Tracking system, tracking device, tracking method, tracking program

Abstract

To provide a tracking system or the like that can more accurately track a target moving object.SOLUTION: A tracking system includes a processor, and tracks a target moving object detected by an external sensor system. The processor is configured to acquire detection information that is information related to a moving object detected in a current cycle. The processor is configured to associate tracking information that is information related to a moving object tracked before a previous cycle with detection information estimated to be related to the same moving object as the tracking information. The processor is configured to update the tracking information in response to the associated detection information in the current cycle. The processor is configured to generate duplicate tracking information where tracking information that satisfies an error condition related to possibility that association between the tracking information and the detection information is incorrect is duplicated in a state before updating, as tracking information.SELECTED DRAWING: Figure 5

Description

Technical Field

[0001] The present disclosure relates to a tracking technique for tracking a target moving body.

Background Art

[0002] Patent Document 1 discloses an image analysis apparatus that tracks a subject between a plurality of frames of an image. This image analysis apparatus associates the same tracking target based on feature amounts between the current image frame and past image frames. When the tracking target does not intersect with other subjects, the image analysis apparatus updates the feature amount of the tracking target, and when it intersects, the image analysis apparatus does not update the feature amount of the tracking target.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In tracking a moving body, part or all of the target moving body to be tracked may be hidden by other moving bodies, resulting in the target moving body not being detected. If another moving body similar to the target moving body appears while the target moving body is not being detected, there is a risk that the detection result of this other moving body will be associated with the past data of the target moving body. As a result, accurate tracking of the target moving body may become impossible.

[0005] An object of the present disclosure is to provide a tracking system capable of more accurately tracking a target moving body. Another object of the present disclosure is to provide a tracking device capable of more accurately tracking a target moving body. Still another object of the present disclosure is to provide a tracking method capable of more accurately tracking a target moving body. Yet another object of the present disclosure is to provide a tracking program capable of more accurately tracking a target moving body.

Means for Solving the Problems

[0006] Hereinafter, the technical means of the present disclosure for solving the problems will be described. Note that the reference numerals in parentheses described in the claims and this column indicate the correspondence with the specific means described in the embodiments to be described in detail later, and do not limit the technical scope of the present disclosure.

[0007] A first aspect of the present disclosure is a tracking system having a processor (102) and tracking a target moving body (3) detected by an external sensor system (10), The processor obtains detection information, which is information about the target moving body detected in the current cycle, associates tracking information, which is information about the target moving body tracked before the previous cycle, with detection information estimated to be about the same target moving body as the tracking information, updates the tracking information according to the detection information associated in the current cycle, generates replicated tracking information in which tracking information for which an error condition regarding the possibility that the association between the tracking information and the detection information is incorrect holds is replicated in the state before the update, as the tracking information, and is configured to execute the above.

[0008] A second aspect of the present disclosure is a tracking device having a processor (102), configured to be mountable on a host moving body (1), and tracking a target moving body (3) detected by an external sensor system (10), The processor Obtaining detection information, which is information about the target moving object detected in the current cycle, Associating tracking information, which is information about the target moving object tracked before the previous cycle, with detection information estimated to be about the same object as the tracking information, Updating the tracking information according to the detection information associated in the current cycle, Generating, as tracking information, replicated tracking information in which the tracking information for which an error condition regarding the possibility that the association between the tracking information and the detection information is incorrect holds is replicated in the state before update, and being configured to execute.

[0009] A third aspect of the present disclosure is a tracking method executed by a processor (102) for tracking a target moving object (3) detected by an external sensor system (10), obtaining detection information, which is information about the target moving object detected in the current cycle, associating tracking information, which is information about the target moving object tracked before the previous cycle, with detection information estimated to be about the same object as the tracking information, updating the tracking information according to the detection information associated in the current cycle, generating, as tracking information, replicated tracking information in which the tracking information for which an error condition regarding the possibility that the association between the tracking information and the detection information is incorrect holds is replicated in the state before update, and including.

[0010] A fourth aspect of the present disclosure is a tracking program stored in a storage medium (101) and including instructions for causing a processor (102) to execute to track a target moving object (3) detected by an external sensor system (10), the instructions causing to obtain detection information, which is information about the target moving object detected in the current cycle, Associating tracking information, which is information about a target moving object tracked before a previous cycle, with detection information estimated to be about the same object as the tracking information. Updating the tracking information according to the detection information associated therewith in the current cycle. Generating replicated tracking information in which tracking information for which an error condition regarding the possibility that the association between the tracking information and the detection information is incorrect holds is replicated in the state before update, as the tracking information. Including.

