuration diagram showing a vehicle tracking system according to a first embodiment of the present invention. It is the top view which showed the tracking attention area | region of the vehicle made into the detection target by the vehicle position detection sensor in 1st Embodiment. It is the top view which showed the example which divided | segmented the tracking attention area | region in 1st Embodiment into the unit area. It is explanatory drawing which represented numerically the detection frequency in which the erroneous detection of each unit area | region in the tracking attention area of 1st Embodiment generate | occur | produces. It is a top view showing the position of each vehicle in time T-1 of a 1st embodiment. It is a top view showing the position of each vehicle in the time T of 1st Embodiment. It is explanatory drawing which showed an example of matching with the exercise | movement locus | trajectory obtained by time T-1 in one lane which exists in the tracking attention area | region of the time T in 1st Embodiment, and a candidate detection position. It is explanatory drawing which showed the area | region where the analysis of the movement locus | trajectory in 1st Embodiment is performed. It is a characteristic view showing an example of an exercise locus calculated based on a vehicle detection result in the first embodiment. It is the characteristic view which showed the relationship between the tracking start position and deviation | shift amount in 1st Embodiment. It is explanatory drawing which showed the exceptional case where the shift | offset | difference D in 1st Embodiment becomes large. It is a characteristic figure which shows the reason why it is necessary to discriminate | determine the erroneous detection steady generation | occurrence | production area | region in 1st Embodiment. It is explanatory drawing which showed the main area | regions in the memory in the vehicle tracking apparatus of FIG. 3 is a flowchart specifically showing the operation of the vehicle tracking system according to the first embodiment. It is a system block diagram showing the vehicle tracking system in the 2nd Embodiment of this invention. It is explanatory drawing which shows the principle of operation of the misdetection locus | trajectory discrimination | determination part in 2nd Embodiment. It is the block diagram which showed the main area | regions in the memory in the vehicle tracking apparatus shown in FIG. 6 is a flowchart specifically illustrating the operation of the vehicle tracking system according to the second embodiment. It is a system block diagram showing the vehicle tracking system in the 3rd Embodiment of this invention. It is explanatory drawing which shows the principle of operation of the tracking locus | trajectory calculation part in 3rd Embodiment. It is the block diagram which showed the main area | regions in the memory in the vehicle tracking apparatus shown in FIG. 12 is a flowchart specifically illustrating the operation of the vehicle tracking system according to the third embodiment. It is a system block diagram showing the vehicle tracking system in the 4th Embodiment of this invention. It is the block diagram which showed the main area | regions in the memory in the vehicle tracking apparatus shown in FIG. It is a flowchart showing the mode of processing of the tracking start judgment part in a 4th embodiment. It is a characteristic view which shows the principle of calculation of the movement locus | trajectory feature-value f in 4th Embodiment. It is a flowchart showing the mode of processing of the false detection locus feature calculation part in a 4th embodiment. It is explanatory drawing which showed the example of regular false detection.

  FIG. 1 shows a claim correspondence diagram of the vehicle tracking system of the present invention. The vehicle tracking system 10 of the present invention includes a sensor 11, a motion trajectory calculation unit 12, an erroneous detection rate registration unit 13, and a vehicle tracking analysis unit 14. Here, the sensor 11 sequentially detects the physical state on the road in each unit area obtained by dividing a predetermined range on the road into a plurality of parts over time. When the physical state of any of the unit areas on the road detected by the sensor 11 is a specific physical state indicating the possibility of the vehicle, the motion trajectory calculation means 12 The motion trajectory is calculated as a temporal position movement path. The false detection rate registration means 13 registers the rate at which the specific physical state described above is constantly generated for each unit area. The vehicle tracking analysis unit 14 analyzes the travel route of the vehicle using the motion trajectory calculated by the motion trajectory calculation unit 12 and the false detection rate of the vehicle for each unit area registered in the erroneous detection rate registration unit 13. .

  FIG. 2 shows a claim correspondence diagram of the vehicle tracking method of the present invention. The vehicle tracking method 20 of the present invention includes a sensor detection step 21, an exercise locus calculation step 22, an erroneous detection rate registration step 23, and a vehicle tracking analysis step 24. Here, in the sensor detection step 21, the physical state on the road in each unit area obtained by dividing a predetermined range of the area on the road into a plurality of parts is sequentially detected using a sensor over time. In the motion trajectory calculation step 22, when any physical state of the unit area detected by the sensor detection step 21 is a specific physical state indicating the possibility of the presence of the vehicle, the time of these unit areas The motion trajectory is calculated as the position movement path. In the erroneous detection rate registration step 23, the rate at which the specific physical state described above is constantly generated is registered for each unit area. In the vehicle tracking analysis step 24, the travel route of the vehicle is analyzed using the motion trajectory calculated in the motion trajectory calculation step 22 and the false detection rate of the vehicle for each unit area registered in the false detection rate registration step 23. .

  FIG. 3 shows a claim correspondence diagram of the vehicle tracking program of the present invention. The vehicle tracking program 30 of the present invention causes a computer to execute a sensor detection process 31, a motion trajectory calculation process 32, an erroneous detection rate registration process 33, and a vehicle tracking analysis process 34 as a vehicle tracking program. . Here, in the sensor detection process 31, the physical state on the road in each unit area obtained by dividing a predetermined range of the area on the road into a plurality of parts is sequentially detected using a sensor over time. In the motion trajectory calculation process 32, when any of the physical states of the unit areas detected by the sensor detection process 31 is a specific physical state indicating the possibility of the presence of the vehicle, the temporal movement of these unit areas The motion trajectory is calculated as the position movement path. In the erroneous detection rate registration process 33, the rate at which the specific physical state described above is constantly generated is registered for each unit area. In the vehicle tracking analysis process 34, the travel path of the vehicle is determined using the motion trajectory calculated by the motion trajectory calculation process 32 and the vehicle false detection rate for each unit area registered in the error detection rate registration process 33. To analyze.

  <First Embodiment of the Invention>

  Next, a first embodiment of the present invention will be described.

  FIG. 4 shows a vehicle tracking system according to the first embodiment of the present invention. The vehicle tracking system 100 according to the present embodiment includes a vehicle position detection sensor 102 installed at a position overlooking a predetermined range of a road 101 from above, for example. The vehicle position detection sensor 102 includes, for example, a camera, an ultrasonic sensor, a loop coil, a laser radar, and a millimeter wave radar. Such a vehicle position detection sensor 102 detects the traveling of the vehicle 103 within a tracking attention area consisting of a predetermined length of a road having a predetermined width (three lanes in the figure). The vehicle position detection sensor 102 does not have to be a single sensor, but may be a set of a plurality of sensors. The detection data 104 of the vehicle position detection sensor 102 is input to a vehicle tracking device 105 for tracking the vehicle in the tracking attention area.

  The vehicle tracking device 105 includes a CPU (Central Processing Unit) 111 and a control unit 113 including a memory 112. The control unit 113 includes a tracking locus calculation unit 115, a tracking locus comparison unit 116, an erroneous detection locus determination unit 117, an erroneous detection steady occurrence region determination unit 118, and an erroneous detection locus deletion unit 119 via a bus 114 such as a data bus. Control. The control unit 113 executes a control program stored in the memory 112 to thereby delete the tracking locus calculation unit 115, the tracking locus comparison unit 116, the erroneous detection locus determination unit 117, the erroneous detection steady occurrence region determination unit 118, and the erroneous detection locus deletion. At least a part of the unit 119 can be realized by software.

  Here, the tracking trajectory calculation unit 115 tracks the moving position of the vehicle 103 traveling on the road 101 in the tracking attention area. The tracking trajectory comparison unit 116 compares the motion trajectories of the plurality of vehicles 103 calculated by the tracking trajectory calculation unit 115. The erroneous detection trajectory determination unit 117 determines a trajectory that is erroneously detected for the vehicle 103. The erroneous detection steady occurrence region discriminating unit 118 discriminates a region where the erroneous detection in the tracking attention region regularly occurs. The erroneous detection trajectory deletion unit 119 excludes the trajectory calculated from the time series data indicating the erroneous detection of the vehicle 103 from the target of the tracking process.

