Track fusion method, fusion device, processor and fusion system

Abstract

The present application provides a track fusion method, a fusion device, a processor, and a fusion system. The method includes: obtaining multiple first track data sets and multiple second track data sets; calculating the sum of the mean square errors of all track data subset pairs in each track data set group to obtain multiple total mean square error sums, wherein the track data set group includes multiple track data subset pairs, a track data subset pair includes a first track data set and a second track data set, and the first track data set and the second track data set of any two track data subset pairs are different; using a track fusion algorithm to fuse the two tracks corresponding to the optimal track data subset pair to obtain multiple optimal tracks, wherein the optimal track data subset pair is the track data subset pair with the smallest total mean square error and the corresponding track data set group. This method solves the problem of low accuracy of the track similarity calculation method used in the prior art for track association.

Description

Technical Field

[0001] The present application relates to the field of track association technology, and in particular to a track fusion method, a fusion device, a processor, and a fusion system. Background Art

[0002] In the field of autonomous driving, in order to meet the needs of environmental perception, vehicles need to be equipped with multi-source heterogeneous sensors to give full play to the advantages of each sensor and avoid the disadvantages of each sensor. Sensor track fusion is an important part of environmental perception fusion. Due to the different characteristics of multi-source heterogeneous sensors, they are easily affected by environmental factors, resulting in large detection errors. The sensor track and system track fusion solution has good versatility and scalability, and is convenient for rapid convergence when accuracy jumps and eliminating the influence of outliers on data. The multi-sensor track fusion algorithm is based on the detection of target track information by multi-source heterogeneous sensors (millimeter wave radar, camera, lidar, corner radar, etc.) after tracking. The sensor target track information is preprocessed, synchronized in time and space, associated, and managed in the world coordinate system, and finally the fused target information is output. However, in the track association process, the existing track methods are computationally complex, especially when there are a large number of scene targets, which is prone to mismatching. The amount of association calculation also increases significantly, seriously affecting the target tracking efficiency. In addition, the existing track methods mostly use Mahalanobis distance and Euclidean distance to represent track similarity. The Mahalanobis distance relies on the target attribute variance value and attribute variance value accuracy provided by the sensor, and has poor applicability. The Euclidean distance represents track similarity and only considers position attribute information, which may cause mismatching problems.

[0003] The above information disclosed in the background technology section is only used to enhance the understanding of the background technology of the technology described in this article. Therefore, the background technology may contain certain information that does not form the prior art known in this country to those skilled in the art. Summary of the Invention

[0004] The main purpose of the present application is to provide a track fusion method, a fusion device, a processor and a fusion system to solve the problem of low accuracy of the track similarity calculation method used in track association in the prior art.

[0005] According to one aspect of an embodiment of the present application, a track fusion method is provided, the method comprising: respectively acquiring track data of multiple tracks of a first sensor and a second sensor to obtain multiple first track data sets and multiple second track data sets, the first track data set comprising multiple track data of an obstacle detected by the first sensor within a time period, the second track data set comprising multiple track data of an obstacle detected by the second sensor within the time period, the time period comprising multiple time nodes, the track data corresponding to the time nodes one by one; calculating the sum of the mean square errors of all track data subset pairs in each track data set group to obtain multiple total mean square error sums, the track data set group comprising multiple track data According to subset pairs, a track data subset pair includes a first track data set and a second track data set, the first track data sets of any two track data subset pairs are different, and the second track data sets of any two track data subset pairs are different. The mean square error of the track data subset pair is the sum of the mean square errors of the track data in the track data subset pair, and the total mean square error sum corresponds one-to-one to the track data set group; a track fusion algorithm is used to fuse the two tracks corresponding to the optimal track data subset pair to obtain multiple optimal tracks, the optimal track data subset pair is the track data subset pair of the track data set group corresponding to the minimum sum of the total mean square error, and the optimal track data subset pair corresponds one-to-one to the optimal track.

[0006] Optionally, after respectively acquiring the track data of multiple tracks of the first sensor and the second sensor to obtain multiple first track data sets and multiple second track data sets, the method further includes: a first acquisition step of acquiring a preliminary track data set group, wherein the preliminary track data set group is any one of the track data set groups; a second acquisition step of acquiring a preliminary track data subset pair, wherein the preliminary track data subset pair is any one of the track data subset pairs of the preliminary track data set group, and the track data in the first track data set of the preliminary track data subset pair corresponds one-to-one to the track data in the second track data set of the preliminary track data subset pair; a judgment step of judging whether the difference between the track data of the first track data set in the preliminary track data subset pair and the corresponding track data in the second track data set is less than a preset difference, and obtaining multiple judgment results; a second acquisition step of acquiring a preliminary track data subset pair, wherein the preliminary track data subset pair is any one of the track data subset pairs of the preliminary track data set group, and the track data in the first track data set of the preliminary track data subset pair corresponds one-to-one to the track data in the second track data set of the preliminary track data subset pair; and a judgment step of judging whether the difference between the track data of the first track data set in the preliminary track data subset pair and the corresponding track data in the second track data set is less than a preset difference, and obtaining multiple judgment results. a repeating step of repeating the second obtaining step and the judging step at least once until the judging of the track data of all the track data subset pairs in the prepared track data set group is completed, and a plurality of the judging results are obtained; a determining step of determining whether the prepared track data set group meets a screening requirement based on all the judging results, wherein the screening requirement is that the difference between the track data in the first track data set and the corresponding track data in the second track data set of all the track data subset pairs in the prepared track data set group is less than a preset difference; repeating the first obtaining step, the first repeating step, and the determining step at least once until the determining of all the track data set groups is completed; deleting the track data set groups that do not meet the screening requirement from all the track data set groups, and obtaining a plurality of track data set groups that meet the screening requirement.

[0007] Optionally, based on all the judgment results, determining whether the preliminary track dataset group meets the screening requirements includes: when all the judgment results are yes, determining that the preliminary track dataset group meets the screening requirements; when at least one of the judgment results is no, determining that the preliminary track dataset group does not meet the screening requirements.

[0008] Optionally, the sum of the mean square errors of all track data subset pairs in each track data set group is calculated to obtain multiple total mean square error sums, including: a third acquisition step of acquiring a target track data set group, where the target track data set group is any one of the track data set groups that meets the screening requirements; a first calculation step of calculating the mean square error of each track data subset pair in the target track data set group to obtain multiple target mean square errors; a second calculation step of calculating the sum of all the target mean square errors to obtain a target total mean square error sum; repeating the third acquisition step, the first calculation step and the second calculation step at least once until the calculation of the sum of the mean square errors of all track data subset pairs in all the track data set groups is completed to obtain multiple total mean square error sums.

[0009] Optionally, the track data includes position data and speed data of the obstacle, and the mean square error of each track data subset pair in the target track data set group is calculated to obtain multiple target mean square errors, including: a fourth acquisition step of obtaining a target track data subset pair, wherein the target track data subset pair is any one of the track data subset pairs in the target track data set group; a third calculation step of calculating the mean square error of the position data of the target track data subset pair to obtain multiple target position mean square errors, calculating the mean square error of the speed data of the target track data subset pair to obtain multiple target position mean square errors, and calculating the mean square error of the speed data of the target track data subset pair to obtain multiple target position mean square errors. target speed mean square deviation; a fourth calculation step, calculating the sum of all the target position mean square deviations to obtain a first target mean square deviation, calculating the sum of all the target speed mean square deviations to obtain a second target mean square deviation; a fifth calculation step, calculating the sum of the first target mean square deviation and the second target mean square deviation to obtain the target mean square deviation; repeating the fourth acquisition step, the third calculation step, the fourth calculation step and the fifth calculation step at least once until the calculation of the mean square deviations of all the track data subset pairs in the target track data set group is completed to obtain multiple target mean square deviations.

[0010] Optionally, before adopting a track fusion algorithm to fuse the two tracks corresponding to the optimal track data subset pair to obtain multiple optimal tracks, the method further includes: determining the minimum value of all the total mean square error sums to obtain the optimal total mean square error sum; and determining the optimal track data subset pair based on the optimal total mean square error sum, the optimal track data subset pair being the track data subset pair of the track data set group corresponding to the optimal total mean square error sum.

[0011] Optionally, the number of the track data in the first track data set and the number of the track data in the second track data set are both equal to a predetermined number.