[0011] According to these first to fourth aspects, replicated tracking information in which tracking information for which an error condition regarding the possibility that the association between the tracking information and the detection information is incorrect holds is replicated in the state before update is generated as the tracking information. Therefore, even when the original tracking information is updated in association with detection information of a target moving object different from the original target moving object that has been tracked until then, in a cycle after the next cycle, the original target moving object can be associated with the replicated tracking information replicated in the state before update. Accordingly, the target moving object can be tracked more accurately.

Brief Description of the Drawings

[0012]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Mode for Carrying Out the Invention

[0013] Hereinafter, a plurality of embodiments of the present disclosure will be described with reference to the drawings. In addition, in each embodiment, the same reference numerals may be given to corresponding components, and redundant explanations may be omitted. Further, when only a part of the configuration is described in each embodiment, the configuration of other embodiments described previously can be applied to other parts of the said configuration. Furthermore, not only the combinations of configurations explicitly shown in the description of each embodiment, but also the configurations of a plurality of embodiments can be partially combined with each other as long as there is no problem with the combination.

[0014] (First Embodiment) The tracking system 100 of the first embodiment shown in FIG. 1 tracks a target moving body 3 as a moving body around the host vehicle 1 shown in FIG. 1 as a host moving body. From the perspective centered on the host vehicle 1, it can be said that the host vehicle 1 is an ego-vehicle. The host vehicle 1 is a moving body such as an automobile that can travel on a road in a state where a passenger is on board. From the perspective centered on the host vehicle 1, it can be said that the target moving body 3 is another road user. The target moving body 3 includes at least one type among, for example, automobiles, trucks, motorcycles, bicycles, autonomous driving robots, pedestrians, and animals.

[0015] In the host vehicle 1, an automatic driving mode is provided that is classified according to the degree of manual intervention of the passenger in the driving task. The automatic driving mode may be realized by autonomous driving control in which the system during operation executes all driving tasks, such as conditional driving automation, highly automated driving, or fully automated driving. The automatic driving mode may be realized by advanced driving assistance control in which the passenger executes some or all of the driving tasks, such as driving assistance or partial driving automation. The automatic driving mode may be realized by either one, a combination, or a switch between the autonomous driving control and the advanced driving assistance control.

[0016] The host vehicle 1 is equipped with the external sensor system 10 shown in FIG. 1. The external sensor system 10 acquires external information available to the tracking system 100 from the external environment that is the periphery of the host vehicle 1. The external sensor system 10 may acquire external information by detecting targets existing in the external environment of the host vehicle 1. The external sensor system 10 of the target detection type is at least one type among, for example, cameras, LiDAR (Light Detection and Ranging / Laser Imaging Detection and Ranging), radars, and sonars.

[0017] The tracking system 100 is connected to the external sensor system 10 via at least one of, for example, a LAN (Local Area Network) line, a wire harness, an internal bus, and a wireless communication line. The tracking system 100 is configured to include at least one dedicated computer.

[0018] The dedicated computer constituting the tracking system 100 may be an operation control ECU (Electronic Control Unit) that controls the operation of the host vehicle 1. The dedicated computer constituting the tracking system 100 may be a navigation ECU that navigates the traveling route of the host vehicle 1. The dedicated computer constituting the tracking system 100 may be a locator ECU that estimates the self-state quantity of the host vehicle 1. The dedicated computer constituting the tracking system 100 may be an actuator ECU that controls the traveling actuator of the host vehicle 1. The dedicated computer constituting the tracking system 100 may be an HCU (HMI (Human Machine Interface) Control Unit) that controls information presentation in the host vehicle 1. The dedicated computer constituting the tracking system 100 may be a computer other than the host vehicle 1 that constitutes, for example, an external center or a mobile terminal capable of communicating with the host vehicle 1.

[0019] The dedicated computer constituting the tracking system 100 may be an integrated ECU (Electronic Control Unit) that integrates the operation control of the host vehicle 1. The dedicated computer constituting the tracking system 100 may be a judgment ECU that judges the operation tasks in the operation control of the host vehicle 1. The dedicated computer constituting the tracking system 100 may be a monitoring ECU that monitors the operation control of the host vehicle 1. The dedicated computer constituting the tracking system 100 may be an evaluation ECU that evaluates the operation control of the host vehicle 1.