FIG. 5 shows the tracking attention area of the vehicle to be detected by the vehicle position detection sensor shown in FIG. 4 with hatching. In the example picked up in the present embodiment, the road 101 is composed of three lanes (lanes) one way. Here, the tracking attention area 131 is a belt-like area from the _first position P ₁ to the second position P ₂ of the road 101 in the traveling direction of the vehicle (not shown). Each position of the vehicle 103 shown in FIG. 4 is tracked by the tracking attention area 131, and erroneous detection data generated in the tracking process is eliminated.

  For this reason, in the present embodiment, data on which position of the tracking attention area 131 the vehicle detection signal is generated is used. This data is formed by dividing the tracking attention area 131 into a plurality of unit areas and representing the vehicle detection frequency in each unit area.

  FIG. 6 shows an example in which the tracking attention area is divided into unit areas. The tracking attention area 131 is divided into a matrix so as not to straddle each lane, as indicated by a broken line. Each cell by this constitutes a unit region 141. In this example, the tracking attention area 131 is equally divided into 10 in the vehicle traveling direction, and each lane is equally divided into 2 in the width direction, and is divided into 60 unit areas 141 in total.

  FIG. 7 shows an example in which the detection frequency of each unit area in the tracking attention area within a predetermined time is represented by a numerical value. Here, in the part represented as “0” in the unit area 141, it means that the vehicle has never been detected. On the other hand, a portion indicated by “89” in the unit area 141 indicates that the vehicle has been detected 89 times within a predetermined time.

  FIG. 8 and FIG. 9 are for explaining the principle of vehicle tracking in a road tracking attention area in the present embodiment. Among these, FIG. 9 shows the position of each vehicle 103 at time T of interest, and FIG. 8 shows the position of each vehicle 103 in the past by unit time “1”. In these drawings, each vehicle 103 has a direction indicated by an arrow 145 as a vehicle traveling direction. This will be described with reference to FIGS.

First, the state at time T-1 in FIG. 8 will be described. In the first lane 151 of the road 101, the vehicle 103 ₁₁ and the vehicle 103 ₁₂ are detected in the displayed positions in the tracking attention area 131. Dashed line leading to each of the vehicle 103 ₁₁ and the vehicle 103 ₁₂ represent the movement locus 161 _11, 161 ₁₂ of these vehicles 103 _11, 103 ₁₂ in the tracking target area 131. Similarly, the second lane 152, the vehicle 103 ₂₁ and the vehicle 103 ₂₂ in the tracking target area 131 is detected at a position that is displayed. Similarly, the third lane 153 is detected at a position where the vehicle 103 ₃₁ is displayed on the tracking target area 131. Other vehicles 103 ₃₂ is detected in front of the position of the tracking target area 131. It is assumed that the motion trajectories 161 ₁₁ , 161 ₁₂ , 161 ₂₁ , 161 ₂₂ , and 161 ₃₁ are determined in the tracking attention area 131.

Attention is paid to first lane 151 shown in FIG. 9 at time T when unit time “1” has elapsed from time T-1 shown in FIG. Here has moved the actual vehicle 103 ₁₁ at a position deviated from the tracking target area 131, the other vehicle 103 _12, moved to a position shifted upward in FIG than the position shown in FIG. 8 It shall be. Similarly, vehicles 103 ₂₁ , 103 ₂₂ , 103 ₃₁ , 103 ₃₂ are located in the second lane 152 and the third lane 153 of FIG.

However, it should be noted that the vehicle 103 ₁₁ in the first lane 151 is out of the tracking attention area 131 and is not a detection target. That is, from the detection result of the vehicle position detection sensor 102 shown in FIG. 4 at the time T, the positions of the vehicles 103 ₁₂ , 103 ₂₁ , 103 ₂₂ , 103 ₃₁ , 103 ₃₂ marked with “X” in the tracking attention area 131. In addition, only what seems to be a vehicle has been detected. In the tracking attention area 131, candidate detection positions 171 ₁₁ , 171 ₁₂ , and 171 ₃₁ of the vehicle 103 are also detected at the locations indicated by the crosses similarly.

Therefore, the vehicle tracking device 105 needs to determine the position of each vehicle 103 at the time T from among these, and extend the motion trajectories 161 ₁₁ , 161 ₁₂ , 161 ₂₁ , 161 ₂₂ , 161 ₃₁ . Also, the vehicle 103 ₃₂ appearing in new tracking target area 131, and starts tracking the trajectory, if the vehicle 103 deviates from the tracking target area 131, it is necessary to terminate the tracking in this region.

  Therefore, a new motion trajectory at time T is generated based on the motion trajectory 161 of each vehicle 103 up to time T-1. For this purpose, the tracking trajectory comparison unit 116 shown in FIG. 4 compares the vehicle tracking trajectories to be compared at time T. Similarly, the erroneous detection trajectory discriminating unit 117 shown in FIG. 4 detects a motion trajectory that cannot be generated based on the time series data of the vehicle position obtained by correctly detecting the position of the vehicle 103 at the time of regular erroneous detection. It is determined that the trajectory is calculated from the series data. For this purpose, the vehicle detection result obtained by the vehicle position detection sensor 102 is reflected in the vehicle detection position distribution, and a steady erroneous detection occurrence region is determined from the distribution information.

The circles marked at the tips of the motion trajectories 161 ₁₁ , 161 ₁₂ , 161 ₂₁ , 161 ₂₂ , and 161 ₃₁ shown in FIG. 9 are the tip positions of the trajectories at time T-1. Based on these end positions marked with ○, the correct vehicle position is associated with each of the candidate detection positions 171 ₁₁ , 171 ₁₂ , and 171 _{31 at} time T and the other positions of the vehicle 103 marked with ×. Become. Incidentally, the candidate detection positions 171 ₁₁ , 171 ₁₂ , and 171 ₃₁ shown in FIG. 9 are positions that are finally determined as errors.

FIG. 10 shows an example of the correspondence between the motion trajectory obtained up to time T-1 in one lane existing in the tracking attention area at time T and the candidate detection position. In this example, there is a problem with the correspondence between the end position indicated by a circle in the motion trajectory 161 and the three candidate detection positions 171 ₁ , 171 ₂ , and 171 ₃ existing in the vicinity thereof.

  As an example of a method of associating with the end position of the motion trajectory, when the distance from the end of the motion trajectory to the detection position is l, the following equation (1) defined using this distance l is used as the correspondence score: Can be used as

D _S = 1 / l (1)

With this (1), in the example of FIG. 10, the most corresponding score D _S distance from the terminal end position is shortest candidate detection position 171 ₁ shown increases in ○ sign of movement trajectory 161. Further, the distance is the farthest candidate detection position 171 ₃ most corresponding score D _S decreases from the end position.

In this way, in this association method, the score is calculated using the distances from all candidate detection positions 171 (all positions marked with “x” in FIG. 10) for each end of the motion trajectory. The 1-position in the ascending order of the corresponding scores, position 2, performs ... and ranking, the corresponding score candidate detection position 171 ₁ D _S corresponds to the 1-position, considered as that corresponding to the movement locus termination of interest.

For a motion trajectory that has been successfully associated with the candidate detection position 171, its end is extended to the corresponding candidate detection position 171. Thereby, the motion trajectory at the current time T is calculated. The movement trajectory 161 of the vehicle 103 (for example, the vehicle 103 ₁₁ in FIG. 9) moved from the inside of the tracking attention area 131 to the outside as time passes disappears at time T. In addition, a movement locus 161 of the vehicle 103 (for example, the vehicle 103 ₃₂ in FIG. 9) moved from the outside to the inside of the tracking attention area 131 is newly generated. Therefore, the number of motion trajectories 161 varies at each time.

  The tracking trajectory comparison unit 116 shown in FIG. 4 performs the tracking trajectory 161 of the predetermined number of vehicles 103 existing in the tracking attention area 131 at the time T shown in FIG. Two motion trajectories 161 waiting for detection of the position of the vehicle 103 are selected. Then, these motion trajectories 161 are compared in a time range having a finite fixed width.

