Multi-target track matching method based on time sequence backtracking
By employing a multi-target trajectory matching method based on time-series backtracking, utilizing a cache eviction algorithm and Euclidean distance for multi-stage matching, and combining confidence weights for data fusion, the problem of limited computing resources and low matching accuracy caused by large differences in sensor target update cycles is solved, thus achieving efficient and reliable multi-target matching.
Patent Information
- Application Number
- CN202511072474.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-01
- Publication Date
- 2025-11-14
AI Technical Summary
In existing avionics systems, the target update cycles of sensors vary greatly, computational resources are scarce, and the limited information provided, such as latitude and longitude coordinates and velocity, results in low matching accuracy.
A multi-target trajectory matching method based on time-series backtracking is adopted. The target data is updated through a cache elimination algorithm, multi-stage matching is performed using Euclidean distance and motion model, and data fusion is combined with confidence weight.
It achieves efficient and reliable multi-target matching with limited computing resources and storage space, and is suitable for scenarios where sensor targets have frequent cycles, thus improving matching accuracy.
Smart Images
Figure CN120951290A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of avionics technology, and more specifically, to a multi-target trajectory matching method based on time-series backtracking. Background Technology
[0002] Currently, multi-target matching technology in avionics systems faces challenges such as large differences in sensor target update cycles, frequent updates leading to computational resource constraints, and the limited information provided by sensors, including latitude and longitude coordinates and velocity. While the paper "A Probabilistic Data Association Method Based on Point Distance" employs a method combining fused point traces and probability ratios, it still fails to determine trajectory similarity and cannot achieve the desired time-series backtracking path fitting effect.
[0003] Specifically, the paper "A Probabilistic Data Association Method Based on Point Distance" uses point distance as the criterion for judging the correlation of point tracks, and then uses the proportion of related point tracks in a track to the total number of tracks as the evaluation standard for track correlation. Although this method calculates the track distance and probability ratio, it does not determine the target point track's direction of travel or the path similarity correlation, which affects the accuracy of target matching.
[0004] No effective solutions have yet been proposed to address the problems in the relevant technologies. Summary of the Invention
[0005] To address the shortcomings of existing technologies, this invention provides a multi-target trajectory matching method based on time-series backtracking, which solves the problem of fast and efficient multi-target matching under the condition that multiple sensor targets only have latitude and longitude coordinates and velocity information.
[0006] Therefore, the specific technical solution adopted by the present invention is as follows:
[0007] A multi-target trajectory matching method based on time-series backtracking, comprising the following steps:
[0008] S1. Update and store the acquired low-frequency periodic target data and high-frequency periodic target data based on the cache eviction algorithm to obtain the low-frequency periodic target array and the high-frequency periodic target array;
[0009] S2. Perform initial matching on the low-frequency periodic target array and the high-frequency periodic target array, and determine whether the initial matching is successful based on the initial matching conditions;
[0010] S3. If the initial matching fails, return to step S2; if the initial matching succeeds, backtrack the low-frequency periodic target array and the high-frequency periodic target array after the initial matching succeeds, and determine whether the backtracking matching is successful based on the backtracking matching conditions.
[0011] S4. If the backtracking match is unsuccessful, return to step S2; if the backtracking match is successful, perform data fusion on the high-frequency periodic target data and low-frequency periodic target data that were successfully backtracked based on the confidence weight to obtain the target point track result.
[0012] Furthermore, updating and storing the acquired low-frequency periodic target data and high-frequency periodic target data based on the cache eviction algorithm to obtain the low-frequency periodic target array and the high-frequency periodic target array includes the following steps:
[0013] S11. Based on the preset low-frequency periodic sensor and high-frequency periodic sensor, acquire low-frequency periodic target data and high-frequency periodic target data;
[0014] S12. Store target information based on a global static array, store low-frequency periodic target data and high-frequency periodic target data based on a circular queue to obtain an initial low-frequency periodic target array and an initial high-frequency periodic target array;
[0015] S13. Using a cache eviction algorithm, update and store the initial low-frequency period target array and the initial high-frequency period target array to obtain the low-frequency period target array and the medium-term high-frequency period target array, respectively.
[0016] S14. Periodically sample the mid-term high-frequency periodic target array to obtain the high-frequency periodic target array.
[0017] Furthermore, a circular queue includes forward pointers, backward pointers, and reserved pointers;
[0018] The forward pointer marks the beginning of the circular queue, and the backward pointer marks the end of the circular queue. When the forward and backward pointers coincide, the circular queue is empty. When the backward pointer minus the forward pointer equals the reserved pointer, the circular queue is full.