[0020] The dedicated computer that constitutes the tracking system 100 has at least one memory 101 and one processor 102. The memory 101 is at least one type of non-transitory tangible storage medium, such as a semiconductor memory, a magnetic medium, and an optical medium, that non-temporarily stores a computer-readable program, data, and the like. Here, storage may be an accumulation in which data is retained even when the host vehicle 1 is powered off, or it may be a temporary storage in which data is erased when the host vehicle 1 is powered off. The processor 102 includes at least one type, such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a RISC (Reduced Instruction Set Computer)-CPU, a DFP (Data Flow Processor), and a GSP (Graph Streaming Processor), as a core.

[0021] In the tracking system 100, the processor 102 executes a plurality of instructions included in the tracking program stored in the memory 101 to track the target moving body 3 around the host vehicle 1. As a result, the tracking system 100 constructs a plurality of functional blocks for tracking the target moving body 3 around the host vehicle 1. The plurality of functional blocks constructed in the tracking system 100 include a detection block 110, a correlation block 120, a track management block 130, and an update block 140, as shown in FIG. 2.

[0022] The detection block 110 detects the target moving body 3 from the external information by the external sensor system 10. The detection block 110 executes this detection process for each tracking cycle to obtain detection information, which is information about the target moving body 3. For example, the detection information includes the state quantity and the feature quantity of the target moving body 3.

[0023] The association block 120 associates a track, which is information about the target moving body 3 tracked before the previous cycle of the tracking cycle, with detection information estimated to be about the same target moving body 3 as the track. The association block 120 performs the association based on the similarity described later.

[0024] Here, the track is information about the target moving body 3 tracked before the previous cycle. For example, the track at least includes state quantities and feature quantities related to the target moving body 3. These state quantities and feature quantities are the state quantities and feature quantities in the last cycle during at least the period of the tracking state. The track may include the time-series state quantities and feature quantities before the cycle. An ID is individually assigned to the track as identification information for distinguishing it from other tracks. The track is an example of "tracking information".

[0025] Regarding the above track, the track management block 130 sets a state related to the tracking process. Specifically, as shown in FIG. 3, the track management block 130 sets one of the tracking state, lost state, and deleted state for the track. The tracking state is set for a track in which there is at least detection information associated in the previous cycle and the update has been executed, that is, a track for which the tracking of the corresponding target moving body 3 continues. The lost state is set for a track in which there is no detection information associated in at least the previous cycle and the update is suspended, that is, a track for which the tracking of the corresponding target moving body 3 has been interrupted. The deleted state is set for a track in which the lost state has continued for a predetermined number of cycles, that is, a track for which the corresponding target moving body 3 is considered to have completely disappeared from the detection range of the external sensor system 10. Note that the data of the track in the deleted state is deleted from a storage medium such as the memory 101 that stores the track. That is, the process of setting the deleted state for the track is substantially equivalent to the process of deleting the track.

[0026] The management block 130 generates a replicated track that is a copy of a track in the state before update (described later), for which an error condition regarding the possibility of incorrect association between the track and the detection information holds. The management block 130 generates the replicated track in a lost state. When there is detection information associated with the replicated track in a subsequent cycle, the management block 130 resumes tracking the target moving body 3 corresponding to the detection information with the replicated track, and assigns a different ID to the original track of the replication.

[0027] The update block 140 updates the track according to the detection information for the track associated with the detection information. Specifically, the update block 140 updates the state quantity and the feature quantity of the track.

[0028] The cooperation of the blocks 110, 120, 130, and 140 described so far will be used to explain the flow of the tracking method (hereinafter referred to as the tracking flow) by which the tracking system 100 tracks the target moving body 3 according to FIGS. 4 and 5 below. This processing flow is repeatedly executed while the host vehicle 1 is running. Each "S" in this processing flow means a plurality of steps executed by a plurality of instructions included in the tracking program.

[0029] First, in S100 of FIG. 4, the detection block 110 detects the target moving body 3 from external information. The detection block 110 may detect the target moving body 3 from the image data of the camera. Alternatively, the detection block 110 may detect the target moving body 3 from the point cloud data of the LiDAR or the reflected wave data of the radar. Alternatively, the detection block 110 may detect the target moving body 3 from the fusion data from at least two of the above-described sensing devices. The detection block 110 inputs the external information to a learned model such as a CNN (Convolutional Neural Network) that has been learned to perform object detection that outputs a detection result of the target moving body 3 for such input of external information, thereby detecting the target moving body 3. The state quantity (detection state quantity) such as the position and shape of the target moving body 3 is detected based on, for example, a bounding box or the like.