[0012] According to another aspect of an embodiment of the present application, a track fusion device is further provided, the device comprising: an acquisition unit, which respectively acquires track data of multiple tracks of a first sensor and a second sensor to obtain multiple first track data sets and multiple second track data sets, the first track data set comprising multiple track data of an obstacle detected by the first sensor within a time period, the second track data set comprising multiple track data of an obstacle detected by the second sensor within the time period, the time period comprising multiple time nodes, the track data corresponding to the time nodes one by one; a calculation unit, which calculates the sum of the mean square errors of all track data subset pairs in each track data set group to obtain multiple total mean square error sums, the track data set group comprising multiple A track data subset pair, wherein the track data subset pair includes a first track data set and a second track data set, the first track data sets of any two track data subset pairs are different, the second track data sets of any two track data subset pairs are different, the mean square error of the track data subset pair is the sum of the mean square errors of the track data in the track data subset pair, and the total mean square error sum corresponds one-to-one to the track data set group; a fusion unit, which adopts a track fusion algorithm to fuse the two tracks corresponding to the optimal track data subset pair to obtain multiple optimal tracks, the optimal track data subset pair is the track data subset pair of the track data set group corresponding to the minimum total mean square error sum, and the optimal track data subset pair corresponds one-to-one to the optimal track.

[0013] According to yet another aspect of the embodiments of the present application, a processor is provided, which is configured to run a program, wherein any one of the methods described is executed when the program is run.

[0014] According to another aspect of an embodiment of the present application, a track fusion system is also provided, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include methods for executing any one of the methods described.

[0015] In the above track fusion method, first, track data of multiple tracks of the first sensor and the second sensor are obtained respectively to obtain multiple first track data sets and multiple second track data sets, wherein the above first track data set includes multiple track data of an obstacle detected by the above first sensor within a time period, and the above second track data set includes multiple track data of an obstacle detected by the above second sensor within the above time period, the above time period includes multiple time nodes, and the above track data correspond to the above time nodes one by one; then, the sum of the mean square errors of all track data subset pairs in each track data set group is calculated to obtain multiple total mean square error sums, the above track data set group includes multiple track data subset pairs, and one above track data set is a plurality of track data subset pairs. The track data subset pair includes the above-mentioned first track data set and the above-mentioned second track data set, the above-mentioned first track data sets of any two of the above-mentioned track data subset pairs are different, the above-mentioned second track data sets of any two of the above-mentioned track data subset pairs are different, the mean square error of the above-mentioned track data subset pair is the sum of the mean square errors of the above-mentioned track data in the above-mentioned track data subset pair, and the above-mentioned total mean square error sum corresponds one-to-one to the above-mentioned track data set group; finally, a track fusion algorithm is used to fuse the two above-mentioned tracks corresponding to the optimal track data subset pair to obtain multiple optimal tracks, the above-mentioned optimal track data subset pair is the above-mentioned track data subset pair of the above-mentioned track data set group corresponding to the minimum sum of the above-mentioned total mean square error, and the above-mentioned optimal track data subset pair corresponds one-to-one to the above-mentioned optimal track. In this method, a track data subset pair is a set of track data of a track detected by a first sensor and a track data of a track detected by a second sensor. The mean square error of the track data subset pair reflects the similarity of the two tracks detected by the two sensors. The tracks detected by the first sensor are matched with the tracks detected by the second sensor pairwise to obtain multiple track data sets. Each track data set group corresponds to a matching method. The total mean square error sum of a track data set group reflects the overall similarity of each matched track pair in a matching method. The matching method with the highest overall similarity of each matched track pair is determined based on the minimum total mean square error sum. The two tracks corresponding to the optimal track data set group with the minimum total mean square error sum are the two tracks of one target obstacle detected by each sensor. These two tracks are fused using a track fusion algorithm to obtain the two correlated tracks of the target obstacle detected by the two sensors. This method solves the problem of low accuracy of track similarity calculation methods used in track correlation in the prior art. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The drawings that constitute part of this application are used to provide a further understanding of this application. The illustrative embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation on this application. In the drawings:

[0017] Figure 1 A flow chart of a track fusion method according to an embodiment of the present application is shown;

[0018] Figure 2 A schematic diagram of a track fusion device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0019] It should be noted that, in the absence of conflict, the embodiments and features of the embodiments in this application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0020] In order to enable those skilled in the art to better understand the present invention, the following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0021] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequential order. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0022] It should be understood that when an element (such as a layer, film, region, or substrate) is described as being "on" another element, the element may be directly on the other element or intervening elements may be present. Moreover, in the specification and claims, when it is described that an element is "connected to" another element, the element may be "directly connected to" the other element or "connected to" the other element through a third element.

[0023] As mentioned in the background technology, the method for calculating track similarity used in the prior art for track association has low accuracy. In order to solve the above problem, a typical embodiment of the present application provides a track fusion method, fusion device, processor and fusion system.

[0024] According to an embodiment of the present application, a track fusion method is provided.

[0025] Figure 1 FIG. 1 is a flow chart of a track fusion method according to an embodiment of the present application. Figure 1 As shown, the method includes the following steps:

[0026] Step S101: acquiring track data of a plurality of tracks of a first sensor and a second sensor, respectively, to obtain a plurality of first track data sets and a plurality of second track data sets, wherein the first track data set includes a plurality of track data of an obstacle detected by the first sensor within a time period, and the second track data set includes a plurality of track data of an obstacle detected by the second sensor within the time period, wherein the time period includes a plurality of time nodes, and the track data correspond to the time nodes in a one-to-one manner;

[0027] Among them, the accuracy of the track similarity calculation method used in track association in the prior art is low. In order to solve the above problem, it is first necessary to obtain track data of multiple tracks of the first sensor and track data of multiple tracks of the second sensor, and then solve the problem of low accuracy of the track similarity calculation method used in track association in the prior art.

[0028] It should be noted that in complex scenarios with multiple target obstacles, the track data detected by each sensor will deviate, leading to the introduction of noise and interference. With a lot of clutter, only considering the track data of a single time period will result in incorrect or unmatched target obstacles. Therefore, it is necessary to preprocess the track data detected by each sensor. Taking the preprocessing of track data of an obstacle detected by the first sensor as an example, a track consists of multiple track points. The track data of a track point includes position data and speed data. The position data includes lateral distance data and longitudinal distance data, and the speed data includes lateral speed data and longitudinal speed data. The first track data set corresponding to the track includes the track data of all track points in the track. The first track data set corresponding to the track in which the sensor detected the obstacle in the previous time period and the first track data set corresponding to the track in which the sensor detected the obstacle in the current time period are obtained. The track points where the track data jump in the two tracks are compared, and the track data corresponding to these unstable track points are deleted from the first track data set of the current time period.

[0029] It should also be noted that the number of track data in the first track data set and the number of track data in the second track data set are both equal to a predetermined number. The predetermined number of track data for each track ensures that each track has a sufficient amount of track data, thereby ensuring the accuracy of the calculated total mean square error (MSS), and improving the correct association rate of target obstacle tracks.

[0030] In addition, in an optional embodiment, after the above step S101, the above method further includes:

[0031] Step S201, a first acquisition step, acquiring a preliminary track data set group, wherein the preliminary track data set group is any one of the aforementioned track data set groups;

[0032] Step S202, a second acquisition step, acquiring a pair of prepared track data subsets, wherein the pair of prepared track data subsets is any one of the pair of track data subsets in the prepared track data set group, wherein the track data in the first track data set of the prepared track data subset pair corresponds one-to-one to the track data in the second track data set of the prepared track data subset pair;

[0033] Step S203, a judgment step, judging whether a difference between the track data of the first track data set in the prepared track data subset pair and the corresponding track data in the second track data set is less than a preset difference, and obtaining multiple judgment results;

[0034] Step S204, a first repetition step, repeating the second obtaining step and the judgment step at least once, until the judgment of the track data of all the track data subset pairs in the preliminary track data set group is completed, and a plurality of judgment results are obtained;

[0035] Step S205, a determination step, determining whether the prepared track data set group meets a screening requirement based on all the above judgment results, wherein the screening requirement is that the difference between the track data in the first track data set and the corresponding track data in the second track data set in all the track data subset pairs in the prepared track data set group is less than the preset difference;

[0036] Step S206, repeating the first obtaining step, the first repeating step, and the determining step at least once until all the track data set groups are determined;

[0037] Step S207 : deleting the track data set groups that do not meet the screening requirements from all the track data set groups, and obtaining a plurality of track data set groups that meet the screening requirements.

[0038] In the above embodiment, before determining the overall matching degree of the matching method corresponding to each track dataset group by calculating the total mean square error corresponding to each track dataset group, it is necessary to preliminarily determine whether each track dataset group meets the screening conditions to preliminarily determine whether the matching method corresponding to each track dataset group has a high overall matching degree, so as to preliminarily screen out matching methods with a high overall matching degree.