Here, the reason why the fixed width time range is used for the comparison processing of the tracking locus comparison unit 116 will be described. For example, depending on the traffic situation such as traffic jam, the time required from the appearance to disappearance of the vehicle 103 of interest in the tracking attention area 131 changes dynamically. Therefore, it is difficult to dynamically adjust the time range used for comparing the movement trajectories according to the appearance and disappearance of the vehicle 103. Therefore, as shown in FIG. 5, the upper limit position of the tracking attention area 131 in the traveling direction of the vehicle 103 on the road 101 is set as P ₁ and the lower limit position is set as P ₂ . Further, the axis representing the vehicle position with the traveling direction of the vehicle 103 facing upward is taken as the vertical axis, and the time is taken as the horizontal axis.

FIG. 11 shows a region where the motion trajectory is analyzed in this example. For example, the first position P ₁ and the second position P ₂ as fixed points are taken in the vertical axis direction as the vehicle traveling direction shown in FIG. On the horizontal axis, _two fixed times of time T ₁ and time T ₂ are taken. In the present embodiment, the motion trajectory is analyzed in a fixed range as a rectangular region 181 surrounded by these positions P ₁ and P ₂ and times T ₁ and T ₂ .

  FIG. 12 shows an example of a motion trajectory calculated based on the vehicle detection result. This will be described with reference to FIGS.

  It is assumed that the vehicle position detection sensor 102 normally detects the vehicle 103 and that the vehicle 103 is traveling in the direction indicated by the arrow 145 on the first lane 151, for example. In this case, the movement trajectory 191 of the vehicle 103 is basically a broken line that rises to the right, which may include a horizontal portion in part, as shown by a solid line in FIG.

On the other hand, in the case of the motion locus calculated based on the result of the erroneous vehicle detection that occurs regularly, the vehicle detection result exists at the same location on the first lane 151, for example. As a result, it is observed that the vehicle is stopped. Accordingly, the motion trajectory 192 is composed of a line that moves forward and backward in the vehicle traveling direction in a predetermined width around the stop position P _A of the vehicle 103 as indicated by the broken line in FIG. 12.

No vehicle corresponding to the motion trajectory 192 indicated by a broken line present in this manner stop position P _A. On the other hand, in the case of the movement locus 191 indicated by a solid line, the vehicle 103 is actually traveling. Therefore, when the motion trajectory 191, will run the first lane 151 without the vehicle 103 is stopped once before the stop position P _A. As a result, as shown in the circle 193 in FIG. 12, the movement trajectories 191 and 192 cross each other.

First, the erroneous detection trajectory discriminating unit 117 shown in FIG. 4 observes the presence or absence of the intersection of the two motion trajectories 191 and 192 shown in FIG. When there is an intersection, attention is paid to a motion trajectory 192 in which the motion trajectory is regarded as horizontal. Then, it calculates the intersecting position P _A of the two motion trajectories 191 and 192, the deviation D between the tracking start position. Here, the tracking start position refers to the vehicle detection position at the time when the tracking of the movement locus 192 is started. In the case of FIG. 12, the tracking start position of the motion trajectory 192 is P _S and the intersection position is P _A. Accordingly deviation D, as shown in FIG. 12, the difference between the positions P _S and the position P _A.

FIG. 13 is a diagram for explaining the relationship between the tracking start position and the shift amount. In FIG. 13, the movement locus 191 is the same as that in FIG. On the other hand, a motion trajectory 195 in which the motion trajectory between time T ₁ and time T ₂ is very similar to the motion trajectory 192 of FIG. 12 represents the actual traveling of the vehicle 103, and at a position P _B for a predetermined time. Only have stopped its running.

In the case of the motion trajectory 195, a misdetection result in which the deviation D between the tracking start position P _C and the motion trajectory 195 at the intersection start point P _B at the tracking start time before the time T ₁ occurs regularly. The displacement D in the movement locus 191 based on the above is greatly exceeded. Therefore, in the case of the movement locus 195, it can be determined that it represents the actual traveling of the vehicle 103.

  In addition, in the case of FIG. 13, the motion locus 191 and the motion locus 195 intersect as in FIG. 12, as shown in the circle 193. Therefore, in the case of FIG. 13, the vehicle 103 indicated by the movement locus 191 and the other vehicle 103 indicated by the movement locus 195 collide at the position indicated inside the circle 193, and the vehicle 103 indicated by the movement locus 191 continues thereafter. This means that the vehicle is moving in the direction of travel. That is, in the case of FIG. 13, it should be determined that the motion trajectory 195 represents the actual detection result of the vehicle 103, and it is not correct to determine that the motion trajectory is a motion trajectory due to erroneous detection.

Meanwhile, in the example shown in FIG. 12, when the motion trajectory 192 is in the position substantially equal to as intersecting position P _A at the tracking start timing predated sufficient time than the time T _1, when is moving path due to an erroneous detection The probability of discrimination increases.

Thus, when a certain motion trajectory includes a “trajectory that can be regarded as horizontal” corresponding to the stop position of the vehicle 103, the deviation D between the tracking start position and the stop position is measured, and the magnitude thereof is calculated. It is effective to determine whether the movement locus is due to erroneous detection. This is because erroneous detections that occur regularly are repeatedly observed at the same location on the lane. Therefore, in the case of a motion trajectory due to such erroneous detection, the intersection position with the other motion trajectory corresponding to the stop position substantially coincides with the tracking start position. On the other hand, when the traveling vehicle actually stops, the vehicle 103 until the vehicle stops until the vehicle stops, such as the tracking start position P _C and the vehicle 103 stop position P _B in the movement locus 195 in FIG. This is because a significant deviation D occurs based on the running.

FIG. 14 shows an exceptional case where the deviation D increases. Second lane 152 as an example of tracking target area 131 illustrated in FIG. 14, it is assumed that the erroneous detection constant generating region 201 _{1-201 5} performs false detection constantly present. In this case, the first erroneous ○ mark detection constant generating region 201 in the _first vehicle traveling direction indicated by the arrow 145 in the tracking target area 131 is tracking start position 202. Also, the last erroneous ○ mark detection constant generating region 201 in ₅ of the vehicle traveling direction is the vehicle stop position 203.

  In this example, a motion trajectory indicated by a broken line connecting the tracking start position 202 and the vehicle stop position 203 is calculated. Since these are along the second lane 152, the deviation D is increased as a result.

  Therefore, the tracking attention area 131 is divided into a plurality of unit areas 141 as shown in FIG. Then, the misdetection locus determination unit 117 shown in FIG. 4 obtains the detection frequency by associating the detection positions of the vehicle 103 obtained from the vehicle position detection sensor 102 with the plurality of unit regions 141, respectively. Update. Thereby, a detection frequency distribution as shown in FIG. 7 is obtained.

Incidentally, in the moving path of interest, the detection frequency in each unit region 141 corresponding to the detected position at the time T and C _t. At this time, the sum C of the detection frequencies at the detection position from the tracking start time T ₀ to the current time T at which the vehicle is considered to have stopped can be expressed by the following equation (2).

  In the case of a motion trajectory calculated from a plurality of erroneously detected steady regions as shown in FIG. 14, the total sum C increases.

  The function F having the above-described deviation D and the total detection frequency C as inputs is expressed by the following equation (3).

  F (D, C) (3)

  Using this equation (3), the erroneous detection trajectory discrimination unit 117 shown in FIG. 4 determines whether or not the motion trajectory of interest is calculated based on the result of erroneous detection. For example, it is assumed that the function F (D, C) is expressed by the following equation (4).

F (D, C) = W ₁ × D + (W ₂ / C) (4)

Here, W ₁ and W ₂ are parameters.

  When the value represented by the equation (4) is smaller than a preset threshold value, it is considered that the observed motion trajectory is a motion trajectory obtained based on the erroneous detection result of the vehicle 103 that is constantly generated.

  The erroneous detection steady occurrence region determination unit 118 illustrated in FIG. 4 pays attention to the detection frequency distribution calculated by the erroneous detection locus determination unit 117. And the area | region near the edge part of the tracking attention area | region 131 among the unit area | regions 141 in which a detection frequency exceeds a threshold value is discriminate | determined as a misdetection steady generation | occurrence | production area | region.