[0030] Next, in S150, the detection block 110 extracts feature quantities for the detected target moving body 3. In the case of image data, the detection block 110 extracts feature quantities by methods such as SIFT (Scale-Invariant Feature Transform) and HOG (Histograms of Oriented Gradients). The feature quantity here is a multi-dimensional quantity composed of a combination of a plurality of numerical values. The detection block 110 inputs the external information to a learned model such as a CNN that has been learned to output a feature quantity for the input of external information about the target moving body 3, thereby extracting the feature quantity.

[0031] Through the above steps, the detection block 110 obtains detection information, which is information about the target moving body 3 in the current cycle. As described above, the detection information includes at least the detection state quantity and the feature quantity.

[0032] In the subsequent S200, the association block 120 predicts the state quantity in the current cycle for each track.

[0033] The state quantity includes at least, for example, position and shape. This predicted state quantity may be referred to as the predicted state quantity hereinafter. For example, the association block 120 calculates the predicted state quantity by executing the prediction step of the Kalman filter.

[0034] Then, in S250, the association block 120 associates the track with the detection information estimated to be related to the same object as the track. The association block 120 performs the association based on the similarity between the detection information and the track. For example, the association block 120 calculates the similarity based on the state quantities and feature quantities of the detection information and the track respectively. For example, the similarity is a value correlated with the difference between the state quantities of the detection information and the track and the difference between the feature quantities. Then, the association block 120 associates the combination of the detection information and the track for which the similarity reaches the association range. The association range is, for example, a range where the similarity is equal to or greater than a predetermined first threshold value or greater than the first threshold value.

[0035] However, when the track is a duplicate track described later, the association block 120 changes the association range from the range in the case of a normal track. Specifically, when the similarity between the duplicate track and the detection information is greater than the similarity between the original track associated in the cycle in which the duplicate track was generated and the detection information, the association block 120 associates the duplicate track and the detection information. In other words, when the track is a duplicate track, the association block 120 makes the association range higher than the similarity between the original track associated in the cycle in which the duplicate track was generated and the detection information.

[0036] When there are a plurality of tracks and detection information, the association block 120 calculates the similarity for all combinations of the track and the detection information. The association block 120 selects a combination pattern in which the total sum of each similarity is the largest from among the combinations for which the similarity reaches the association range, and associates the track with the detection information.

[0037] In the subsequent S300, a track state update process is executed. Details of the track state update process will be described with reference to FIG. 5.

[0038] First, in S301, the track management block 130 determines the state of any one of the tracks for which the state update process is not completed. Specifically, the track management block 130 determines whether the target track is in a tracked state or a lost state. If it is determined that the track is in a tracked state, this flow proceeds to S302.

[0039] In S302, the track management block 130 determines whether there is detection information associated with the track in the tracked state. If it is determined that there is associated detection information, this flow proceeds to S303.

[0040] In S303, the track management block 130 determines whether an error condition regarding the possibility that the association between the track in the tracked state and the detection information associated with the track is incorrect is satisfied. The error condition is a condition that is satisfied when the possibility that the association between the track and the detection information is incorrect is at an unacceptable level.

[0041] Specifically, the track management block 130 determines whether the similarity between the track and the detection information associated with the track is within an error range. The error range is a similarity range whose upper limit is defined to be higher than the lower limit of the association range. Specifically, the error range is a similarity range that is equal to or lower than a second threshold value higher than a first threshold value or equal to or higher than the second threshold value. Determining that the similarity is within the error range is equivalent to determining that the error condition is satisfied.

[0042] When it is determined that the similarity is within the error range, this flow proceeds to S304. In S304, the track management block 130 generates a replicated track, which is a track obtained by replicating the track for which the error condition holds. This replicated track is replicated from the track in the state before the update of the feature amount described later. At this time, the track management block 130 defines the state of the replicated track as the lost state. That is, in the cycle in which the replicated track is generated, the track management block 130 does not perform the association with the detection information of the replicated track. As a result, the update of the state amount and the feature amount in S314 and S315 described later is postponed for the replicated track in the generated cycle. The replicated track is an example of "replicated tracking information".

