A target trajectory association method

Abstract

This invention discloses a target trajectory association method, relating to the field of target tracking, comprising: acquiring multiple ground truth trajectories and multiple sensing trajectories of the target in full time sequence; determining the association cost between ground truth trajectories and sensing trajectories in different time periods, wherein the association cost is negatively correlated with the length of the intersection time period of the ground truth trajectory and the sensing trajectory to be associated; and determining the sensing trajectory associated with each ground truth trajectory based on the determined association cost. This invention has the advantage of reducing erroneous associations between the target's ground truth trajectories and sensing trajectories.

Description

Technical Field

[0001] This invention relates to the field of target tracking technology, and more specifically to a target trajectory association method. Background Technology

[0002] In automated perception performance evaluation before vehicles leave the factory and in some offline multi-target tracking tasks, it is a very important step to correctly "correlate" the ground truth trajectory of the target with the target trajectory output by perception.

[0003] Both real-time and offline association technologies in the present technology have the problem that the association results do not match the actual degree of association. Summary of the Invention

[0004] This invention aims to address, to a certain extent, one of the technical problems in related technologies. To this end, this invention provides a target trajectory association method, which has the advantage of reducing erroneous associations between the true trajectory and the perceived trajectory of the target.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] A target trajectory association method, comprising:

[0007] Acquire multiple truth trajectories and multiple sensing trajectories of the target throughout the entire time series;

[0008] Determine the correlation value between the true value trajectory and the perceived trajectory in different time periods, wherein the correlation value is negatively correlated with the length of the intersection time period of the true value trajectory and the perceived trajectory to be correlated.

[0009] Based on the determined association value, the perceptual trajectory associated with each truth value trajectory is determined.

[0010] Optionally, the associated cost is positively correlated with the positional deviation between the true trajectory and the perceived trajectory to be associated.

[0011] Optionally, the cost of determining the correlation between the true trajectory and the perceived trajectory in different time periods includes:

[0012] Based on the obtained true trajectory and sensing trajectory, calculate the length of the intersection time period between the sensing trajectory to be associated and the true trajectory, as well as the positional deviation between the sensing trajectory to be associated and the true trajectory.

[0013] The association cost is calculated based on the intersection time period length and the positional deviation value.

[0014] Optionally, the association cost is calculated based on the intersection time period length and the positional deviation value, including:

[0015] Obtain a first weight coefficient corresponding to the length of the intersection time period and a second weight coefficient corresponding to the position deviation value; wherein the sum of the first weight coefficient and the second weight coefficient is 1;

[0016] The association value is obtained by summing the product of the intersection time period length and the first weight coefficient with the product of the position deviation value and the first weight coefficient.

[0017] Optionally, based on the determined association cost, the perceptual trajectory associated with each truth trajectory is determined, including:

[0018] Each truth trajectory is associated with the corresponding perceptual trajectory with the lowest associated cost, resulting in multiple first trajectory association pairs; wherein each first trajectory association pair includes a truth trajectory and a perceptual trajectory.

[0019] Optionally, the step of associating each truth trajectory with the corresponding perceptual trajectory that has the lowest associated cost to obtain multiple first trajectory association pairs includes:

[0020] If the intersection time period between the sensing trajectory to be associated and the true value trajectory is empty, cancel the trajectory association between the corresponding sensing trajectory and the true value trajectory.

[0021] Optionally, after associating each truth trajectory with the corresponding perceptual trajectory that has the lowest association cost to obtain multiple first trajectory association pairs, the process includes:

[0022] In the unassociated perceptual trajectories, the truth trajectories in the first trajectory association pair, and the unassociated truth trajectories, each truth trajectories is associated with the perceptual trajectory with the lowest associated cost.

[0023] The trajectory association is iterated until the trajectory association result is empty, resulting in a second trajectory association pair; wherein the second trajectory association pair includes a true trajectory and at least one perceived trajectory.