FIG. 15 is for explaining the reason why it is necessary to determine the erroneously detected steady occurrence region. As shown in the candidate detection position 171 ₁₂ of FIG. 9, there may be a motion trajectory calculated from a steady false detection at the end point of the tracking attention area 131. In such a case, as shown in FIG. 15, the motion trajectory sticks to the boundary of the analysis range. As a result, discrimination based on the intersection of the motion trajectories described in FIG. 12 or FIG. 13 becomes impossible.

  In the unit area 141 corresponding to regular erroneous detection, the vehicle detection frequency naturally increases. Therefore, by combining the detection frequency of the vehicle 103 for each unit area 141, it is possible to discriminate even if a steady erroneous detection occurs in an arbitrary place in the tracking attention area 131. The erroneous detection trajectory deletion unit 119 excludes the motion trajectory calculated based on the vehicle erroneous detection result that is constantly generated by the erroneous detection trajectory determination unit 117 from the vehicle tracking process.

  FIG. 16 shows main areas in the memory in the vehicle tracking device of FIG. In the memory 112, several areas are set in addition to the program storage area 211 for storing programs executed by the CPU 111. The detection data storage buffer 212 temporarily stores the detection data 104 sent from the vehicle position detection sensor 102 shown in FIG. The observation data storage area 213 stores data excluding the detection data 104 sent from the vehicle position detection sensor 102 that corresponds to the erroneous detection steady occurrence area described with reference to FIG. The erroneous detection steady occurrence registration area 214 registers an erroneous detection steady occurrence area. The list GL215 stores all combinations of motion trajectories. The list Err216 registers a misdetection trajectory of the motion trajectory.

  FIG. 17 specifically shows the operation of the vehicle tracking system. This will be described with reference to FIGS. 4, 6, 7, 12 and 16.

  First, the control unit 113 in the vehicle tracking device 105 inputs the detection data 104 at the current time T from the vehicle position detection sensor 102 and stores it in the detection data storage buffer 212 of the memory 112. This detection data includes image data of a predetermined number of unit areas 141 obtained by dividing the tracking attention area 131 into a matrix as shown in FIG. As described with reference to FIG. 7, these image data include those that have a high probability of erroneous detection and are registered in the erroneous detection steady occurrence registration region 214. The control unit 113 uses the M pieces of image data excluding the image data of the unit area 141 corresponding to the erroneous detection steady occurrence area as observation data at the current time T indicating the vehicle detection result, and the observation data storage area of the memory 112. It is stored in 213 (step S301).

Next, the processing of the tracking trajectory calculation unit 115 is performed. The tracking trajectory calculation unit 115 calls the observation data obtained at time T-1 one stage before the current time T, and associates it with N _T-1 observation data at the current time T. Then, the motion trajectory at time T is calculated (step S302). Let N _T be the number of calculated motion trajectories.

Next, the processing of the tracking locus comparison unit 116 is performed. The tracking trajectory comparison unit 116 registers a motion trajectory having an end on the target lane of the road 101 in the list GL215 (step S303). The registered movement locus and G _T. Next, the tracking trajectory comparison unit 116 selects an unselected one from a combination of two arbitrary motion trajectories from the list GL215 (step S304).

  Next, the process proceeds to the process of the misdetection locus determination unit 117. The erroneous detection trajectory determination unit 117 compares the two motion trajectories selected in step S304. That is, the erroneous detection trajectory determination unit 117 determines whether or not there is an intersection of the motion trajectories 191 and 192 described with reference to FIG. 12 (step S305). If there is an intersection between the movement trajectories 191 and 192 (Y), the erroneous detection trajectory determination unit 117 checks whether there is a movement trajectory of the stopped vehicle (step S306).

  When there is a motion trajectory of the stopped vehicle as described with reference to FIG. 12 (step S306: Y), the erroneous detection trajectory determination unit 117 determines the detection position for each of the two traversing motion trajectories 191 and 192. The deviation D is calculated (step S307). And the misdetection locus | trajectory discrimination | determination part 117 determines the locus | trajectory with small deviation D as the exercise | movement locus | trajectory corresponding to a stop vehicle. Thereafter, the erroneous detection trajectory determination unit 117 updates the detection frequency of the unit area corresponding to the detection position of the vehicle 103 (step S308), and calculates the function F of the above-described equation (3) (step S309).

As a result, when the function F calculated in step S309 is less than the predetermined threshold value Thr ₁ (step S310: Y), the erroneous detection trajectory discriminating unit 117 displays the erroneous detection trajectory in the motion trajectory. Registration in the list Err216 (step S311). The erroneous detection trajectory determination unit 117 determines whether all combinations of motion trajectories listed in the list GL 215 have been selected (step S312). If there is an unselected combination (N), the process returns to step S304, and an unselected one is selected from the remaining two motion trajectory combinations. And the process after step S305 is repeated similarly.

If it is determined in step S310 that the function F is equal to or greater than the predetermined threshold Thr ₁ (N), the process proceeds to step S312 without performing registration in the list Err216 in step S311. If the erroneous detection trajectory discriminating unit 117 does not discriminate the presence of the motion trajectory of the stopped vehicle in Step S306 (N), the process proceeds to Step S312 at that time. Furthermore, if it is determined in step S305 that there is no intersection between the two motion trajectories (N), the process proceeds to step S312 at that time as well.

  If it is determined in step S312 that the tracking trajectory comparison unit 116 has selected all the combinations of motion trajectories (Y), the tracking trajectory comparison unit 116 determines whether all lanes (lanes) on the road 101 have been noticed. A determination is made (step S313). If there is a lane (lane) that has not been noticed yet (N), the tracking trajectory comparison unit 116 changes the current noticed lane to one of the remaining lanes (step S314). Then, returning to step S303, the same processing as described above is started.

  In this way, when the processing for all the lanes (lanes) of the road 101 is completed (step S313: Y), the unit area 141 in which erroneous detection occurs at a detection frequency equal to or higher than a predetermined threshold is detected as the erroneously detected steady occurrence registration area 214. (Step S315). Then, the motion trajectory registered in the list Err 216 is deleted from the tracking target (step S316), and the process ends (END).

  In the vehicle tracking system 100 according to the first embodiment of the present invention described above, it is possible to prevent the erroneous detection of determining that the vehicles 103 collide with each other in the same lane, and to accurately grasp the collision accident. There is. In addition, there is an effect that the accuracy of the erroneous detection steady occurrence region can be increased using the erroneous detection result.

  In the first embodiment described above, the erroneous detection trajectory discriminating unit 117 uses a condition for discriminating whether or not the two motion trajectories of interest completely intersect as an erroneous detection. It is not limited. It may be considered that a crossing has occurred when the two movement trajectories are close to each other to a certain degree and become an unnatural state as a normal inter-vehicle distance.

  In the first embodiment, the erroneous detection trajectory determination unit 117 determines whether or not the motion trajectory of interest is calculated based on the result of the erroneous detection using the equation (3). . The function F represented by the equation (3) may have any definition as long as it is a function that decreases as the input D decreases and decreases as the input S increases. Furthermore, although the erroneous detection steady occurrence region determination unit 118 uses the detection frequency distribution in each unit region in the first embodiment, the present invention is not limited to this. For example, the detection frequency of the vehicle 103 may be given as a probability by dividing by the time range used for the comparison of the movement trajectory.

  <Second Embodiment of the Invention>

  Next, a second embodiment of the present invention will be described.

  FIG. 18 shows a vehicle tracking system according to the second embodiment of the present invention. In FIG. 18, the same parts as those in FIG.

  A vehicle tracking system 100A according to the second embodiment includes a vehicle position detection sensor 102 and a vehicle tracking device 105A that inputs detection data 104 thereof.

  The vehicle tracking device 105A includes a CPU 111 and a control unit 113A including a memory 112A. The control unit 113A controls the tracking locus calculation unit 115, the erroneous detection locus determination unit 117A, and the erroneous detection locus deletion unit 119 via a bus 114 such as a data bus. By executing the control program stored in the memory 112A, the control unit 113 can realize at least a part of the tracking locus calculation unit 115, the erroneous detection locus determination unit 117A, and the erroneous detection locus deletion unit 119 by software.