[0019] Furthermore, using a cache eviction algorithm, the initial low-frequency period target array and the initial high-frequency period target array are updated and stored to obtain the low-frequency period target array and the medium-term high-frequency period target array, respectively, including the following steps:
[0020] S131. Determine the maximum number of the initial low-frequency periodic target array and the initial high-frequency periodic target array;
[0021] S132. When the number of initial low-frequency periodic target arrays and initial high-frequency periodic target arrays reaches the maximum number, determine the target data that has not been updated for the longest time.
[0022] S133. Replace the oldest unupdated target data with the new target data, update and store the initial low-frequency period target array and the initial high-frequency period target array, and obtain the low-frequency period target array and the medium-term high-frequency period target array respectively.
[0023] Furthermore, the initial matching of the low-frequency periodic target array and the high-frequency periodic target array, and the determination of whether the initial matching is successful based on the initial matching conditions, include the following steps:
[0024] S21. Calculate the Euclidean distance between the latest track points in the low-frequency periodic target array and the high-frequency periodic target array;
[0025] S22. Based on the Euclidean distance between the track points at the latest time and combined with the initial matching conditions, determine whether the initial matching was successful.
[0026] Furthermore, based on the Euclidean distance between the waypoints at the latest moment, and in conjunction with the initial matching conditions, the determination of whether the initial match was successful includes:
[0027] The initial radius is determined by using the latest track point in the low-frequency periodic target array as the center. If the Euclidean distance between the latest track point in the high-frequency periodic target array and the latest track point in the low-frequency periodic target array is greater than or equal to the initial radius, the initial match is considered unsuccessful. If the Euclidean distance between the latest track point in the high-frequency periodic target array and the latest track point in the low-frequency periodic target array is less than the initial radius, the initial match is considered successful.
[0028] Furthermore, if the initial matching is successful, a backtracking match is performed on the low-frequency periodic target array and the high-frequency periodic target array after the initial matching, and the determination of whether the backtracking match is successful is based on the backtracking match conditions, including the following steps:
[0029] A motion model is constructed using historical points from a low-frequency periodic target array; and based on the motion model, the average distance error for the corresponding time period in the high-frequency periodic target array is calculated.
[0030] Based on the average distance error of the corresponding time period in the high-frequency periodic target array, and combined with the backtracking matching conditions, it is determined whether the backtracking matching is successful.
[0031] Furthermore, a motion model is constructed using historical points from the low-frequency periodic target array; and based on the motion model, the average distance error for the corresponding time period in the high-frequency periodic target array is calculated, including the following steps:
[0032] Calculate the average velocity vector from the latest point to the previous point in the low-frequency periodic target array;
[0033] A motion model is constructed based on the average velocity vector from the latest point to the previous point in the low-frequency periodic target array.
[0034] Using a motion model, the predicted position of each trajectory point in the corresponding time period in the high-frequency periodic target array is calculated;
[0035] The distance error is calculated based on the actual position of each trajectory point in the high-frequency periodic target array within the corresponding time period and the corresponding predicted position of each trajectory point within the corresponding time period.
[0036] Based on the distance error, the average distance error of the corresponding time period in the high-frequency periodic target array is calculated.
[0037] Furthermore, using a motion model, the predicted position of each trajectory point within the corresponding time period in the high-frequency periodic target array is calculated, including:
[0038] Calculate the time difference based on the starting point and the current time.
[0039] By utilizing the time difference and combining it with the average velocity vector from the latest point to the previous point, the position of motion can be obtained;
[0040] Based on the movement position and the position of the latest point, the predicted position at the current time point is obtained.
[0041] Furthermore, based on the average distance error of the corresponding time period in the high-frequency periodic target array, and combined with the backtracking matching conditions, the determination of whether the backtracking matching is successful includes:
[0042] The matching score is calculated based on the average distance error of the corresponding time period in the high-frequency periodic target array;
[0043] When the matching score is less than the preset backtracking matching score, the backtracking matching is considered unsuccessful.
[0044] When the matching score is greater than or equal to the preset backtracking matching score, the backtracking matching is considered successful.
[0045] Compared with existing technologies, this invention provides a multi-target trajectory matching method based on time-series backtracking, which has the following beneficial effects:
[0046] (1) Based on the mature Euclidean distance and nearest neighbor matching algorithm, this invention proposes a multi-target track matching method based on time backtracking. By using multi-stage matching and reasonable time sampling, the increased computational resources caused by the large difference between fast and slow cycles (i.e., high frequency and low frequency cycles) are eliminated, thus achieving efficient and reliable track matching.