[0039] Here, taking the determination of whether a track dataset group meets the screening requirements, that is, whether the overall matching degree of the matching method corresponding to the track dataset group meets the matching degree requirements of the preliminary screening stage as an example, if the difference between the track data in the first track dataset and the corresponding track data in the second track dataset of all track data subset pairs in the track dataset group is less than a preset difference, it is determined that the track dataset group meets the screening requirements, that is, it is determined that the matching method meets the matching degree requirements of the preliminary screening stage. If the difference between the track data in the first track dataset and the corresponding track data in the second track dataset in at least one track data subset pair in the track dataset group is greater than or equal to the preset difference, it is determined that the matching method does not meet the matching degree of the preliminary screening stage. In this case, there is no need to further determine the overall matching degree of the matching method based on the total mean square error corresponding to the track dataset group corresponding to the matching method. Preliminary screening can reduce the amount of calculation of the total mean square error sum, improve computational efficiency, and screen out matching methods with higher matching degrees, thereby increasing the probability of track matching, ensuring optimal track allocation when tracking target obstacles, and reducing track mismatches.

[0040] It should be noted that a track data includes lateral distance data, longitudinal distance data, lateral speed data, longitudinal speed data and motion state. For the convenience of description, the first track data represents the track data in the first track data set, and the second track data represents the track data in the second track data set. The specific process of judging whether the difference between the track data in the first track data set and the corresponding track data in the second track data set of the track data subset pair is less than the preset difference includes: judging whether the difference between the lateral distance data corresponding to the first track data and the second track data satisfies the preset lateral distance difference, judging whether the difference between the longitudinal distance data corresponding to the first track data and the second track data satisfies the preset longitudinal distance difference, judging whether the difference between the lateral speed data corresponding to the first track data and the second track data satisfies the preset lateral speed difference, judging whether the difference between the longitudinal speed data corresponding to the first track data and the second track data satisfies the preset longitudinal speed difference, and considering whether the motion state of the obstacle corresponding to the first track data and the second track data is the same motion state, and the motion state includes stationary and moving. When the difference between the track data corresponding to the first track data and the second track data is less than the preset difference and the motion state of the obstacle corresponding to the first track data and the second track data is the same, it is determined that the first track data and the second track data meet the similarity requirement for track data in the preliminary screening stage. Similarly, it can be determined whether other corresponding track data in the two track data sets meet the similarity requirement for track data in the first screening stage. Then, the number of track data in the first track data set and the second track data set that meet the similarity requirement for track data in the first screening stage is counted. In actual practice, when the number of track data in the two track data sets that meet the similarity requirement for track data in the first screening stage reaches 80%, it can be determined that the similarity of the two track data sets meets the similarity requirement for track data in the preliminary screening stage. When the number of track data in the two track data sets that meet the similarity requirement for track data in the first screening stage is less than 80%, it can be determined that the similarity of the two track data sets does not meet the similarity requirement for track data in the preliminary screening stage, that is, it is preliminarily determined that the similarity of the two tracks is low.

[0041] It should also be noted that the settings of the sizes of the four preset differences, namely, preset lateral distance difference, preset lateral distance difference, preset lateral distance difference and preset lateral distance difference, are related to the distance from the obstacle to the vehicle. The closer the distance from the obstacle to the vehicle, the higher the accuracy requirement, the smaller the preset difference is set; the farther the distance from the obstacle to the vehicle, the larger the preset difference is set.

[0042] Optionally, the present application does not limit the specific process of determining whether the above-mentioned preliminary track data set group meets the screening requirements based on all the above-mentioned judgment results, and any feasible method falls within the scope of protection of the present application.

[0043] For example, in another optional implementation, the above step S205 includes:

[0044] Step S2051 : if all the above judgment results are yes, it is determined that the above prepared track dataset group meets the above screening requirement; if at least one of the above judgment results is no, it is determined that the above prepared track dataset group does not meet the above screening requirement.

[0045] In the above embodiment, if the difference between the track data in the first track data set and the corresponding track data in the second track data set for all track data subset pairs in the track data set group is less than the preset difference, the track data set group is determined to meet the screening requirements, that is, it is determined that the matching method corresponding to the track data set group meets the matching degree requirements of the preliminary screening stage. If, in at least one track data subset pair in the track data set group, the difference between the track data in the first track data set and the corresponding track data in the second track data set is greater than or equal to the preset difference, it is determined that the matching method corresponding to the track data set group does not meet the matching degree requirements of the preliminary screening stage. In this case, there is no need to further determine the overall matching degree of the matching method corresponding to the track data set group based on the total mean square error corresponding to the track data set group. Preliminary screening can reduce the amount of calculation of the total mean square error sum, improve computational efficiency, and screen out matching methods with higher matching degrees, thereby increasing the probability of track matching, ensuring optimal track allocation when tracking target obstacles, and reducing track mismatches.

[0046] Step S102, calculating the sum of the mean square errors of all track data subset pairs in each track data set group to obtain a plurality of total mean square error sums, where the track data set group includes a plurality of track data subset pairs, one track data subset pair includes a first track data set and a second track data set, the first track data sets of any two track data subset pairs are different, and the second track data sets of any two track data subset pairs are different. The mean square error of the track data subset pair is the sum of the mean square errors of the track data in the track data subset pair, and the total mean square error sum corresponds one-to-one to the track data set group;

[0047] Among them, after screening out the track data set groups that meet the screening conditions, that is, preliminarily screening out the matching methods with higher overall matching degrees, further, the total mean square error corresponding to each track data set group is calculated to determine the overall matching degree of the matching methods corresponding to each track data set group. This can improve the track matching probability, ensure that the optimal track allocation is achieved when tracking the target obstacle, and reduce track mismatching.

[0048] Optionally, the present application does not limit the specific process of calculating the sum of the mean square errors of all track data subset pairs in each track data set group to obtain multiple total mean square error sums, and any feasible method falls within the scope of protection of the present application.

[0049] For example, in another optional implementation, the above step S102 includes:

[0050] Step S1021, a third acquisition step, acquiring a target track dataset group, wherein the target track dataset group is any one of the track dataset groups that meets the screening requirements;

[0051] Step S1022, a first calculation step, calculating the mean square error of each pair of track data subsets in the target track data set group to obtain a plurality of target mean square errors;

[0052] Step S1023, a second calculation step, calculating the sum of all the above target mean square errors to obtain a target total mean square error sum;

[0053] Repeat the third obtaining step, the first calculating step and the second calculating step at least once until the calculation of the sum of the mean square errors of all track data subset pairs in all the track data set groups is completed to obtain multiple total mean square error sums.

[0054] In the above embodiment, it is necessary to calculate the total mean square error sum corresponding to each track data set group to determine the overall matching degree of the matching method corresponding to each track data set group. Taking the calculation of the target total mean square error sum corresponding to the target track data set group as an example, first, it is necessary to calculate the mean square error of each track data subset pair in the target track data set group. The mean square error of each track data subset pair represents the similarity between the two track data corresponding to the track data subset pair. The smaller the mean square error, the higher the similarity. Then, the sum of the mean square error of each track data subset pair in the target track data set group is calculated to obtain the target total mean square error sum. The target total mean square error sum represents the overall matching degree of each pair of matched tracks in the matching method corresponding to the target track data set group.

[0055] Optionally, the present application does not limit the specific process of calculating the mean square error of each pair of track data subsets in the target track data set group to obtain multiple target mean square errors, and any feasible method falls within the scope of protection of the present application.

[0056] For example, in another optional implementation, the above step S1022 includes:

[0057] Step S10221, a fourth acquisition step, acquiring a target track data subset pair, wherein the target track data subset pair is any one of the track data subset pairs of the target track data set group;

[0058] Step S10222, a third calculation step, calculating the mean square error of the position data of the target track data subset pair to obtain a plurality of target position mean square errors, and calculating the mean square error of the speed data of the target track data subset pair to obtain a plurality of target speed mean square errors;

[0059] Step S10223, a fourth calculation step, calculating the sum of all the target position mean square errors to obtain a first target mean square error, and calculating the sum of all the target velocity mean square errors to obtain a second target mean square error;

[0060] Step S10224, a fifth calculation step, calculating the sum of the first target mean square error and the second target mean square error to obtain the target mean square error;

[0061] Step S10225, repeat the fourth acquisition step, the third calculation step, the fourth calculation step and the fifth calculation step at least once until the calculation of the mean square errors of all the track data subset pairs in the target track data set group is completed, and multiple target mean square errors are obtained.