FIG. 19 is a diagram for explaining the operation principle of the erroneous detection locus determination unit in the second embodiment. In FIG. 19, the same parts as those in FIG. 9 are denoted by the same reference numerals, and description thereof will be omitted as appropriate. In FIG. 19, candidate detection positions 171 ₁₃ , 171 ₂₃ , and 171 ₃₃ of the vehicle 103 are detected at the locations indicated by the crosses. Here, the candidate detection position 171 ₁₃ is at the position q ₁ of the first lane 151. The candidate detection position 171 ₃₃ is at the position q ₂ of the third lane 153. The candidate detection position 171 ₂₃ exists in the second lane 152, and the position is between the position q ₁ and the position q ₂ . First to third lanes 151 to 153 are lanes in which the vehicle travels in the direction indicated by arrow 145.

Assume that the candidate detection positions 171 ₁₃ , 171 ₂₃ , and 171 ₃₃ are erroneously detected in this order from the first lane 151 to the third lane 153 in the range between the positions q ₁ and q ₂ . In such a case, depending on the detection time order, a motion trajectory 401 as indicated by a broken line is obtained.

Now, it is assumed that the distance between the position q ₁ and the position q ₂ is, for example, about the length of one ordinary vehicle. In this case, a broken movement trajectory 401 shown in FIG. 19 means that the lane change is performed twice while the vehicle travels in the traveling direction at most by the vehicle length. Such a lane change is clearly not possible in reality. Therefore, the motion trajectory 401 can be determined to be a motion trajectory calculated based on the erroneous detection result.

The misdetection trajectory discriminating unit 117A shown in FIG. 18 is based on the principle as described above, and the shift between the tracking start position of the motion trajectory 401 and the end position at the current time (in the case of FIG. 19, the position q ₁ And the distance between the position q ₂ ). When the deviation value (distance) is smaller than a predetermined threshold value and the lane change history is two or more times (three or more in the number of lanes straddled), the motion trajectory 401 of interest is determined based on the erroneous detection result. It is determined that the motion trajectory is calculated.

  FIG. 20 shows main areas in the memory in the vehicle tracking apparatus shown in FIG. In the memory 112A, several areas are set in addition to the program storage area 211A for storing programs executed by the CPU 111. The detection data storage buffer 212 temporarily stores the detection data 104 sent from the vehicle position detection sensor 102 shown in FIG. The observation data storage area 213 stores data excluding the detection data 104 sent from the vehicle position detection sensor 102 that corresponds to the erroneous detection steady occurrence area described with reference to FIG. The erroneous detection steady occurrence registration area 214 registers an erroneous detection steady occurrence area. The list Err216 registers a misdetection trajectory of the motion trajectory.

  FIG. 21 corresponds to FIG. 17 and specifically shows the operation of the vehicle tracking system according to the second embodiment. The parts in FIG. 21 where the same processing as in FIG. 17 is performed are represented as the same steps, and description thereof will be omitted as appropriate. This will be described with reference to FIGS. 6 and 18 to 20.

  First, the control unit 113A in the vehicle tracking device 105A receives the detection data 104 at the current time T from the vehicle position detection sensor 102 and stores it in the detection data storage buffer 212 of the memory 112A. This detection data includes image data of a predetermined number of unit areas 141 obtained by dividing the tracking attention area 131 into a matrix as shown in FIG. The control unit 113A uses the M pieces of image data excluding the image data of the unit area 141 corresponding to the erroneous detection steady occurrence area 214 as observation data at the current time T indicating the vehicle detection result, and the observation data (not shown) of the memory 112. Save in the save area (step S301).

Next, the tracking trajectory calculation unit 115 calls the observation data obtained at time T-1 one stage before the current time T, and associates it with N _T-1 observation data at the current time T. . Then, the motion trajectory at time T is calculated (step S302). Let N _T be the number of calculated motion trajectories.

  Next, the misdetection trajectory discriminating unit 117A selects one motion trajectory of interest from unselected motion trajectories (step S501). Next, the erroneous detection trajectory discriminating unit 117A checks the lane change history of the selected motion trajectory, and checks whether there are two or more lane change histories (step S502). When there are two or more lane change histories (Y), a deviation D between the tracking start position of the motion trajectory and the trajectory end position at the current time T is calculated (step S503).

The misdetection locus determination unit 117A determines whether or not the calculated deviation D is smaller than a preset threshold value Thr ₂ (step S504). If it is small, as described above, this motion trajectory is due to erroneous detection. Therefore, in this case (step S504: Y), the erroneous detection trajectory determination unit 117A registers the motion trajectory in the list Err216 as a list of erroneous detection trajectories (step S311).

The misdetection trajectory discriminating unit 117A discriminates whether or not all the motion trajectories have been noticed after performing the registration in step S311 (step S505). If there is an unprocessed motion trajectory (N), the process returns to step S501 and the same process is performed. Incidentally, when the deviation D calculated in step S503 was the threshold Thr ₂ or more (step S504: N), it performs the determination in step S505 without performing the processing in step S311. Also, if it is determined in step S502 that there is no lane change history more than once (N), the determination in step S505 is performed without performing the process in step S311.

  When all the motion trajectories have been checked as described above (step S505: Y), the erroneous detection trajectory deletion unit 119 deletes the motion trajectory registered in the list Err 216 from the tracking target (step S316), and ends the processing. (End) In the second embodiment, the list GL215 that the tracking locus comparison unit 116 used in the first embodiment for registering the movement locus is not used, but the list GL215 is used in the same manner. Also good. In this case, the tracking trajectory comparison unit 116 may be provided for registration of the list GL 215, or the erroneous detection trajectory determination unit 117A may also use this function.

  In the vehicle tracking system 100A according to the second embodiment of the present invention described above, when a motion trajectory in which the vehicle 103 moves rapidly across a plurality of different lanes is generated, it is determined whether this is a false detection. There is an effect that can be done. In addition, there is an effect that the accuracy of the erroneous detection steady occurrence region can be increased using the erroneous detection result.

  <Third Embodiment of the Invention>

  Next, a third embodiment of the present invention will be described.

  FIG. 22 shows a vehicle tracking system according to the third embodiment of the present invention. In FIG. 22, the same parts as those in FIG.

  A vehicle tracking system 100B according to the third embodiment includes a vehicle position detection sensor 102 and a vehicle tracking device 105B that inputs detection data 104 thereof.

  The vehicle tracking device 105B includes a CPU 111 and a control unit 113B including a memory 112B. The control unit 113B controls the tracking trajectory calculation unit 115B and the erroneous detection steady occurrence region determination unit 118B via a bus 114 such as a data bus. By executing the control program stored in the memory 112B, the control unit 113B can realize at least a part of the tracking trajectory calculation unit 115B and the erroneous detection steady occurrence region determination unit 118B with software.

FIG. 23 is a diagram for explaining the operating principle of the tracking trajectory calculation unit in the third embodiment. FIG. 23 shows the detection state of the vehicle 103 at time T ′, time T ′ + 1, and time T ′ + 2 in the first lane 151 as an example. Vehicle 103 is traveling in the vehicle traveling direction indicated by the arrow 145, the motion trajectory 161 ₁₅ is extended through the tracking target area 131 in the traveling direction with time. The circles marked at the tips of the movement trajectories 161 ₁₅ shown in FIG. 23 are the tip positions of the trajectories at a time one step before the time points T ′, T ′ + 1, and T ′ + 2. is there. Further, x marks 601 ₁ , 601 ₂ , and 601 ₃ positioned at the rear of the vehicle 103 are candidate detection positions that are candidates for the vehicle 103 at the time T ′, the time T ′ + 1, and the time T ′ + 2, respectively.

In FIG. 23, another candidate detection position 171 ₁₅ of the vehicle 103 is detected in front of the vehicle 103. The candidate detection position 171 ₁₅ does not change at each time T ′, time T ′ + 1, and time T ′ + 2.