[0024] Optionally, the step of associating each truth value trajectory with the corresponding perception trajectory with the lowest association cost among the unassociated perception trajectories and the first trajectory association pair includes:

[0025] If the time periods of unassociated sensing trajectories and the sensing trajectories in the trajectory association pair overlap, cancel the trajectory association between the corresponding sensing trajectory and the ground truth trajectory.

[0026] Furthermore, the present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the target trajectory association method described in any of the above claims.

[0027] In addition, the present invention also provides a computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the target trajectory association method described in any of the above claims.

[0028] These features and advantages of the present invention will be disclosed in detail in the following specific embodiments and accompanying drawings. The preferred embodiments or means of the present invention will be shown in detail in conjunction with the accompanying drawings, but are not intended to limit the technical solutions of the present invention. In addition, each of these features, elements and components appearing in the following text and drawings is a plurality of, and different symbols or numbers are used for convenience of representation, but all represent parts with the same or similar construction or function. Attached Figure Description

[0029] The present invention will be further described below with reference to the accompanying drawings:

[0030] Figure 1 This is a schematic diagram illustrating the steps of a target trajectory association method according to an embodiment of the present invention;

[0031] Figure 2 This is a schematic diagram of the perceived trajectory point and the true trajectory point at time T1 in the real-time correlation method.

[0032] Figure 3 This is a schematic diagram of the perceived trajectory point and the true trajectory point at time T1 in the real-time correlation method.

[0033] Figure 4 This is a schematic diagram of the perceived trajectory point and the true trajectory point at time T1 in the real-time correlation method.

[0034] Figure 5 This is a schematic diagram of the perceived trajectory and the ground truth trajectory in the offline association method;

[0035] Figure 6 This is a schematic diagram illustrating the trajectory association between the perceived trajectory and the true trajectory in the above embodiments;

[0036] Figure 7 This is a schematic diagram of the storage device module in the above embodiments of the present invention;

[0037] Figure 8 This is a schematic diagram of a computer-readable medium in the above embodiments of the present invention. Detailed Implementation

[0038] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described are intended to explain the present invention and should not be construed as limiting the invention.

[0039] The terms "an embodiment," "example," or "trademark" used in this specification refer to a particular feature, structure, or characteristic described in connection with the embodiment itself that may be included in at least one embodiment disclosed in this patent. The phrase "in an embodiment" appearing in various places throughout the specification does not necessarily refer to the same embodiment.

[0040] Existing methods for target trajectory association include real-time and offline methods. For example, in real-time association, only the data of the current frame is used. The association cost is calculated based on the positional relationship between the perceived trajectory (det) and the ground truth trajectory (GT). Multiple association cost values ​​are constructed into an association cost matrix, and finally, the matching result between the perceived trajectory and the ground truth trajectory is obtained using the Hungarian algorithm. While real-time association has low computational cost and high speed, it lacks historical and future information, making it prone to errors in complex scenarios, especially when the perceived results fluctuate. Figures 2 to 4 As shown, from the perspective of minimizing the association cost of trajectory points at each time step, det2 can be associated with GT1. However, from the perspective of obtaining the overall trajectory association result, det1 should be associated with GT1, and det2 should be associated with GT2.

[0041] In offline association, for example, temporal information of the perceived trajectory and the ground truth trajectory can be created separately. Based on the temporal information of the perceived trajectory and the ground truth trajectory, the corresponding association cost is calculated, and then a greedy algorithm is used to obtain the matching result of the perceived trajectory and the ground truth trajectory in each frame. Although this method solves the problem of single-frame incorrect association caused by real-time association not utilizing historical and future information, such as... Figure 5 As shown, within the circle, if real-time correlation is used, the GT green should be correlated with the det black, but from the offline full-time-series trajectory, it should be correlated with the det brown. However, insufficient consideration in offline correlation calculations can also introduce problems. For example... Figure 3 In the previous calculation method, the result of the association cost between det red and GT green is the smallest. However, since the intersection time period of det red and GT green is short, the association result does not match the actual degree of association.

[0042] In the above scenario, different association algorithms or thresholds will yield different association results. However, only one result is correct. If the calculated association result is incorrect, it will affect the results of subsequent algorithms.