[0043] Incidentally, hereinafter, the original track of the replicated track may be referred to as the original track. After the step of S304, this flow proceeds to S314 described later.

[0044] On the other hand, when it is determined in S303 that the similarity is outside the error range, this flow proceeds to S314 while skipping the process of S304. In other words, when it is determined that the similarity is outside the error range, the generation process of the replicated track is interrupted.

[0045] Incidentally, when it is determined in S302 that there is no detection information associated with the track in the tracking state, this flow proceeds to S305. In S305, the track management block 130 changes the state of the track from the tracking state to the lost state. After the process of S305, this subroutine process ends, and the flow of FIG. 4 proceeds to S350.

[0046] Also, when it is determined in S301 that the track is in the lost state, this flow proceeds to S306. In S306, similar to S302, the track management block 130 determines whether there is detection information associated with the track in the tracking state. When it is determined that there is associated detection information, this flow proceeds to S307.

[0047] In S307, the track management block 130 determines whether the track for which associated detection information has been determined has been a replicated track generated before the previous cycle. If it is determined to be a replicated track, this flow proceeds to S308.

[0048] In S308, the track management block 130 changes the state of the replicated track in the lost state to the tracking state. In addition, the track management block 130 changes the ID given to the original track, which is the replication source of the replicated track, up to the previous cycle, to a new ID. In other words, the track management block 130 defines that the original track is for information regarding a different target moving body 3 than before. After the process of S308, this flow proceeds to S314 described later.

[0049] On the other hand, if it is determined in S307 that the track for which associated detection information has been determined is not a replicated track, this flow proceeds to S309. In S309, the track management block 130 changes the state of the track from the lost state to the tracking state.

[0050] Subsequently, in S310, the track management block 130 determines whether the degree of similarity with the associated detection information for the track is within the error range. If it is determined to be within the error range, this flow proceeds to S311. In S311, the track management block 130 generates a replicated track of the track in the lost state, similar to S304.

[0051] Also, if it is determined in S306 that there is no associated detection information, this flow proceeds to S312. In S312, the track management block 130 determines whether the lost period, which is the period defined as the lost state for the track without associated detection information, is outside the allowable period. The allowable period is, for example, the period after which a predetermined number of cycles have elapsed since the lost state.

[0052] When it is determined in S312 that the lost period is outside the allowable period, this flow proceeds to S313. In S313, the track management block 130 changes the state of the track from the lost state to the deleted state. In other words, the track management block 130 deletes the track for which the allowable period has elapsed in the lost state. After the process of S313, this subroutine process ends, and the flow of FIG. 4 proceeds to S350. Also, when it is determined in S312 that it is within the predetermined period, the process of S313 is skipped and the process of this subroutine ends.

[0053] After the negative determination process in S303, the process of S304, the process of S308, the process of S311, or the negative determination process in S310, this flow proceeds to S314. In S314, the update block 140 updates the state quantity of the track based on the associated detection information. Specifically, the update block 140 estimates the state quantity (estimated state quantity) in the current cycle based on the predicted state quantity of the track and the detected state quantity based on the detection information associated with the track. Specifically, the update block 140 estimates the estimated state quantity by executing the filtering step of the Kalman filter. The update block 140 sets the estimated state quantity as the state quantity of the track in the current cycle.

[0054] In the subsequent S315, the update block 140 updates the feature quantity of the track based on the associated detection information. The update block 140 updates the feature quantity of the track in the previous cycle with the feature quantity of the detection information acquired in S150 at a prescribed update rate.

[0055] Furthermore, in S350, the track management block 130 determines whether the track state update process has been completed for all tracks. If it is determined that the track state update process has not been completed for all tracks, the process of S300 is executed for the tracks for which the process is not completed. On the other hand, if it is determined that the track state update process has been completed for all tracks, this flow proceeds to S400. In S400, the track management block 130 generates a new track corresponding to the detection information that has not been associated with the existing tracks.