In the procedure described in the operation of the tracking trajectory calculation unit 115 described in the first embodiment, the positional relationship with each terminal (circle mark) of the motion trajectory 161 ₁₅ corresponding to the vehicle 103 illustrated in FIG. Consider the corresponding score ranks of candidate detection positions indicated by marks 601 ₁ , 601 ₂ , and 601 ₃ and candidate detection positions 171 ₁₅ . Then, from the equation (1), the candidate detection position score indicated by the X marks 601 ₁ , 601 ₂ , and 601 ₃ having a short distance from each end (◯ mark) of the movement locus 161 ₁₅ corresponding to the vehicle 103 is 1st. Become. Therefore, the score of the erroneous detection position as the candidate detection position 171 ₁₅ is 2nd or lower.

Thus, during the time range to associate the motion trajectory 161 ₁₅ and the vehicle detected position of the vehicle 103, the corresponding ranking always false detection position is position 2 or less. Therefore, by reflecting the occurrence frequency of the candidate detection position 171 ₁₅ in the unit area 141 shown in FIG. 6, it is possible to identify the unit area 141 in which erroneous detection occurs regularly.

In the erroneous detection steady occurrence region discriminating unit 118B, the tracking trajectory calculation unit 115B checks the occurrence frequency of the detection position where the correspondence score with the motion trajectory is second or lower for each unit region 141 set on the road in advance. When the occurrence frequency is equal to or higher than a predetermined threshold Thr ₃ , it is determined as an erroneously detected steady occurrence region.

FIG. 24 shows main areas in the memory in the vehicle tracking apparatus shown in FIG. In the memory 112B, several areas are set in addition to the program storage area 211B for storing programs executed by the CPU 111. The detection data storage buffer 212 temporarily stores the detection data 104 sent from the vehicle position detection sensor 102 shown in FIG. The observation data storage area 213 stores data excluding the detection data 104 sent from the vehicle position detection sensor 102 that corresponds to the erroneous detection steady occurrence area described with reference to FIG. Misdetection constant generating registration area 214B registers the threshold Thr ₃ or more unit areas.

  FIG. 25 corresponds to FIG. 17 and specifically shows the operation of the vehicle tracking system according to the third embodiment. In FIG. 24, the same processing as in FIG. 17 is represented as the same step, and description thereof will be omitted as appropriate. This will be described with reference to FIGS. 6 and 22 to 24.

  First, the control unit 113B in the vehicle tracking device 105B receives the detection data 104 at the current time T from the vehicle position detection sensor 102 and stores it in the detection data storage buffer 212 of the memory 112. This detection data includes image data of a predetermined number of unit areas 141 obtained by dividing the tracking attention area 131 into a matrix as shown in FIG. The control unit 113B uses the M pieces of image data excluding the image data of the unit area 141 corresponding to the erroneous detection steady occurrence area 214B as the observation data at the current time T indicating the vehicle detection result, and the observation data storage area of the memory 112 It is stored in 213 (step S301).

Next, the tracking trajectory calculation unit 115 calls the observation data obtained at time T-1 one stage before the current time T, and associates it with N _T-1 observation data at the current time T. . Then, the motion trajectory at time T is calculated (step S302). Let N _T be the number of calculated motion trajectories.

Next, the erroneous detection steady occurrence region discriminating unit 118B updates the occurrence frequency of the detection position where the corresponding score with the motion trajectory is second or lower for each unit region 141 set in advance on the road (step S701). Then, the unit area 141 having a detection frequency equal to or higher than the threshold Thr ₃ is registered in the erroneous detection steady occurrence registration area 214B (step S702), and the process ends (END).

  In the vehicle tracking system 100B according to the third embodiment of the present invention described above, when a plurality of candidate detection positions are detected in the lane where the end of the movement locus of the vehicle exists, the erroneous detection candidate detection positions are easily determined. There is an effect that it can be specified. In addition, there is an effect that the accuracy of the erroneous detection steady occurrence region can be increased using the erroneous detection result.

  <Fourth Embodiment of the Invention>

  FIG. 26 shows a vehicle tracking system according to the fourth embodiment of the present invention. In FIG. 26, the same reference numerals are given to the same portions as those in FIG.

  A vehicle tracking system 100C according to the fourth embodiment includes a vehicle position detection sensor 102 and a vehicle tracking device 105C that inputs detection data 104 thereof.

  The vehicle tracking device 105C includes a CPU 111 and a control unit 113C including a memory 112C. The control unit 113C includes a tracking trajectory calculation unit 115, a tracking start determination unit 801, a tracking trajectory comparison unit 116, an erroneous detection trajectory determination unit 117, an erroneous detection trajectory feature calculation unit 802, and an erroneous detection via a bus 114 such as a data bus. The steady occurrence region discriminating unit 118 and the erroneous detection locus deleting unit 119 are controlled. The control unit 113C executes the control program stored in the memory 112C, whereby a tracking locus calculation unit 115, a tracking start determination unit 801, a tracking locus comparison unit 116, an erroneous detection locus determination unit 117, and an erroneous detection locus feature calculation unit 802. In addition, at least a part of the erroneous detection steady occurrence region determination unit 118 and the erroneous detection locus deletion unit 119 can be realized by software.

  Here, the newly provided tracking start determination unit 801 calculates the feature amount of the motion trajectory calculated by the tracking trajectory calculation unit 115 for a motion trajectory that has only elapsed for a short time from the start of tracking. If it matches the erroneously detected trajectory feature database registered by the erroneously detected trajectory feature calculation unit 802, it is deleted from the motion trajectory to be tracked. The erroneous detection trajectory feature calculation unit 802 calculates the feature amount of the motion detection trajectory determined from the erroneous detection result by the erroneous detection trajectory determination unit 117. These are registered in the erroneous detection trajectory feature database.

  The feature amount calculated by the false detection trajectory feature calculation unit 802 is a motion trajectory curve such as a tracking start position, a detection position fluctuation range, a detection position fluctuation frequency, an average velocity calculated from the detection position, and an average acceleration. Any object may be used as long as it represents the feature of the shape.

  FIG. 27 shows main areas in the memory in the vehicle tracking apparatus shown in FIG. In the memory 112 </ b> C, several areas are set in addition to the program storage area 211 </ b> C for storing programs executed by the CPU 111. Compared to FIG. 16 in the first embodiment, a false detection locus feature database 811 is newly added to the memory 112C.

  Such a memory 112C need not be a single physical memory or storage device, and may be a combination of a plurality of storage devices and storage elements. The same applies to the memories 112, 112A, and 112B in the embodiments already described.

  FIG. 28 shows the process of the tracking start determination unit. First, the tracking start determination unit 801 selects an initial tracking trajectory with a small elapsed time from the start of tracking (step S901). Next, the tracking start determination unit 801 calculates a motion trajectory feature value f of the selected motion trajectory (step S902).

  FIG. 29 is for explaining the principle of calculation of the motion trajectory feature quantity f. In FIG. 29, the vertical axis represents the traveling direction of the vehicle 103 (see FIG. 26), and the horizontal axis represents the passage of time.

The tracking start position of the vehicle 103 is P _S, and the movement width of the detection position is v. It is assumed that the detection position of the vehicle 103 in the section from the time T ₁ to the time T ₂ varies as illustrated in the movement width v. The characteristic of the curve shape of the motion trajectory is the motion trajectory feature amount f such as the fluctuation frequency F of the detection position of the vehicle 103, the average speed V (MV) calculated from the detection position, or the average speed α (MV). Can be calculated as

Returning to FIG. 28, the description will be continued. After calculating the motion trajectory feature quantity f of the selected motion trajectory in step S902, in the next step S903, the similarity (Similarity) is calculated. For calculating the similarity, the motion trajectory feature amount calculated in step S302 and the i-th motion trajectory feature amount PS ′ (i), v ′ (i) registered in the erroneous detection trajectory feature database 811 are used. , F ′ (i), average speed V (MV) (i), and average speed α (MV) (i). The similarity (Similarity) is calculated by the following equation (5). Here, α _{1 to} α ₅ are parameters.

Similarity = α ₁ | PS (S) −PS ′ (i) | + α ₂ | V (S) −V ′ (i) | + α ₃ | F (S) −F ′ (i) | + α ₄ | V (MV ) (S) −V (MV) ′ (i) | + α ₅ | α (MV) (S) −α (MV) ′ (i) | (5)

In step S903, the (5) whether the value of the expression is present than smaller i is the threshold value Thr ₄ determination is made. If i that is smaller than the threshold Thr ₄ exists (Y), the process proceeds to step S904, the target motion trajectory is deleted from the tracking target, and the processing of the tracking start determination unit 801 is ended (END).