[0043] Therefore, as a first aspect of the present invention, a target trajectory association method is provided, such as... Figure 1 As shown, the method includes:

[0044] In step S110, multiple true-value trajectories and multiple sensing trajectories of the target in full time series are obtained;

[0045] In step S120, the correlation value between the true value trajectory and the perceived trajectory in different time periods is determined, wherein the correlation value is negatively correlated with the length of the intersection time period of the true value trajectory and the perceived trajectory to be correlated.

[0046] In step S130, the perception trajectory associated with each truth trajectory is determined based on the determined association value.

[0047] The target trajectory method of this invention is based on an offline algorithm, thereby acquiring multiple true-value trajectories and multiple sensing trajectories of the target in full-time sequence in step S110. Here, "offline" refers to the algorithm used in automobiles, which can only compute information at time t and cannot utilize information at time t+1. Furthermore, due to limitations in computing resources (n), it often only uses data after time tn and cannot use data before tn-1. However, the offline algorithm can acquire full-time sequence information without considering the limitations of computing resources.

[0048] In step S120, the association cost is negatively correlated with the length of the intersection time period between the true trajectory and the perceived trajectory to be associated. When the intersection time period is short, the perceived trajectory and the true trajectory are prone to losing connection, i.e., erroneous association occurs. When the intersection time period is long, the degree of association is correspondingly higher. Therefore, the length of the intersection time period is used as an influencing factor of the association cost to reduce erroneous associations between the perceived trajectory and the true trajectory.

[0049] The calculation method for the length of the intersection time period is shown in formula (1):

[0050] T inter =max(T) det1 ,T GT1 ),min(T detn ,T GTn (1)

[0051] T inter T represents the length of the intersection time period. det1 T represents the start time of the perceived trajectory. GT1 T represents the start time of the truth locus. detn T represents the endpoint time of the sensing trajectory. GTn This represents the endpoint time of the true value trajectory.

[0052] In step S130, the perceived trajectory is associated with a suitable truth trajectory based on the determined association cost value. In this invention, there are no special limitations on how the perceived trajectory is associated with a suitable truth trajectory. For example, an association cost matrix can be constructed using the determined association cost value, and the aforementioned Hungarian algorithm can be used to associate the perceived trajectory with a suitable truth trajectory based on the association cost matrix.

[0053] To make it easier to associate GT and det with similar motion trends, optionally, the association cost is positively correlated with the positional deviation value of the true trajectory and the perceived trajectory to be associated.

[0054] The positional deviation value is calculated as shown in formula (2):

[0055]

[0056] Corres represents the positional deviation between the true trajectory and the perceived trajectory within the intersection time period. i To introduce the correlation coefficient, corres i The larger the value, the closer the perceived trajectory is to the true trajectory, and the smaller the position deviation value.

[0057] For example, the correlation coefficient for lateral positions is calculated as shown in formula (3):

[0058]

[0059] in, σ represents the covariance of the lateral positions of the true trajectory and the perceived trajectory. GT_lateral σ det_lateral It represents the standard deviation of the lateral position of the true trajectory and the perceived trajectory.

[0060] As an optional implementation, step S120 includes:

[0061] Based on the obtained true trajectory and sensing trajectory, calculate the length of the intersection time period between the sensing trajectory to be associated and the true trajectory, as well as the positional deviation between the sensing trajectory to be associated and the true trajectory.

[0062] The association cost is calculated based on the intersection time period length and positional deviation.

[0063] Specifically, a first weighting coefficient corresponding to the length of the intersection time period and a second weighting coefficient corresponding to the position deviation value are set according to the actual situation.

[0064] The correlation cost is obtained by summing the product of the intersection time period length and the first weight coefficient with the product of the position deviation and the first weight coefficient; where the sum of the first weight coefficient and the second weight coefficient is 1.

[0065] The calculation method for the associated cost value is shown in formula (4):

[0066]

[0067] cost represents the correlation cost, μ represents the first weighting coefficient, and 1-μ represents the second weighting coefficient. The larger the intersection time period length and the smaller the positional deviation, the smaller the correlation cost, meaning the higher the correlation between the perceived trajectory and the true trajectory. inter This represents the proportion of the intersection time interval length in the entire time interval of the true value trajectory, and is calculated as shown in formula (5):

[0068]

[0069] Here, γ is a number between 0 and 1, which is the scale correction for rat io.