[0062] In the above implementation manner, to calculate the target total mean square error and corresponding to the target track data subset pair, it is necessary to calculate the mean square error of each track data subset pair in the target track data subset pair. Taking the calculation of the mean square error of the target track data subset pair as an example, assuming that the target track data subset pair includes 2N track data, one track data includes position data and speed data, the position data includes lateral distance data and longitudinal distance data, and the speed data includes lateral speed data and longitudinal speed data, first, calculate the mean of the 2N lateral distance data, the mean of the 2N longitudinal distance data, the mean of the 2N lateral speed data, and the mean of the 2N longitudinal speed data in the target track data subset pair, and then calculate the mean square error of the 2N lateral distance data, the mean of the 2N longitudinal distance data in the target track data subset pair, and the mean square error of the 2N longitudinal distance data. Variance, mean square error of 2N lateral velocity data and mean square error of 2N longitudinal velocity data, then calculate the sum of mean square errors of 2N lateral distance data, the sum of mean square errors of 2N longitudinal distance data, the sum of mean square errors of 2N lateral velocity data and the sum of mean square errors of 2N longitudinal velocity data. Finally, calculate the sum of mean square errors of lateral distance data, the sum of mean square errors of longitudinal distance data, the sum of mean square errors of lateral velocity data and the sum of mean square errors of longitudinal velocity data to obtain the sum of mean square errors of 2N track data, that is, the sum of mean square errors of target track data subset pairs is obtained. The smaller the sum of mean square errors of target track data subset pairs, the higher the similarity of the two tracks corresponding to the target track data subset pairs. Taking the sum of mean square errors of lateral distance data of 2N track points as an example, is the mean of 2N horizontal distance data, through the formula Calculate the mean square error S of the lateral distance data of the i-th track point i , through the formula The sum of the mean square deviations d of the lateral distance data of the 2N track points is calculated. Similarly, the sum of the mean square deviations of the longitudinal distance data of the 2N track points, the sum of the mean square deviations of the lateral velocity data of the 2N track points, and the sum of the mean square deviations of the longitudinal velocity data of the 2N track points can be calculated.

[0063] Step S103, using a track fusion algorithm to fuse the two tracks corresponding to the optimal track data subset pair to obtain multiple optimal tracks, wherein the optimal track data subset pair is the track data subset pair with the smallest total mean square error and the corresponding track data set group, and the optimal track data subset pair corresponds one-to-one to the optimal track.

[0064] Among them, the total mean square error corresponding to each track data set group represents the overall matching degree of the matching method corresponding to each track data set group. The higher the total mean square error corresponding to a track data set group, the higher the overall matching degree of the matching method corresponding to the track data set group. The overall matching degree of the matching method corresponding to the track data set group with the smallest total mean square error is the highest. The two tracks corresponding to each optimal track data subset pair of the track data set group with the smallest total mean square error are the two tracks of one of the target obstacles detected by the two sensors respectively. The two tracks are fused by the track fusion algorithm, that is, the track after the two tracks of the target obstacle detected by the two sensors are associated is obtained.

[0065] In addition, in an optional embodiment, before the above step S103, the above method further includes:

[0066] Step S301, determining the minimum value of all the above total mean square errors to obtain the optimal total mean square error;

[0067] Step S302 : determining the optimal track data subset pair according to the optimal total mean square error sum, wherein the optimal track data subset pair is the track data subset pair of the track data set group corresponding to the optimal total mean square error sum.

[0068] In the above embodiment, the total mean square error and the sum corresponding to each track data set group represent the overall matching degree of the matching method corresponding to each track data set group. The higher the total mean square error corresponding to a track data set group, the higher the overall matching degree of the matching method corresponding to the track data set group. The present application generates a similarity matrix by taking the inverse of the sum of the mean square errors of each track data subset pair as the element of the matrix, and then solves it based on the KM matching algorithm to obtain the highest weight, and further obtains the optimal total mean square error and the sum. The overall matching degree of the matching method corresponding to the track data set group corresponding to the optimal total mean square error and the sum is the highest. The two tracks corresponding to each optimal track data subset pair of the track data set group corresponding to the optimal total mean square error and the two tracks of one of the target obstacles detected by the two sensors respectively. The KM algorithm can find the global optimal solution under the constraints of the complex scene of multiple target obstacles, thereby ensuring the correct association rate of the tracks of the target obstacles.

[0069] In another optional embodiment, the track fusion method of the present application is executed as follows: Step 1: spatial synchronization and track preprocessing are performed on the sensor tracks respectively; Step 2: the track data of the sensors are synchronized to the same fusion moment in combination with time and motion mode, and the track data of the target obstacle is updated; Step 3: the track data of the target obstacle detected by sensor 1 and sensor 2 are obtained; Step 4: the track data of the target obstacle detected by sensor 1 and sensor 2 are fused, and a new global track is generated based on the fusion result; Step 5: the new sensor track data is obtained and fused with the generated global track; Step 6: the track information of the target obstacle after the track data of all sensors are fused is output. This method can give full play to the advantages of multi-sensor fusion, increase the robustness of the system, achieve effective tracking of the target obstacle, solve the error problem caused by changes in sensor accuracy, and thus improve the accuracy of track fusion. In an environment where the target obstacle tracks are dense or sparse, the target obstacle tracks can be effectively fused, thereby improving the accuracy of the sensor in detecting the target obstacle tracks.

[0070] In another optional embodiment, each track has a track number to ensure that the tracking ID of the same obstacle remains unchanged.

[0071] In the above track fusion method, first, track data of multiple tracks of the first sensor and the second sensor are obtained respectively to obtain multiple first track data sets and multiple second track data sets, wherein the above first track data set includes multiple track data of an obstacle detected by the above first sensor within a time period, and the above second track data set includes multiple track data of an obstacle detected by the above second sensor within the above time period, the above time period includes multiple time nodes, and the above track data correspond to the above time nodes one by one; then, the sum of the mean square errors of all track data subset pairs in each track data set group is calculated to obtain multiple total mean square error sums, the above track data set group includes multiple track data subset pairs, and one above track data set is a plurality of track data subset pairs. The track data subset pair includes the above-mentioned first track data set and the above-mentioned second track data set, the above-mentioned first track data sets of any two of the above-mentioned track data subset pairs are different, the above-mentioned second track data sets of any two of the above-mentioned track data subset pairs are different, the mean square error of the above-mentioned track data subset pair is the sum of the mean square errors of the above-mentioned track data in the above-mentioned track data subset pair, and the above-mentioned total mean square error sum corresponds one-to-one to the above-mentioned track data set group; finally, a track fusion algorithm is used to fuse the two above-mentioned tracks corresponding to the optimal track data subset pair to obtain multiple optimal tracks, the above-mentioned optimal track data subset pair is the above-mentioned track data subset pair of the above-mentioned track data set group corresponding to the minimum sum of the above-mentioned total mean square error, and the above-mentioned optimal track data subset pair corresponds one-to-one to the above-mentioned optimal track. In this method, a track data subset pair is a set of track data of a track detected by a first sensor and a track data of a track detected by a second sensor. The mean square error of the track data subset pair reflects the similarity of the two tracks detected by the two sensors. The tracks detected by the first sensor are matched with the tracks detected by the second sensor pairwise to obtain multiple track data sets. Each track data set group corresponds to a matching method. The total mean square error sum of a track data set group reflects the overall similarity of each matched track pair in a matching method. The matching method with the highest overall similarity of each matched track pair is determined based on the minimum total mean square error sum. The two tracks corresponding to the optimal track data set group with the minimum total mean square error sum are the two tracks of one target obstacle detected by each sensor. These two tracks are fused using a track fusion algorithm to obtain the two correlated tracks of the target obstacle detected by the two sensors. This method solves the problem of low accuracy of track similarity calculation methods used in track correlation in the prior art.

[0072] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0073] The present application also provides a track fusion device. It should be noted that the track fusion device of the present application can be used to execute the track fusion method provided in the present application. The track fusion device provided in the present application is introduced below.

[0074] Figure 2 Schematic diagram of a track fusion device according to an embodiment of the present application. Figure 2 As shown, the device includes:

[0075] an acquiring unit 10, respectively acquiring track data of a plurality of tracks of a first sensor and a second sensor, to obtain a plurality of first track data sets and a plurality of second track data sets, wherein the first track data set includes a plurality of track data of an obstacle detected by the first sensor within a time period, and the second track data set includes a plurality of track data of an obstacle detected by the second sensor within the time period, wherein the time period includes a plurality of time nodes, and the track data correspond to the time nodes in a one-to-one manner;

[0076] Among them, the accuracy of the track similarity calculation method used in track association in the prior art is low. In order to solve the above problem, it is first necessary to obtain track data of multiple tracks of the first sensor and track data of multiple tracks of the second sensor, and then solve the problem of low accuracy of the track similarity calculation method used in track association in the prior art.

[0077] It should be noted that in complex scenarios with multiple target obstacles, the track data detected by each sensor will deviate, leading to the introduction of noise and interference. With a lot of clutter, only considering the track data of a single time period will result in incorrect or unmatched target obstacles. Therefore, it is necessary to preprocess the track data detected by each sensor. Taking the preprocessing of track data of an obstacle detected by the first sensor as an example, a track consists of multiple track points. The track data of a track point includes position data and speed data. The position data includes lateral distance data and longitudinal distance data, and the speed data includes lateral speed data and longitudinal speed data. The first track data set corresponding to the track includes the track data of all track points in the track. The first track data set corresponding to the track in which the sensor detected the obstacle in the previous time period and the first track data set corresponding to the track in which the sensor detected the obstacle in the current time period are obtained. The track points where the track data jump in the two tracks are compared, and the track data corresponding to these unstable track points are deleted from the first track data set of the current time period.