[0056] Next, a comparison between the operation example by the tracking process of the comparative example and the operation example by the tracking process of the present embodiment will be described with reference to FIGS. 6 to 11. In the following, an example of tracking a plurality of persons as the target moving body 3 at times T1 to T4 will be described. Specifically, as shown in FIGS. 6 and 9, time T1 is a scene in which persons A and B are detected. Time T2 is a scene in which person A hides behind person B and only person B is detected. Time T3 is a scene in which person A is hidden by person B and a new person C is detected near the location where person A existed at time T1. Time T4 is a scene in which person A appears from behind person B and becomes detectable, and all of persons A, B, and C are detected. It is assumed that person A has been tracked by being associated with track α and person B has been tracked by being associated with track β since before time T1. Also, in each operation example, the association range is set to a range where the similarity is 0.5 or more, and in the operation example of the present embodiment, the error range is set to a range where the similarity is 0.5 or more and less than 0.6.

[0057] First, the tracking example by the tracking process of the comparative example will be described. In this tracking process of the comparative example, it is assumed that no duplicate track is generated even if the similarity of the associated track and detection information is within the error range.

[0058] At time T1, for the combination of the detection information of each person and each track, the similarity is calculated as shown in the top table of FIG. 7. As a result, the association is executed so that the similarity reaches the association range and the sum thereof is maximized. The detection information of person A is associated with track α, and the detection information of person B is associated with track β. Track α and track β are updated according to the associated detection information, respectively.

[0059] At time T2, the detection information of person B is associated with track β, which has a higher similarity than track α. Since person A is not detected and there is no detection information associated with track α, track α is defined as being in a lost state.

[0060] At time T3, person A is not detected, and the similarity between track α and person C is 0.5 or more. Therefore, as a result of the association so that the sum of the similarities is maximized, person C is associated with track α and person B is associated with track β. In this case, track α is updated with the feature amount of person C. Therefore, track α is in a state including the information of person C.

[0061] At time T4, person A is detected again. However, due to the association result at time T3, the information of person C is included in track α, and the similarity with track α is higher for person C than for person A. Therefore, as a result of the association so that the sum of the similarities is maximized, person C is associated with track α and person B is associated with track β. Since there is no track associated with person A, a newly generated track γ is associated.

[0062] As described above, in the tracking process of the comparative example, as shown in FIGS. 6 and 8, when person C appears while person A is not detected at time T3, track α is associated with the detection information of person C, which is different from person A who should originally be associated with track α at time T3. Then, at time T4 as well, when person A is redetected, track α, which should originally be associated with the detection information of person A, is associated with the detection information of person C.

[0063] Next, a tracking example by the tracking process of this embodiment will be described. The processes at time T1 and time T2 are substantially the same as those of the comparative example.

[0064] At time T3, person C is associated with track α, and person B is associated with track β. However, since the similarity between track α and person C is within the error range, track α in the lost state is replicated.

[0065] At time T4, as a result of the association such that the total similarity is maximized, since the similarity between the replicated track α and person A is greater than the similarity between track α and person C at time T3, person A is associated with the replicated track α (see FIG. 10). The original track α has its ID changed to a new ID. As a result, the original track α is changed to a new track γ.

[0066] That is, as shown in FIGS. 9 and 11, in the tracking process of this embodiment, at time T3, like the comparative example, track α is associated with the detection information of person C, which is different from person A with whom track α should originally be associated. However, as shown in FIG. 11, in the tracking process of this embodiment, the replicated track α can be associated again with the detection information of person A, which is the correct association destination, at time T4.

[0067] According to the above first embodiment, a replicated track in which a track for which an error condition regarding the possibility of an incorrect association between the track and the detection information holds is replicated in the state before update is generated as a track. Therefore, even when the original track is updated by being associated with the detection information of a target moving body 3 different from the original target moving body 3 that has been tracked until then, in a subsequent cycle, the original target moving body 3 can be associated with the replicated track replicated in the state before update. Therefore, the target moving body 3 can be tracked more accurately.

[0068] (Second Embodiment) As shown in FIG. 12, the second embodiment is a modification of the first embodiment.

[0069] In the second embodiment, the track management block 130 includes, in an error condition, that a track that was not associated with detection information in the previous cycle is associated with the detection information in the current cycle. In this case, as shown in FIG. 12, when the state of the track is changed from the lost state to the tracking state in S309, this flow proceeds to S311. That is, the process of S310 in the first embodiment is skipped, and the track is replicated regardless of the similarity between the track and the detection information.

[0070] According to the above second embodiment, when a track that was not associated with detection information in the previous cycle is associated with the detection information in the current cycle, a replicated track is generated. Therefore, when returning from a lost state where there is a relatively high possibility that detection information of a target moving body 3 different from the original target moving body 3 is associated, the target moving body 3 can be more reliably tracked.