If there is no i in which the value of the expression (5) is smaller than the threshold Thr ₄ (step S903: N), the processing of the tracking start determination unit 801 ends without performing the processing of step S904 (end). ).

  FIG. 30 shows a state of processing of the erroneous detection locus feature calculation unit. First, the erroneous detection trajectory feature calculation unit 802 selects a trajectory determined as an exercise trajectory calculated based on the erroneous detection result (step S921). Next, the erroneous detection trajectory feature calculation unit 802 calculates the motion trajectory feature amount f of the selected motion trajectory. This is the same processing as step S902 in FIG.

  When the above processing is completed, the erroneous detection trajectory feature calculation unit 802 registers the motion trajectory feature amount calculated in step S922 in the erroneous detection trajectory feature database (DB) 811 (step S923), and ends the processing (end). ).

  In the vehicle tracking system 100C according to the fourth embodiment of the present invention described above, the feature amount of the motion trajectory that only passes for a short time from the start of tracking is calculated. I decided to delete it from the trajectory of the subject. Thereby, there is an effect that the load of the tracking process of the motion trajectory of the vehicle 103 can be reduced.

  In each of the embodiments described above, the unit region 141 determined as the erroneously detected steady occurrence region by the erroneously detected steady occurrence region determining unit 118 does not need to be fixed for a long time. For example, it may be determined whether or not a false detection steady occurrence region occurs every time a certain time elapses. At this time, the past discrimination situation may be taken into account, or weather conditions such as sunshine at the time of discrimination may be taken into account.

  Furthermore, the motion trajectory determined to be erroneously detected may be registered in the dictionary as an erroneous detection pattern, and may be used to determine whether or not it is a false detection when a similar motion trajectory appears. When registering in the dictionary, register the status of the tracking start position, detected position fluctuation range, detected position fluctuation frequency, average speed calculated from the detected position, average acceleration, etc. It is preferable to do. Thereby, the process of the similarity of the motion locus when the motion locus of the same pattern is generated can be performed more appropriately.

  As an example, it is determined whether or not there is a pattern match between the movement locus of interest and the movement locus registered in the dictionary. If the two patterns match, the tracking start position, detected position fluctuation range, detected position fluctuation frequency, average speed calculated from the detected position, and average acceleration are registered in the dictionary. Compared with each item. The various items registered in the dictionary are the tracking start position, the detection position fluctuation range, the detection position fluctuation frequency, the average speed calculated from the detection position, and the average acceleration of the movement locus of the erroneous detection pattern. And when it corresponds with either of these matters registered in the dictionary, you may perform the process which does not use the detection position of the attention vehicle as a tracking start position.

  If the similarity between the movement trajectory of the vehicle of interest and the movement trajectory of the false detection pattern registered in the dictionary is calculated and the similarity is high, what has been tracked as the movement trajectory of the vehicle 103 until now is the false detection. The possibility of being determined as a result increases. In such a case, it is also effective for the tracking process of the vehicle 103 to end the tracking of the motion trajectory and narrow down other candidate motion trajectories as the motion trajectory of the vehicle 103.

  It should be noted that the present invention described above can be regarded as the technical idea of the method as follows.

(First technical concept)
In a vehicle tracking method for discriminating between a false detection vehicle detection position and a true vehicle detection position that occur regularly and repeatedly at the same position, a shape characteristic of a motion trajectory calculated using the vehicle detection position, and a vehicle detection position A vehicle tracking method characterized in that the above-described discrimination is executed based on a feature amount calculated from an appearance distribution on a road surface.

(Second technical concept)
In the vehicle tracking method shown as the first technical idea, when one movement locus stays at a position within a certain range when the movement locus of two vehicles intersects on the same lane, the staying movement Using the amount of deviation between the tracking start position of the trajectory and the staying position as the motion trajectory shape feature, using the sum of the vehicle detection frequencies at the vehicle detection positions constituting the motion trajectory as the feature amount calculated from the vehicle detection position appearance distribution, A vehicle that performs erroneous detection determination on vehicle detection position time-series data corresponding to the staying movement locus based on a value calculated from a function calculated from the sum of the deviation amount and the detection frequency. Tracking method.

(Third technical idea)
In the vehicle tracking method shown as the second technical idea, when a function that increases monotonously as the discriminant function is used when the deviation amount described above decreases and the sum of the detection frequencies increases, the discriminant function value exceeds a threshold value A vehicle tracking method comprising: performing erroneous detection determination on vehicle detection position time-series data corresponding to the staying movement locus described above.

(Fourth technical idea)
In the vehicle tracking method shown as the second technical idea, when a function that monotonously decreases when the above-described deviation amount decreases and the sum of the above-described detection frequencies increases is used as a discriminant function, and the value of the discriminant function falls below a threshold value A vehicle tracking method comprising: performing erroneous detection determination on vehicle detection position time-series data corresponding to the staying movement locus described above.

(Fifth technical idea)
In the vehicle tracking method shown as the first technical idea, as the vehicle detection position appearance distribution, the road area is divided into a plurality of partial areas, and within the range of the count start time and count end time specified in advance, A vehicle tracking method comprising adding the appearance frequencies of vehicle detection positions.

(Sixth technical idea)
In the vehicle tracking method shown as the fifth technical concept, when the value of the vehicle detection position appearance distribution is equal to or greater than a predetermined threshold value, the corresponding partial area is determined as an area where erroneous detection of the vehicle position occurs regularly. A vehicle tracking method.

(Seventh technical idea)
In the vehicle tracking method shown as the first technical idea, the number of lanes straddling when the vehicle's motion trajectory travels a certain short distance in the traveling direction is used as the motion trajectory shape feature, and when the number of lanes is 3 or more, A vehicle tracking method, characterized in that time-series data of vehicle detection positions corresponding to the movement trajectory is determined as an erroneous detection pattern.

(Eighth technical idea)
In the vehicle tracking method shown as the first technical idea, when associating the already calculated motion trajectory calculated up to the current time with the vehicle detection position obtained at the current time, the road region is divided into a plurality of partial regions, A vehicle tracking method, wherein an appearance frequency value is added to a partial region corresponding to a vehicle position having a correspondence score of second or lower for use in association, and used as a feature amount calculated from a vehicle detection position appearance distribution.

(Ninth technical idea)
In the vehicle tracking method shown as the eighth technical idea, the reciprocal number is used for the distance between the end position of the already calculated motion trajectory and the current time vehicle detection position as the corresponding score, and the score ranking in descending order of the corresponding score. A vehicle tracking method, characterized by:

(10th technical idea)
According to the vehicle tracking method shown as the eighth technical concept, the above-described partial region in which the value of the vehicle detection position appearance distribution is greater than or equal to a threshold value is determined as a region where erroneous detection of the vehicle occurs regularly. Vehicle tracking method.

(Eleventh technical idea)
In the vehicle tracking method shown as one of the first to tenth technical ideas, it is determined again whether or not it is an erroneous detection occurrence region every time a certain time has elapsed since it was determined as an erroneous detection occurrence region. Vehicle tracking method.

(Twelfth technical idea)
A vehicle tracking method according to any one of the first to eleventh technical ideas, wherein a motion trajectory determined to be erroneously detected is registered in a dictionary as an erroneously detected pattern.

(13th technical idea)
In the vehicle tracking method shown as the twelfth technical concept, the tracking start position, detected position fluctuation range, detected position fluctuation frequency, average speed calculated from the detected position, and average acceleration are dictionary The vehicle tracking method characterized by registering in.

(14th technical idea)
In the vehicle tracking method shown as the twelfth or thirteenth technical idea, the similarity between the vehicle trajectory of interest and the trajectory of the false detection pattern registered in the dictionary is calculated, and when the similarity is high, the vehicle trajectory of interest is A vehicle tracking method characterized in that the subsequent tracking is terminated after determining the erroneous detection pattern.