[0078] It should also be noted that the number of track data in the first track data set and the number of track data in the second track data set are both equal to a predetermined number. The predetermined number of track data for each track ensures that each track has a sufficient amount of track data, thereby ensuring the accuracy of the calculated total mean square error (MSS), and improving the correct association rate of target obstacle tracks.

[0079] In addition, in an optional embodiment, the track fusion device further includes:

[0080] A first acquiring unit is configured to acquire a preliminary track data set group, wherein the preliminary track data set group is any one of the aforementioned track data set groups;

[0081] a second acquiring unit, acquiring a prepared track data subset pair, wherein the prepared track data subset pair is any one of the track data subset pairs in the prepared track data set group, wherein the track data in the first track data set of the prepared track data subset pair corresponds one-to-one to the track data in the second track data set of the prepared track data subset pair;

[0082] a judgment unit, judging whether a difference between the track data of the first track data set in the pair of preliminary track data subsets and the corresponding track data in the second track data set is less than a preset difference, and obtaining a plurality of judgment results;

[0083] a first iterative unit, repeating the second obtaining step and the judging step at least once until the judging of the track data of all the track data subset pairs in the preliminary track data set is completed, and a plurality of the judging results are obtained;

[0084] a first determining unit, determining, based on all the judgment results, whether the prepared track data set group meets a screening requirement, wherein the screening requirement is that a difference between the track data in the first track data set and the corresponding track data in the second track data set in all pairs of the track data subsets in the prepared track data set group is less than a preset difference;

[0085] a second iterative unit, repeating the first obtaining step, the first repeating step, and the determining step at least once until all the track data set groups are determined;

[0086] The deleting unit deletes the track data set groups that do not meet the screening requirements from all the track data set groups, thereby obtaining a plurality of track data set groups that meet the screening requirements.

[0087] In the above embodiment, before determining the overall matching degree of the matching method corresponding to each track dataset group by calculating the total mean square error corresponding to each track dataset group, it is necessary to preliminarily determine whether each track dataset group meets the screening conditions to preliminarily determine whether the matching method corresponding to each track dataset group has a high overall matching degree, so as to preliminarily screen out matching methods with a high overall matching degree.

[0088] Here, taking the determination of whether a track dataset group meets the screening requirements, that is, whether the overall matching degree of the matching method corresponding to the track dataset group meets the matching degree requirements of the preliminary screening stage as an example, if the difference between the track data in the first track dataset and the corresponding track data in the second track dataset of all track data subset pairs in the track dataset group is less than a preset difference, it is determined that the track dataset group meets the screening requirements, that is, it is determined that the matching method meets the matching degree requirements of the preliminary screening stage. If the difference between the track data in the first track dataset and the corresponding track data in the second track dataset in at least one track data subset pair in the track dataset group is greater than or equal to the preset difference, it is determined that the matching method does not meet the matching degree of the preliminary screening stage. In this case, there is no need to further determine the overall matching degree of the matching method based on the total mean square error corresponding to the track dataset group corresponding to the matching method. Preliminary screening can reduce the amount of calculation of the total mean square error sum, improve computational efficiency, and screen out matching methods with higher matching degrees, thereby increasing the probability of track matching, ensuring optimal track allocation when tracking target obstacles, and reducing track mismatches.

[0089] It should be noted that a track data includes lateral distance data, longitudinal distance data, lateral speed data, longitudinal speed data and motion state. For the convenience of description, the first track data represents the track data in the first track data set, and the second track data represents the track data in the second track data set. The specific process of judging whether the difference between the track data in the first track data set and the corresponding track data in the second track data set of the track data subset pair is less than the preset difference includes: judging whether the difference between the lateral distance data corresponding to the first track data and the second track data satisfies the preset lateral distance difference, judging whether the difference between the longitudinal distance data corresponding to the first track data and the second track data satisfies the preset longitudinal distance difference, judging whether the difference between the lateral speed data corresponding to the first track data and the second track data satisfies the preset lateral speed difference, judging whether the difference between the longitudinal speed data corresponding to the first track data and the second track data satisfies the preset longitudinal speed difference, and considering whether the motion state of the obstacle corresponding to the first track data and the second track data is the same motion state, and the motion state includes stationary and moving. When the difference between the track data corresponding to the first track data and the second track data is less than the preset difference and the motion state of the obstacle corresponding to the first track data and the second track data is the same, it is determined that the first track data and the second track data meet the similarity requirement for track data in the preliminary screening stage. Similarly, it can be determined whether other corresponding track data in the two track data sets meet the similarity requirement for track data in the first screening stage. Then, the number of track data in the first track data set and the second track data set that meet the similarity requirement for track data in the first screening stage is counted. In actual practice, when the number of track data in the two track data sets that meet the similarity requirement for track data in the first screening stage reaches 80%, it can be determined that the similarity of the two track data sets meets the similarity requirement for track data in the preliminary screening stage. When the number of track data in the two track data sets that meet the similarity requirement for track data in the first screening stage is less than 80%, it can be determined that the similarity of the two track data sets does not meet the similarity requirement for track data in the preliminary screening stage, that is, it is preliminarily determined that the similarity of the two tracks is low.

[0090] It should also be noted that the settings of the sizes of the four preset differences, namely, preset lateral distance difference, preset lateral distance difference, preset lateral distance difference and preset lateral distance difference, are related to the distance from the obstacle to the vehicle. The closer the distance from the obstacle to the vehicle, the higher the accuracy requirement, the smaller the preset difference is set; the farther the distance from the obstacle to the vehicle, the larger the preset difference is set.

[0091] Optionally, the present application does not limit the specific process of determining whether the above-mentioned preliminary track data set group meets the screening requirements based on all the above-mentioned judgment results, and any feasible method falls within the scope of protection of the present application.

[0092] For example, in another optional implementation, the determining unit includes:

[0093] The determination module determines that the prepared track dataset group meets the screening requirement when all the above judgment results are yes, and determines that the prepared track dataset group does not meet the screening requirement when at least one of the above judgment results is no.

[0094] In the above embodiment, if the difference between the track data in the first track data set and the corresponding track data in the second track data set for all track data subset pairs in the track data set group is less than the preset difference, the track data set group is determined to meet the screening requirements, that is, it is determined that the matching method corresponding to the track data set group meets the matching degree requirements of the preliminary screening stage. If, in at least one track data subset pair in the track data set group, the difference between the track data in the first track data set and the corresponding track data in the second track data set is greater than or equal to the preset difference, it is determined that the matching method corresponding to the track data set group does not meet the matching degree requirements of the preliminary screening stage. In this case, there is no need to further determine the overall matching degree of the matching method corresponding to the track data set group based on the total mean square error corresponding to the track data set group. Preliminary screening can reduce the amount of calculation of the total mean square error sum, improve computational efficiency, and screen out matching methods with higher matching degrees, thereby increasing the probability of track matching, ensuring optimal track allocation when tracking target obstacles, and reducing track mismatches.

[0095] The calculation unit 20 calculates the sum of the mean square errors of all track data subset pairs in each track data set group to obtain multiple total mean square error sums, where the track data set group includes multiple track data subset pairs, one track data subset pair includes the first track data set and the second track data set, the first track data sets of any two track data subset pairs are different, and the second track data sets of any two track data subset pairs are different. The mean square error of the track data subset pair is the sum of the mean square errors of the track data in the track data subset pair, and the total mean square error sums correspond one-to-one to the track data set group.

[0096] Among them, after screening out the track data set groups that meet the screening conditions, that is, preliminarily screening out the matching methods with higher overall matching degrees, further, the total mean square error corresponding to each track data set group is calculated to determine the overall matching degree of the matching methods corresponding to each track data set group. This can improve the track matching probability, ensure that the optimal track allocation is achieved when tracking the target obstacle, and reduce track mismatching.

[0097] Optionally, the present application does not limit the specific process of calculating the sum of the mean square errors of all track data subset pairs in each track data set group to obtain multiple total mean square error sums, and any feasible method falls within the scope of protection of the present application.

[0098] For example, in another optional embodiment, the calculation unit includes:

[0099] A first acquisition module acquires a target track dataset group, wherein the target track dataset group is any one of the track dataset groups that meets the screening requirements;

[0100] A first calculation module calculates the mean square error of each pair of the track data subsets in the target track data set group to obtain multiple target mean square errors;

[0101] The second calculation module calculates the sum of all the above target mean square deviations to obtain the target total mean square deviation sum;

[0102] The iterative module repeats the third acquisition step, the first calculation step, and the second calculation step at least once until the calculation of the sum of the mean square errors of all track data subset pairs in all the track data set groups is completed to obtain multiple total mean square error sums.