[0071] (Third Embodiment) As shown in FIG. 13, the third embodiment is a modification of the first embodiment.

[0072] In the third embodiment, the track management block 130 includes, in an error condition, that a state in which the degree of decrease of the similarity with the detection information associated in a specific cycle after the specific cycle is outside the allowable decrease range continues for a predetermined number of cycles. The allowable decrease range is a range in which the magnitude of the decrease amount of the similarity is equal to or greater than a predetermined threshold value. For example, the track management block 130 determines that the error condition is satisfied when the amount of decrease of the similarity in the current cycle from the similarity in the previous cycle is outside the allowable decrease range. This corresponds to the case where the predetermined number of cycles described above is 1 cycle.

[0073] In this case, as shown in the figure, in S303a and S310a, the track management block 130 determines whether the amount of decrease of the similarity in the current cycle from the similarity in the previous cycle is outside the allowable decrease range.

[0074] According to the above-described third embodiment, a duplicate track can be generated for a track whose similarity has decreased since the previous cycle. Therefore, it is possible to more reliably track the target moving body 3.

[0075] (Fourth Embodiment) As shown in FIGS. 14 and 15, the fourth embodiment is a modification of the first embodiment.

[0076] In the fourth embodiment, the association block 120 acquires the degree of overlap between the target moving body 3 corresponding to the detection information and other target moving bodies 3. For example, the association block 120 may calculate the degree of overlap based on a predetermined image recognition process.

[0077] Then, the association block 120 prohibits the association between the detection information regarding an object whose degree of overlap with other objects is outside the allowable overlap range and the track. The allowable overlap range is, for example, a range where the degree of overlap is equal to or greater than a predetermined threshold value.

[0078] Further, the update block 140 updates the track in an update mode according to the degree of overlap between the target moving body 3 corresponding to the detection information and other target moving bodies 3. As an example of the update mode according to the degree of overlap, the update block 140 calculates the state quantity of the track according to the amount of observation noise correlated with the degree of overlap. Specifically, the update block 140 sets a larger amount of observation noise when calculating the estimated state quantity in the filtering step of the Kalman filter as the degree of overlap is larger.

[0079] Further, as another example of the update mode according to the degree of overlap, the update block 140 updates the feature quantity of the track at an update rate correlated with the degree of overlap. Specifically, the update block 140 decreases the update rate of the feature quantity in the track as the degree of overlap is larger.

[0080] In this case, as shown in FIG. 14, this flow proceeds to S260 after S200. In S260, the association block 120 calculates the similarity and the degree of overlap, and performs an association based on the similarity and the degree of overlap.

[0081] Then, as shown in FIG. 15, instead of the process of S314, the process of S316 is executed. In S316, the update block 140 updates the state quantity of the track by calculating the estimated state quantity with the amount of observation noise correlated with the degree of overlap. In S317 following S316, the update block 140 updates the feature quantity with the update rate correlated with the degree of overlap.

[0082] According to the above fourth embodiment, the track is updated in an update mode according to the degree of overlap between the target moving body 3 corresponding to the detection information and other target moving bodies 3. Therefore, it is possible to suppress the inclusion of information of other target moving bodies 3 in the track.

[0083] Also, according to the fourth embodiment, the association between the detection information regarding the target moving body 3 whose degree of overlap with other target moving bodies 3 is outside the allowable overlap range and the track is prohibited. Therefore, it is possible to avoid the association with the detection information in which the information of other target moving bodies 3 is likely to be mixed.

[0084] (Other Embodiments) Although a plurality of embodiments have been described above, the present disclosure is not construed as being limited to those embodiments, and can be applied to various embodiments and combinations without departing from the gist of the present disclosure.

[0085] In the modified example, the dedicated computer constituting the tracking system 100 may have at least one of a digital circuit and an analog circuit as a processor. Here, the digital circuit is, for example, at least one of ASIC (Application Specific Integrated Circuit), FPGA (Field Programmable Gate Array), SOC (System on a Chip), PGA (Programmable Gate Array), and CPLD (Complex Programmable Logic Device). Further, such a digital circuit may have a memory storing a program.