[0070] As an optional implementation, step S130 includes:

[0071] Each true trajectory is associated with the corresponding perceptual trajectory with the lowest associated cost, resulting in multiple first trajectory association pairs; each first trajectory association pair includes a true trajectory and a perceptual trajectory.

[0072] An association cost matrix is ​​established based on the association cost value, and the Hungarian matching algorithm is used to perform maximum matching between the true trajectory and the associated trajectory to obtain multiple first trajectory association pairs.

[0073] Optionally, each truth trajectory is associated with the corresponding perceptual trajectory that has the lowest association cost, resulting in multiple first trajectory association pairs, including:

[0074] If the intersection time period between the sensing trajectory to be associated and the true value trajectory is empty, cancel the trajectory association between the corresponding sensing trajectory and the true value trajectory.

[0075] When there is no overlap in time period between the sensing trajectory to be associated and the true trajectory, the corresponding sensing trajectory and the true trajectory are not associated.

[0076] To avoid the reduction in the correlation time between the sensing trajectory and the truth trajectory due to interruptions in the sensing trajectory, as an optional implementation method, such as... Figure 6 As shown, after associating each truth trajectory with the corresponding perceptual trajectory that has the lowest associated cost, resulting in multiple first trajectory association pairs, the process includes:

[0077] In the unassociated perceptual trajectories, the truth trajectories in the first trajectory association pair, and the unassociated truth trajectories, each truth trajectories is associated with the perceptual trajectory with the lowest associated cost.

[0078] The trajectory association is iterated until the trajectory association result is empty, and a second trajectory association pair is obtained; the second trajectory association pair includes a true value trajectory and at least one perceived trajectory.

[0079] like Figure 6 As shown, a new association cost matrix is ​​constructed using all ground truth trajectories and unassociated perceptual trajectories. If an unassociated perceptual trajectory intersects with an associated perceptual trajectory in the ground truth trajectories in time, the corresponding association cost is set to the maximum value to prevent association. Then, the Hungarian algorithm is used to complete a new round of matching. The matching results still contain: associated ground truth trajectory-perceptual trajectory pairs, unassociated ground truth trajectories, and unassociated perceptual trajectories. This process is iterated until the associated ground truth trajectory-perceptual trajectory pairs are empty.

[0080] Extract all associated ground truth trajectories and perception trajectories as the final association results. The pairing results include one-to-one, where one ground truth trajectory is associated with one perception trajectory, and one-to-many, where one ground truth trajectory is associated with multiple perception trajectories, and the multiple perception trajectories do not overlap in time.

[0081] By constructing an association cost matrix using associated cost values ​​and employing an iteratively cascaded Hungarian matching algorithm, a true value trajectory can be associated with multiple sensing trajectories. This avoids the reduction in the association time between sensing trajectories and true value trajectories when multiple sensing trajectories are formed due to sensing interruptions (such as occlusion or missed detection), thus improving the accuracy of association.

[0082] Meanwhile, this embodiment also provides an electronic device, such as... Figure 7 As shown, it includes:

[0083] One or more processors 101;

[0084] The memory 102 stores one or more computer programs that, when executed by the one or more processors 101, cause the one or more processors 101 to implement the target trajectory association method according to the first aspect of the invention.

[0085] The electronic device may also include one or more I / O interfaces 103 connected between the processor 101 and the memory 102, configured to enable information interaction between the processor 101 and the memory 102.

[0086] The processor 101 is a device with data processing capabilities, including but not limited to a central processing unit (CPU); the memory 102 is a device with data storage capabilities, including but not limited to random access memory (RAM, more specifically SDRAM, DDR, etc.), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), and flash memory (FLASH); the I / O interface 103 (read-write interface) is connected between the processor 101 and the memory 102, enabling information exchange between the processor 101 and the memory 102, including but not limited to a data bus (Bus).