[0103] In the above embodiment, it is necessary to calculate the total mean square error sum corresponding to each track data set group to determine the overall matching degree of the matching method corresponding to each track data set group. Taking the calculation of the target total mean square error sum corresponding to the target track data set group as an example, first, it is necessary to calculate the mean square error of each track data subset pair in the target track data set group. The mean square error of each track data subset pair represents the similarity between the two track data corresponding to the track data subset pair. The smaller the mean square error, the higher the similarity. Then, the sum of the mean square error of each track data subset pair in the target track data set group is calculated to obtain the target total mean square error sum. The target total mean square error sum represents the overall matching degree of each pair of matched tracks in the matching method corresponding to the target track data set group.

[0104] Optionally, the present application does not limit the specific process of calculating the mean square error of each pair of track data subsets in the target track data set group to obtain multiple target mean square errors, and any feasible method falls within the scope of protection of the present application.

[0105] For example, in another optional embodiment, the first calculation module includes:

[0106] An acquisition submodule, which acquires a target track data subset pair, wherein the target track data subset pair is any one of the track data subset pairs in the target track data set group;

[0107] A first calculation submodule calculates the mean square error of the position data of the target track data subset pair to obtain a plurality of target position mean square errors, and calculates the mean square error of the speed data of the target track data subset pair to obtain a plurality of target speed mean square errors;

[0108] The second calculation submodule calculates the sum of all the target position mean square errors to obtain a first target mean square error, and calculates the sum of all the target velocity mean square errors to obtain a second target mean square error;

[0109] A third calculation submodule calculates the sum of the first target mean square error and the second target mean square error to obtain the target mean square error;

[0110] The iterative submodule repeats the fourth acquisition step, the third calculation step, the fourth calculation step, and the fifth calculation step at least once until the calculation of the mean square errors of all the track data subset pairs in the target track data set group is completed, thereby obtaining a plurality of target mean square errors.

[0111] In the above implementation manner, to calculate the target total mean square error and corresponding to the target track data subset pair, it is necessary to calculate the mean square error of each track data subset pair in the target track data subset pair. Taking the calculation of the mean square error of the target track data subset pair as an example, assuming that the target track data subset pair includes 2N track data, one track data includes position data and speed data, the position data includes lateral distance data and longitudinal distance data, and the speed data includes lateral speed data and longitudinal speed data, first, calculate the mean of the 2N lateral distance data, the mean of the 2N longitudinal distance data, the mean of the 2N lateral speed data, and the mean of the 2N longitudinal speed data in the target track data subset pair, and then calculate the mean square error of the 2N lateral distance data, the mean of the 2N longitudinal distance data in the target track data subset pair, and the mean square error of the 2N longitudinal distance data. Variance, mean square error of 2N lateral velocity data and mean square error of 2N longitudinal velocity data, then calculate the sum of mean square errors of 2N lateral distance data, the sum of mean square errors of 2N longitudinal distance data, the sum of mean square errors of 2N lateral velocity data and the sum of mean square errors of 2N longitudinal velocity data. Finally, calculate the sum of mean square errors of lateral distance data, the sum of mean square errors of longitudinal distance data, the sum of mean square errors of lateral velocity data and the sum of mean square errors of longitudinal velocity data to obtain the sum of mean square errors of 2N track data, that is, the sum of mean square errors of target track data subset pairs is obtained. The smaller the sum of mean square errors of target track data subset pairs, the higher the similarity of the two tracks corresponding to the target track data subset pairs. Taking the sum of mean square errors of lateral distance data of 2N track points as an example, is the mean of 2N horizontal distance data, through the formula Calculate the mean square error S of the lateral distance data of the i-th track point i , through the formula The sum of the mean square deviations d of the lateral distance data of the 2N track points is calculated. Similarly, the sum of the mean square deviations of the longitudinal distance data of the 2N track points, the sum of the mean square deviations of the lateral velocity data of the 2N track points, and the sum of the mean square deviations of the longitudinal velocity data of the 2N track points can be calculated.

[0112] The fusion unit 30 adopts a track fusion algorithm to fuse the two tracks corresponding to the optimal track data subset pair to obtain multiple optimal tracks, wherein the optimal track data subset pair is the track data subset pair with the smallest total mean square error and the corresponding track data set group, and the optimal track data subset pair corresponds one-to-one to the optimal track.

[0113] Among them, the total mean square error corresponding to each track data set group represents the overall matching degree of the matching method corresponding to each track data set group. The higher the total mean square error corresponding to a track data set group, the higher the overall matching degree of the matching method corresponding to the track data set group. The overall matching degree of the matching method corresponding to the track data set group with the smallest total mean square error is the highest. The two tracks corresponding to each optimal track data subset pair of the track data set group with the smallest total mean square error are the two tracks of one of the target obstacles detected by the two sensors respectively. The two tracks are fused by the track fusion algorithm, that is, the track after the two tracks of the target obstacle detected by the two sensors are associated is obtained.

[0114] In addition, in an optional embodiment, the track fusion device further includes:

[0115] The second determining unit determines the minimum value among all the above total mean square error sums to obtain the optimal total mean square error sum;

[0116] A third determining unit determines the optimal track data subset pair according to the optimal total mean square error sum, where the optimal track data subset pair is the track data subset pair of the track data set group corresponding to the optimal total mean square error sum.

[0117] In the above embodiment, the total mean square error and the sum corresponding to each track data set group represent the overall matching degree of the matching method corresponding to each track data set group. The higher the total mean square error corresponding to a track data set group, the higher the overall matching degree of the matching method corresponding to the track data set group. The present application generates a similarity matrix by taking the inverse of the sum of the mean square errors of each track data subset pair as the element of the matrix, and then solves it based on the KM matching algorithm to obtain the highest weight, and further obtains the optimal total mean square error and the sum. The overall matching degree of the matching method corresponding to the track data set group corresponding to the optimal total mean square error and the sum is the highest. The two tracks corresponding to each optimal track data subset pair of the track data set group corresponding to the optimal total mean square error and the two tracks of one of the target obstacles detected by the two sensors respectively. The KM algorithm can find the global optimal solution under the constraints of the complex scene of multiple target obstacles, thereby ensuring the correct association rate of the tracks of the target obstacles.

[0118] In another optional embodiment, the track fusion method of the present application is executed as follows: Step 1: spatial synchronization and track preprocessing are performed on the sensor tracks respectively; Step 2: the track data of the sensors are synchronized to the same fusion moment in combination with time and motion mode, and the track data of the target obstacle is updated; Step 3: the track data of the target obstacle detected by sensor 1 and sensor 2 are obtained; Step 4: the track data of the target obstacle detected by sensor 1 and sensor 2 are fused, and a new global track is generated based on the fusion result; Step 5: the new sensor track data is obtained and fused with the generated global track; Step 6: the track information of the target obstacle after the track data of all sensors are fused is output. This method can give full play to the advantages of multi-sensor fusion, increase the robustness of the system, achieve effective tracking of the target obstacle, solve the error problem caused by changes in sensor accuracy, and thus improve the accuracy of track fusion. In an environment where the target obstacle tracks are dense or sparse, the target obstacle tracks can be effectively fused, thereby improving the accuracy of the sensor in detecting the target obstacle tracks.

[0119] In another optional embodiment, each track has a track number to ensure that the tracking ID of the same obstacle remains unchanged.

[0120] In the above-mentioned track fusion device, the acquisition unit acquires the track data of multiple tracks of the first sensor and the second sensor respectively, and obtains multiple first track data sets and multiple second track data sets, the above-mentioned first track data set includes multiple track data of an obstacle detected by the above-mentioned first sensor within a time period, and the above-mentioned second track data set includes multiple track data of an obstacle detected by the above-mentioned second sensor within the above-mentioned time period, the above-mentioned time period includes multiple time nodes, and the above-mentioned track data correspond to the above-mentioned time nodes one by one; the calculation unit calculates the sum of the mean square errors of all track data subset pairs in each track data set group to obtain multiple total mean square error sums, the above-mentioned track data set group includes multiple track data subset pairs, and one above-mentioned The track data subset pair includes the above-mentioned first track data set and the above-mentioned second track data set, the above-mentioned first track data sets of any two of the above-mentioned track data subset pairs are different, the above-mentioned second track data sets of any two of the above-mentioned track data subset pairs are different, the mean square error of the above-mentioned track data subset pair is the sum of the mean square errors of the above-mentioned track data in the above-mentioned track data subset pair, and the above-mentioned total mean square error sum corresponds one-to-one to the above-mentioned track data set group; the fusion unit adopts a track fusion algorithm to fuse the two above-mentioned tracks corresponding to the optimal track data subset pair to obtain multiple optimal tracks, the above-mentioned optimal track data subset pair is the above-mentioned track data subset pair of the above-mentioned track data set group corresponding to the minimum sum of the above-mentioned total mean square error, and the above-mentioned optimal track data subset pair corresponds one-to-one to the above-mentioned optimal track. The device's track data subset pairs are a collection of track data from a track detected by a first sensor and a track detected by a second sensor. The mean squared error (MSD) of the track data subset pairs reflects the similarity between the two tracks detected by the two sensors. Pairwise matching is performed on the tracks detected by the first sensor and the tracks detected by the second sensor to obtain multiple track data sets, each corresponding to a matching method. The total mean squared error (MSD) of a track data set group reflects the overall similarity between each pair of matched track pairs within a matching method. The matching method with the highest overall similarity between each pair of matched track pairs is determined based on the minimum total mean squared error. The optimal track data set corresponding to the minimum total mean squared error corresponds to two tracks of a target obstacle detected by each of the two sensors. These two tracks are fused using a track fusion algorithm to obtain the associated track of the target obstacle detected by the two sensors. This device solves the problem of low accuracy in the track similarity calculation methods used in prior art track association.