(15th technical idea)
In the vehicle tracking method shown as the thirteenth technical idea and the fourteenth technical idea, the tracking start position, detected position fluctuation range, detected position fluctuation frequency, average speed calculated from the detected position, average If any of the accelerations is the same as the tracking start position, the detection position fluctuation range, the detection position fluctuation frequency, the average speed calculated from the detection position, and the average acceleration of the motion locus of the false detection pattern registered in the dictionary, the attention A vehicle tracking method characterized by not using a vehicle position as a tracking start position.

10, 100, 100A, 100B, 100C Vehicle tracking system 11 Sensor 12 Movement trajectory calculation means 13 Error detection ratio registration means 14 Vehicle tracking analysis means 20 Vehicle tracking method 21 Sensor detection step 22 Motion trajectory calculation step 23 Error detection ratio registration step 24 Vehicle tracking analysis step 30 Vehicle tracking program 31 Sensor detection process 32 Motion trajectory calculation process 33 False detection rate registration process 34 Vehicle tracking analysis process 101 Road 102 Vehicle position detection sensor 103 Vehicle 104 Detection data 105, 105A, 105B, 105C Vehicle tracking device 111 CPU
112, 112A, 112B, 112C Memory 113, 113A, 113B, 113C Control unit 115, 115B Tracking trajectory calculation unit 116 Tracking trajectory comparison unit 117, 117A Misdetection trajectory discrimination unit 118, 118B False detection steady occurrence region discrimination unit 119 False detection Trajectory deletion unit 131 Tracking target area 141 Unit area 151 First lane 152 Second lane 153 Third lane 161, 191 and 192 Motion trajectory 171 Candidate detection positions 211 and 211A Program storage area 212 Detection data storage buffer 213 Observation data storage area 214 Error detection steady occurrence registration area 215 List GL
216 List Err
801 Tracking start determination unit 802 Error detection trajectory feature calculation unit 811 Error detection trajectory feature database

Claims (16)

A sensor that sequentially detects the physical state on the road in each unit area obtained by dividing a predetermined range on the road into a plurality of areas over time;
When the physical state of any of the unit areas on the road detected by this sensor is a specific physical state indicating the possibility of the presence of a vehicle, An exercise trajectory calculating means for calculating an exercise trajectory;
An erroneous detection rate registration means for registering a rate at which the specific physical state is constantly generated for each unit region;
Vehicle tracking analysis means for analyzing the travel route of the vehicle using the motion trajectory calculated by the motion trajectory calculating means and the false detection rate of the vehicle for each unit area registered in the erroneous detection rate registering means. A vehicle tracking system characterized by that. When the motion trajectory calculated by the motion trajectory calculating means intersects in the same lane on the road, a staying discriminating means for discriminating whether one of the motion trajectories is staying at a place where the intersection occurs,
When this staying discriminating means discriminates staying, a deviation in the vehicle traveling direction between the start position of the motion trajectory and the staying position in the region of the predetermined range of the staying motion trajectory Deviation amount determining means for determining the amount;
And a staying presence / absence determining means for determining whether or not the vehicle is staying depending on whether or not the amount of deviation determined by the deviation amount determining means exceeds the range of the detection error of the position in the traveling direction of the vehicle by the sensor. The vehicle tracking system according to claim 1.   When the deviation amount does not exceed the detection error range of the position of the vehicle in the traveling direction by the sensor, one of the motion trajectories indicating stay is discriminated as a false detection motion trajectory. The vehicle tracking system according to claim 2, further comprising means.   When the deviation exceeds the range of the detection error of the position of the vehicle in the traveling direction by the sensor, it comprises a collision detection unit for staying that determines that two vehicles have collided in the same lane. The vehicle tracking system according to claim 2. Lane stride amount determining means for determining whether the motion trajectory calculated by the motion trajectory calculating means has moved two or more lanes on the road;
When it is determined that two or more lanes on the road have been moved by the lane straddling amount determination means, it is necessary to determine whether the time required for this is within a predetermined time required for movement between lanes of the vehicle. Time discrimination means;
3. A multi-lane movement erroneous detection determination means for determining that the movement trajectory is an erroneous detection movement trajectory when the required time determination means determines that it is within the predetermined time. The vehicle tracking system described.   The false detection rate registered by the false detection rate registering unit is added to each of the unit areas corresponding to the motion trajectory that the false detection detection unit during dwelling or the false detection determination unit during the movement of multiple lanes determined to be erroneous detection. 6. The vehicle tracking system according to claim 3, further comprising an error detection time addition means.   The movement trajectory calculating means calculates a plurality of candidate detection positions for the vehicle based on the current output of the sensor, and the vehicle movement is calculated based on the output of the sensor up to the previous time in the traveling direction of the vehicle. 2. The vehicle tracking according to claim 1, wherein the vehicle tracking is a means for ranking the corresponding scores in order from the shortest distance from the end of the trajectory and connecting the motion trajectory from the end to the candidate detection position of the first score. system.   The error detection addition means for adding the error detection ratio registered by the error detection ratio registration means for each of the unit regions corresponding to the candidate detection positions of the second and subsequent scores. The vehicle tracking system according to claim 3 and claim 7.   Each of the unit regions in which the false detection rate registered by the false detection rate registering unit exceeds a preset threshold is registered as a false detection steady occurrence region in which erroneous detection of the vehicle occurs regularly. The vehicle tracking system according to claim 1.   2. The vehicle tracking system according to claim 1, wherein the registration content of the erroneous detection rate registration means is updated at a predetermined timing.   The vehicle tracking system according to claim 1, further comprising a dictionary that registers a motion trajectory determined to be erroneously detected as an erroneously detected pattern.   In the dictionary, any one of the tracking start position of the motion trajectory determined to be erroneously detected, the detected position fluctuation range, the frequency of the detected position fluctuation, the average speed calculated from the detected position, and the average acceleration is set as the false detection pattern. The vehicle tracking system according to claim 11, wherein the vehicle tracking system is registered as an item attached to.   Similarity calculation means for calculating the similarity between any of the false detection patterns registered in the dictionary and the candidate motion trajectory pattern, and the similarity calculated by the similarity calculation means is higher than a predetermined value. 12. The vehicle tracking system according to claim 11, further comprising tracking vehicle means for ending tracking of the motion trajectory after determining that the candidate motion trajectory pattern is a false detection pattern.   A pattern comparison unit that compares whether any of the false detection patterns registered in the dictionary matches the pattern of the candidate motion trajectory, and when the pattern comparison unit detects a pattern match, the candidate motion trajectory Any of the tracking start position of the motion trajectory, the detected position fluctuation range, the frequency of the detected position fluctuation, the average velocity calculated from the detected position, and the average acceleration, which accompany the pattern of And a start of tracking that does not adopt the detection position of the candidate motion trajectory as a tracking start position when the accompanying matter comparison means determines whether or not any of the accompanying matters is matched by the accompanying matter comparison means. The vehicle tracking system according to claim 11, further comprising position non-employment means. A sensor detection step of sequentially detecting the physical state on the road in each unit area obtained by dividing a predetermined range of the predetermined area on the road with time using a sensor;
When the physical state of any one of the unit areas detected by the sensor detection step is a specific physical state indicating the possibility of the presence of the vehicle, a motion trajectory as a movement path of the temporal position of these unit areas A motion trajectory calculation step for calculating
An erroneous detection rate registration step of registering a rate at which the specific physical state is constantly generated for each unit region;
A vehicle tracking analysis step of analyzing a travel route of the vehicle using the motion trajectory calculated in the motion trajectory calculation step and a vehicle false detection rate for each unit area registered in the error detection rate registration step. A vehicle tracking method. On the computer,
A sensor detection process for sequentially detecting the physical state on the road in each unit area obtained by dividing a predetermined range of the predetermined area on the road with time using a sensor;
When any physical state of the unit area detected by the sensor detection process is a specific physical state indicating the possibility of the presence of the vehicle, a motion trajectory as a movement path of the temporal position of these unit areas Motion trajectory calculation processing for calculating
A false detection rate registration process for registering a rate at which the specific physical state occurs regularly for each unit region;
A vehicle tracking analysis process for analyzing a travel route of the vehicle using the motion trajectory calculated by the motion trajectory calculation process and a vehicle false detection ratio for each unit area registered in the error detection ratio registration process is executed. A vehicle tracking program characterized by that.

Images