[0087] In some embodiments, the processor 101, memory 102, and I / O interface 103 are interconnected via bus 104, and thus connected to other components of the computing device.

[0088] As a third aspect of the present invention, a computer-readable medium, such as... Figure 8 As shown, a computer program is stored thereon, which, when executed by a processor, implements the target trajectory association method provided in the first aspect of this disclosure.

[0089] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. Accordingly, the computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can implement the methods of any of the above embodiments. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0090] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Those skilled in the art should understand that the present invention includes, but is not limited to, the contents described in the accompanying drawings and the specific embodiments above. Any modifications that do not depart from the functional and structural principles of the present invention will be included within the scope of the claims.

Claims

1. A target trajectory association method, characterized in that, include: Acquire multiple truth trajectories and multiple sensing trajectories of the target throughout the entire time series; The association cost between the true trajectory and the perceived trajectory in different time periods is determined. The true trajectory includes the target's actual motion trajectory GT, and the perceived trajectory includes the target's predicted motion trajectory det. The association cost is calculated based on the intersection time period length and positional deviation of the true trajectory and the perceived trajectory to be associated. The association cost is negatively correlated with the intersection time period length of the true trajectory and the perceived trajectory. The formula for calculating the intersection time period length includes: =max( , ),min( , ), This represents the length of the intersection time period. Indicates the start time of the perceived trajectory. Indicates the start time of the truth locus. Indicates the end time of the perceived trajectory. The formula for calculating the position deviation value, representing the endpoint time of the true trajectory, includes: , This represents the positional deviation between the true trajectory and the perceived trajectory within the intersection time period. This represents the introduced correlation coefficient. The larger the value, the smaller the positional deviation value. The formula for calculating the associated cost value includes: , Indicates associated value. This represents the first weighting coefficient. This represents the second weighting coefficient. This represents the proportion of the intersection time interval length value within the entire time interval of the truth trajectory. The formula for calculating this proportion includes: , It is a number between 0 and 1; Based on the determined correlation value, determine the perceptual trajectory associated with each truth value trajectory; Among them, based on the determined association value, the perceptual trajectory associated with each truth value trajectory is determined, including: Each true trajectory is associated with the corresponding perceptual trajectory with the lowest associated cost, resulting in multiple first trajectory association pairs; wherein each first trajectory association pair includes a true trajectory and a perceptual trajectory. The step of associating each true trajectory with the corresponding perceptual trajectory that has the lowest associated cost to obtain multiple first trajectory association pairs includes: If the intersection time period between the sensing trajectory to be associated and the true value trajectory is empty, cancel the trajectory association between the corresponding sensing trajectory and the true value trajectory.

2. The target trajectory association method according to claim 1, characterized in that, The step of associating each truth trajectory with the corresponding perceptual trajectory that has the lowest associated cost to obtain multiple first trajectory association pairs includes: In the set consisting of unassociated perceptual trajectories, truth trajectories in the first trajectory association pair, and unassociated truth trajectories, for each truth trajectories in the set, each truth trajectories is associated with the perceptual trajectory with the lowest association cost determined from the unassociated perceptual trajectories. The trajectory association is iterated until the trajectory association result is empty, resulting in a second trajectory association pair; wherein the second trajectory association pair includes a true trajectory and at least one perceived trajectory.

3. The target trajectory association method according to claim 2, characterized in that, The step of associating each truth value trajectory with the corresponding perception trajectory with the lowest association cost among the unassociated perception trajectories, the truth value trajectories in the first trajectory association pair, and the unassociated truth value trajectories includes: When an unassociated sensing trajectory overlaps with an already associated sensing trajectory in the first trajectory association pair in terms of time period, the association qualification between the unassociated sensing trajectory and the corresponding truth value trajectory is cancelled.

4. An electronic device, characterized in that, include: One or more processors; A memory having stored one or more computer programs that, when executed by one or more processors, cause the one or more processors to implement the target trajectory association method according to any one of claims 1 to 3.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the target trajectory association method as described in any one of claims 1 to 3.

Images