[0121] The above-mentioned track fusion device includes a processor and a memory. The above-mentioned acquisition unit, calculation unit and fusion unit are all stored in the memory as program units, and the processor executes the above-mentioned program units stored in the memory to realize corresponding functions.

[0122] The processor includes a core that retrieves the corresponding program unit from the memory. One or more cores can be provided, and kernel parameters can be adjusted to address the low accuracy of the track similarity calculation method used in track association in the prior art.

[0123] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM, and the memory includes at least one memory chip.

[0124] An embodiment of the present application further provides a processor, which is used to run a program, wherein the track fusion method is executed when the program is run.

[0125] An embodiment of the present application further provides a track fusion system, comprising: one or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include instructions for executing any one of the above methods. When the processor executes the program, at least the following steps are implemented:

[0126] Step S101: acquiring track data of a plurality of tracks of a first sensor and a second sensor, respectively, to obtain a plurality of first track data sets and a plurality of second track data sets, wherein the first track data set includes a plurality of track data of an obstacle detected by the first sensor within a time period, and the second track data set includes a plurality of track data of an obstacle detected by the second sensor within the time period, wherein the time period includes a plurality of time nodes, and the track data correspond to the time nodes in a one-to-one manner;

[0127] Step S102, calculating the sum of the mean square errors of all track data subset pairs in each track data set group to obtain a plurality of total mean square error sums, where the track data set group includes a plurality of track data subset pairs, one track data subset pair includes a first track data set and a second track data set, the first track data sets of any two track data subset pairs are different, and the second track data sets of any two track data subset pairs are different. The mean square error of the track data subset pair is the sum of the mean square errors of the track data in the track data subset pair, and the total mean square error sum corresponds one-to-one to the track data set group;

[0128] Step S103, using a track fusion algorithm to fuse the two tracks corresponding to the optimal track data subset pair to obtain multiple optimal tracks, wherein the optimal track data subset pair is the track data subset pair with the smallest total mean square error and the corresponding track data set group, and the optimal track data subset pair corresponds one-to-one to the optimal track.

[0129] The present application also provides a computer program product, which, when executed on a data processing device, is adapted to execute a program for initializing at least the following method steps:

[0130] Step S101: acquiring track data of a plurality of tracks of a first sensor and a second sensor, respectively, to obtain a plurality of first track data sets and a plurality of second track data sets, wherein the first track data set includes a plurality of track data of an obstacle detected by the first sensor within a time period, and the second track data set includes a plurality of track data of an obstacle detected by the second sensor within the time period, wherein the time period includes a plurality of time nodes, and the track data correspond to the time nodes in a one-to-one manner;

[0131] Step S102, calculating the sum of the mean square errors of all track data subset pairs in each track data set group to obtain a plurality of total mean square error sums, where the track data set group includes a plurality of track data subset pairs, one track data subset pair includes a first track data set and a second track data set, the first track data sets of any two track data subset pairs are different, and the second track data sets of any two track data subset pairs are different. The mean square error of the track data subset pair is the sum of the mean square errors of the track data in the track data subset pair, and the total mean square error sum corresponds one-to-one to the track data set group;

[0132] Step S103, using a track fusion algorithm to fuse the two tracks corresponding to the optimal track data subset pair to obtain multiple optimal tracks, wherein the optimal track data subset pair is the track data subset pair with the smallest total mean square error and the corresponding track data set group, and the optimal track data subset pair corresponds one-to-one to the optimal track.

[0133] In the above embodiments of the present application, the description of each embodiment has its own focus. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0134] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only exemplary. For example, the division of the above-mentioned units can be a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0135] The units described above as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0136] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0137] If the above-mentioned integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the above-mentioned methods of each embodiment of the present application. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), mobile hard disk, magnetic disk or optical disk, etc. Various media that can store program codes.

[0138] From the above description, it can be seen that the above embodiments of the present application achieve the following technical effects:

[0139] 1) In the track fusion method of the present application, first, track data of multiple tracks of the first sensor and the second sensor are obtained respectively to obtain multiple first track data sets and multiple second track data sets, wherein the first track data set includes multiple track data of an obstacle detected by the first sensor within a time period, and the second track data set includes multiple track data of an obstacle detected by the second sensor within the time period, wherein the time period includes multiple time nodes, and the track data correspond to the time nodes one by one; then, the sum of the mean square errors of all track data subset pairs in each track data set group is calculated to obtain multiple total mean square error sums, wherein the track data set group includes multiple track data subset pairs, and a The above-mentioned track data subset pair includes the above-mentioned first track data set and the above-mentioned second track data set, the above-mentioned first track data sets of any two of the above-mentioned track data subset pairs are different, the above-mentioned second track data sets of any two of the above-mentioned track data subset pairs are different, the mean square error of the above-mentioned track data subset pair is the sum of the mean square errors of the above-mentioned track data in the above-mentioned track data subset pair, and the above-mentioned total mean square error sum corresponds one-to-one to the above-mentioned track data set group; finally, a track fusion algorithm is used to fuse the two above-mentioned tracks corresponding to the optimal track data subset pair to obtain multiple optimal tracks, the above-mentioned optimal track data subset pair is the above-mentioned track data subset pair of the above-mentioned track data set group corresponding to the minimum sum of the above-mentioned total mean square error, and the above-mentioned optimal track data subset pair corresponds one-to-one to the above-mentioned optimal track. In this method, a track data subset pair is a set of track data of a track detected by a first sensor and a track data of a track detected by a second sensor. The mean square error of the track data subset pair reflects the similarity of the two tracks detected by the two sensors. The tracks detected by the first sensor are matched with the tracks detected by the second sensor pairwise to obtain multiple track data sets. Each track data set group corresponds to a matching method. The total mean square error sum of a track data set group reflects the overall similarity of each matched track pair in a matching method. The matching method with the highest overall similarity of each matched track pair is determined based on the minimum total mean square error sum. The two tracks corresponding to the optimal track data set group with the minimum total mean square error sum are the two tracks of one target obstacle detected by each sensor. These two tracks are fused using a track fusion algorithm to obtain the two correlated tracks of the target obstacle detected by the two sensors. This method solves the problem of low accuracy of track similarity calculation methods used in track correlation in the prior art.

[0140] 2) In the track fusion device of the present application, the acquisition unit acquires track data of multiple tracks of the first sensor and the second sensor respectively to obtain multiple first track data sets and multiple second track data sets, wherein the first track data set includes multiple track data of an obstacle detected by the first sensor within a time period, and the second track data set includes multiple track data of an obstacle detected by the second sensor within the time period, wherein the time period includes multiple time nodes, and the track data correspond to the time nodes one by one; the calculation unit calculates the sum of the mean square errors of all track data subset pairs in each track data set group to obtain multiple total mean square error sums, wherein the track data set group includes multiple track data subset pairs, and one The above-mentioned track data subset pairs include the above-mentioned first track data set and the above-mentioned second track data set, the above-mentioned first track data sets of any two of the above-mentioned track data subset pairs are different, the above-mentioned second track data sets of any two of the above-mentioned track data subset pairs are different, the mean square error of the above-mentioned track data subset pairs is the sum of the mean square errors of the above-mentioned track data in the above-mentioned track data subset pairs, and the above-mentioned total mean square error sum corresponds one-to-one to the above-mentioned track data set group; the fusion unit adopts a track fusion algorithm to fuse the two above-mentioned tracks corresponding to the optimal track data subset pair to obtain multiple optimal tracks, the above-mentioned optimal track data subset pair is the above-mentioned track data set group corresponding to the minimum sum of the above-mentioned total mean square error, and the above-mentioned optimal track data subset pair corresponds one-to-one to the above-mentioned optimal track. The device's track data subset pairs are a collection of track data from a track detected by a first sensor and a track detected by a second sensor. The mean squared error (MSD) of the track data subset pairs reflects the similarity between the two tracks detected by the two sensors. Pairwise matching is performed on the tracks detected by the first sensor and the tracks detected by the second sensor to obtain multiple track data sets, each corresponding to a matching method. The total mean squared error (MSD) of a track data set group reflects the overall similarity between each pair of matched track pairs within a matching method. The matching method with the highest overall similarity between each pair of matched track pairs is determined based on the minimum total mean squared error. The optimal track data set corresponding to the minimum total mean squared error corresponds to two tracks of a target obstacle detected by each of the two sensors. These two tracks are fused using a track fusion algorithm to obtain the associated track of the target obstacle detected by the two sensors. This device solves the problem of low accuracy in the track similarity calculation methods used in prior art track association.