[0086] In the modified example, the host moving body to which the tracking system 100 is applied may be, for example, an autonomous mobile robot capable of carrying luggage or collecting information by autonomous driving or remote driving. In addition to the description forms so far, the above-described embodiments and modified examples may be implemented in the form of a processing circuit (for example, a processing ECU or the like) or a semiconductor device (for example, a semiconductor chip or the like) as a tracking device that is configured to be mountable on a host moving body and has at least one of a processor 102 and a memory 101.

Description of Reference Numerals

[0087] 1: Host vehicle (host moving body), 3: Target moving body, 10: External sensor system, 100: Tracking system, 101: Memory (storage medium), 102: Processor

Claims

1. A tracking system having a processor (102) for tracking a target moving body (3) detected by an external sensor system (10), wherein the processor acquires detection information which is information regarding the target moving body detected in the current cycle, associates tracking information which is information regarding the target moving body tracked before the previous cycle with the detection information estimated to be regarding the same target moving body as the tracking information, updates the tracking information according to the detection information associated in the current cycle, generates, as the tracking information, replicated tracking information in which the tracking information before update is replicated in a state where an error condition regarding the possibility that the association between the tracking information and the detection information is incorrect holds, and is configured to execute the above. The tracking system

2. The associating the tracking information with the detection information includes associating the tracking information and the detection information when the similarity between the tracking information and the detection information reaches an association range, The generating the replicated tracking information includes generating the replicated tracking information when the similarity is within an error range defined by an upper limit higher than the lower limit of the association range. The tracking system according to Claim 1

3. The associating the tracking information with the detection information includes when the tracking information is the replicated tracking information, associating the replicated tracking information and the detection information within the association range where the similarity is higher than the similarity between the original tracking information from which the replicated tracking information was generated and the detection information associated in the cycle in which the replicated tracking information was generated. The tracking system according to Claim 2

4. The generating the replicated tracking information includes generating the replicated tracking information when the tracking information not associated with the detection information in the previous cycle is associated with the detection information in the current cycle. The tracking system according to any one of Claims 1 to 3

5. The generating the replicated tracking information includes generating the replicated tracking information when a state where a degree of decrease from the specific cycle of the similarity between the tracking information and the detection information associated in the specific cycle is outside an allowable decrease range continues for a predetermined number of cycles. The tracking system according to Claim 1

6. Updating the tracking information is The tracking system according to any one of claims 1 to 5, wherein the tracking information is updated in an update mode according to the degree of overlap between the target moving body corresponding to the detection information and the other target moving bodies.

7. Associating the tracking information with the detection information is The tracking system according to any one of claims 1 to 6, wherein the association between the detection information regarding the target moving body whose degree of overlap with the other target moving bodies is outside the allowable overlap range and the tracking information is prohibited.

8. A tracking device having a processor (102), configured to be mounted on a host moving body (1), and tracking a target moving body (3) detected by an external sensor system (10), The processor Obtaining detection information, which is information regarding the target moving body detected in the current cycle, Associating tracking information, which is information regarding the target moving body tracked before the previous cycle, with the detection information estimated to be regarding the same object as the tracking information, Updating the tracking information according to the detection information associated in the current cycle, Generating, as the tracking information, replicated tracking information in which the tracking information in a state before update is replicated when an error condition regarding the possibility that the association between the tracking information and the detection information is incorrect is satisfied, A tracking device configured to execute the above.

9. A tracking method executed by a processor (102) for tracking a target moving body (3) detected by an external sensor system (10), Obtaining detection information, which is information regarding the target moving body detected in the current cycle, Associating tracking information, which is information regarding the target moving body tracked before the previous cycle, with the detection information estimated to be regarding the same object as the tracking information, Updating the tracking information according to the detection information associated in the current cycle, Generating, as the tracking information, replicated tracking information in which the tracking information in a state before update is replicated when an error condition regarding the possibility that the association between the tracking information and the detection information is incorrect is satisfied, A tracking method including the above.

10. A tracking program stored in a storage medium (101) and executed by a processor (102) to track a target moving body (3) detected by an external sensor system (10), wherein the instructions are: to acquire detection information which is information about the target moving body detected in the current cycle; to associate tracking information which is information about the target moving body tracked before the previous cycle with the detection information estimated to be about the same object as the tracking information; to update the tracking information according to the detection information associated in the current cycle; to generate, as the tracking information, replicated tracking information in which the tracking information before update is replicated in a state where an error condition regarding the possibility that the association between the tracking information and the detection information is incorrect holds; and the tracking program includes the above.

Images