[0141] 3) The track fusion system of the present application includes: one or more processors, a memory and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the one or more processors, and the one or more programs include functions for executing any one of the above methods. The system's track data subset pairs are a collection of track data from a track detected by a first sensor and a track detected by a second sensor. The mean squared error (MSD) of the track data subset pairs reflects the similarity between the two tracks detected by the two sensors. Pairwise matching is performed on the tracks detected by the first sensor and the second sensor to obtain multiple track data sets, each corresponding to a matching method. The total mean squared error (MSD) of a track data set group reflects the overall similarity between each pair of matched track pairs within a matching method. The matching method with the highest overall similarity between each pair of matched track pairs is determined based on the minimum total mean squared error. The optimal track data set corresponding to the minimum total mean squared error corresponds to two tracks of a target obstacle detected by each sensor. These two tracks are fused using a track fusion algorithm to obtain the associated track of the target obstacle detected by the two sensors. This system solves the problem of low accuracy in the track similarity calculation methods used in track association in the prior art.

[0142] The above description is merely a preferred embodiment of the present application and is not intended to limit the present application. Various modifications and variations are possible for those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present application shall be included within the scope of protection of the present application.

Claims

1. A track fusion method, characterized in that: The method comprises: Acquiring track data of a plurality of tracks of a first sensor and a second sensor respectively to obtain a plurality of first track data sets and a plurality of second track data sets, wherein the first track data set includes a plurality of track data of an obstacle detected by the first sensor within a time period, and the second track data set includes a plurality of track data of an obstacle detected by the second sensor within the time period, wherein the time period includes a plurality of time nodes, and the track data correspond to the time nodes in a one-to-one manner; Calculating the sum of the mean square errors of all track data subset pairs in each track data set group to obtain multiple total mean square error sums, where the track data set group includes multiple track data subset pairs, one track data subset pair includes one first track data set and one second track data set, the first track data sets of any two track data subset pairs are different, and the second track data sets of any two track data subset pairs are different, the mean square error of the track data subset pair is the sum of the mean square errors of the track data in the track data subset pair, and the total mean square error sum corresponds one-to-one to the track data set group; A track fusion algorithm is used to fuse the two tracks corresponding to the optimal track data subset pair to obtain multiple optimal tracks, wherein the optimal track data subset pair is the track data subset pair with the smallest total mean square error and the corresponding track data set group, and the optimal track data subset pair corresponds one-to-one to the optimal track.

2. The method according to claim 1, characterized in that After respectively acquiring track data of a plurality of tracks of the first sensor and the second sensor to obtain a plurality of first track data sets and a plurality of second track data sets, the method further includes: A first acquisition step is to acquire a preliminary track dataset group, wherein the preliminary track dataset group is any one of the track dataset groups; a second acquiring step of acquiring a pair of preliminary track data subsets, wherein the pair of preliminary track data subsets is any one of the pair of track data subsets in the preliminary track data set group, wherein the track data in the first track data set of the preliminary track data subset pair corresponds one-to-one to the track data in the second track data set of the preliminary track data subset pair; a judging step of judging whether a difference between the track data of the first track data set in the pair of preliminary track data subsets and the corresponding track data in the second track data set is less than a preset difference, and obtaining a plurality of judging results; a first repetition step of repeating the second acquisition step and the judgment step at least once until the judgment of the track data of all the track data subset pairs in the preliminary track data set group is completed, and a plurality of judgment results are obtained; a determining step of determining whether the prepared track data set group meets a screening requirement based on all the judgment results, wherein the screening requirement is that the difference between the track data in the first track data set and the corresponding track data in the second track data set in all pairs of the track data subsets in the prepared track data set group is less than a preset difference; Repeating the first obtaining step, the first repeating step, and the determining step at least once until all the track data set groups are determined; The track data set groups that do not meet the screening requirements are deleted from all the track data set groups, so as to obtain a plurality of track data set groups that meet the screening requirements.

3. The method according to claim 2, characterized in that Determining whether the preliminary track data set meets the screening requirements based on all the judgment results includes: If all the judgment results are yes, it is determined that the preliminary track dataset group meets the screening requirement; if at least one judgment result is no, it is determined that the preliminary track dataset group does not meet the screening requirement.

4. The method according to claim 2, characterized in that The sum of the mean square errors of all track data subset pairs in each track data set group is calculated to obtain multiple total mean square error sums, including: A third acquisition step is to acquire a target track dataset group, wherein the target track dataset group is any one of the track dataset groups that meets the screening requirements; The first calculation step is to calculate the mean square error of each pair of track data subsets in the target track data set group to obtain multiple target mean square errors; The second calculation step is to calculate the sum of all the target mean square errors to obtain the target total mean square error sum; Repeat the third obtaining step, the first calculating step and the second calculating step at least once until the calculation of the sum of the mean square errors of all track data subset pairs in all track data set groups is completed to obtain a plurality of total mean square error sums.

5. The method according to claim 4, characterized in that The track data includes position data and speed data of the obstacle, and the mean square error of each track data subset pair in the target track data set group is calculated to obtain multiple target mean square errors, including: A fourth acquisition step is to acquire a target track data subset pair, wherein the target track data subset pair is any one of the track data subset pairs in the target track data set group; a third calculation step of calculating the mean square error of the position data of the target track data subset pair to obtain a plurality of target position mean square errors, and calculating the mean square error of the speed data of the target track data subset pair to obtain a plurality of target speed mean square errors; A fourth calculation step is to calculate the sum of all the target position mean square errors to obtain a first target mean square error, and calculate the sum of all the target speed mean square errors to obtain a second target mean square error; a fifth calculation step of calculating the sum of the first target mean square error and the second target mean square error to obtain the target mean square error; Repeat the fourth obtaining step, the third calculating step, the fourth calculating step, and the fifth calculating step at least once until the calculation of the mean square errors of all the track data subset pairs in the target track data set group is completed, and a plurality of target mean square errors are obtained.

6. The method according to claim 1, characterized in that Before using a track fusion algorithm to fuse the two tracks corresponding to the optimal track data subset to obtain multiple optimal tracks, the method further includes: Determine the minimum value of all the total mean square errors to obtain the optimal total mean square error; The optimal track data subset pair is determined according to the optimal total mean square error sum, where the optimal track data subset pair is the track data subset pair of the track data set group corresponding to the optimal total mean square error sum.

7. The method according to claim 1, characterized in that The number of the track data in the first track data set and the number of the track data in the second track data set are both equal to a predetermined number.

8. A track fusion device, characterized in that: The device comprises: an acquiring unit, configured to acquire track data of a plurality of tracks of a first sensor and a second sensor, respectively, to obtain a plurality of first track data sets and a plurality of second track data sets, wherein the first track data set includes a plurality of track data of an obstacle detected by the first sensor within a time period, and the second track data set includes a plurality of track data of an obstacle detected by the second sensor within the time period, wherein the time period includes a plurality of time nodes, and the track data correspond to the time nodes in a one-to-one manner; a calculation unit, calculating the sum of the mean square errors of all track data subset pairs in each track data set group to obtain a plurality of total mean square error sums, wherein the track data set group includes a plurality of track data subset pairs, one track data subset pair includes a first track data set and a second track data set, the first track data sets of any two track data subset pairs are different, and the second track data sets of any two track data subset pairs are different, the mean square error of the track data subset pair is the sum of the mean square errors of the track data in the track data subset pair, and the total mean square error sum corresponds one-to-one to the track data set group; A fusion unit adopts a track fusion algorithm to fuse the two tracks corresponding to the optimal track data subset pair to obtain multiple optimal tracks, wherein the optimal track data subset pair is the track data subset pair with the smallest total mean square error and the corresponding track data set group, and the optimal track data subset pair corresponds to the optimal track one by one.

9. A processor, characterized in that: The processor is configured to run a program, wherein the program executes the method according to any one of claims 1 to 7 when running.

10. A track fusion system, characterized in that: include: One or more processors, a memory, and one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, and the one or more programs include instructions for executing the method according to any one of claims 1 to 7.

Images