[0047] (2) This invention solves the problem of fast and efficient multi-target matching under the condition that multiple sensor targets only have latitude and longitude coordinates and velocity information; it is suitable for scenarios where the sensor provides target period frequently and the amount of data is large; it is also suitable for application scenarios where CPU computing resources and storage space are limited, but matching calculation is required in real time. Attached Figure Description
[0048] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0049] Figure 1 This is one of the flowcharts of a multi-target trajectory matching method based on time-series backtracking according to an embodiment of the present invention;
[0050] Figure 2 This is a second flowchart of a multi-target trajectory matching method based on time-series backtracking according to an embodiment of the present invention;
[0051] Figure 3 This is a schematic diagram of the target array and the circular queue of track points in the multi-target track matching method based on time-series backtracking according to an embodiment of the present invention.
[0052] Figure 4 This is a schematic diagram of the initial matching in the multi-target trajectory matching method based on time-series backtracking according to an embodiment of the present invention;
[0053] Figure 5 This is a schematic diagram of backtracking matching in a multi-target trajectory matching method based on time-series backtracking according to an embodiment of the present invention. Detailed Implementation
[0054] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention. The components in the drawings are not drawn to scale, and similar component symbols are generally used to represent similar components.
[0055] According to an embodiment of the present invention, a multi-target trajectory matching method based on time-series backtracking is provided.
[0056] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figures 1-5 As shown, the multi-target trajectory matching method based on time-series backtracking according to an embodiment of the present invention includes the following steps:
[0057] S1. Based on the cache eviction algorithm, update and store the acquired low-frequency periodic target data and high-frequency periodic target data to obtain the low-frequency periodic target array and the high-frequency periodic target array.
[0058] It should be explained that the inputs of this invention are slow-cycle target data (i.e., low-frequency cycle target data) and fast-cycle target data (i.e., high-frequency cycle target data). Multiple target data are received through a fast-cycle timing task.
[0059] In this optional embodiment, updating and storing the acquired low-frequency periodic target data and high-frequency periodic target data based on the cache eviction algorithm to obtain the low-frequency periodic target array and the high-frequency periodic target array includes the following steps:
[0060] S11. Based on the preset low-frequency periodic sensor and high-frequency periodic sensor, acquire low-frequency periodic target data and high-frequency periodic target data;
[0061] S12. Store target information based on a global static array, store low-frequency periodic target data and high-frequency periodic target data based on a circular queue to obtain an initial low-frequency periodic target array and an initial high-frequency periodic target array;
[0062] S13. Using a cache eviction algorithm, update and store the initial low-frequency period target array and the initial high-frequency period target array to obtain the low-frequency period target array and the medium-term high-frequency period target array, respectively.
[0063] S14. Periodically sample the mid-term high-frequency periodic target array to obtain the high-frequency periodic target array.
[0064] It should be explained that the fast-cycle target point track data is updated too frequently, and the position changes little in each update cycle. The data of fast-cycle targets is sampled according to a cycle in which a significant change in position and distance can be measured, and the sampled point track data is stored in the final circular queue.
[0065] When the number of fast periodic targets reaches the maximum space of the static array, the LRU algorithm (Least Recently Unused Algorithm, i.e., cache eviction algorithm) is used to update the fast periodic targets. That is, after the fast periodic target data is received, the LRU algorithm is first used to find the index position of the global periodic target array, and then useful data is selected based on the second sampling.
[0066] In this optional embodiment, the circular queue includes a forward pointer, a backward pointer, and a reserved pointer;
[0067] The forward pointer marks the beginning of the circular queue, and the backward pointer marks the end of the circular queue. When the forward and backward pointers coincide, the circular queue is empty. When the backward pointer minus the forward pointer equals the reserved pointer, the circular queue is full.
[0068] It needs to be explained that, for example Figure 3As shown, this invention uses a global static array to store target information, and the point trajectory data of a single target is stored using a circular queue according to the time sequence.
[0069] The structure for storing target number information is a "target array". The target point track data is stored in a circular queue, with front (forward pointer), rear (backward pointer), and reserve (reserved pointer). front is the beginning of the circular queue, and rear is the end of the circular queue. When front and rear coincide, the circular queue is empty. When rear minus front equals reserve, the circular queue is full.
[0070] In this optional embodiment, the initial low-frequency period target array and the initial high-frequency period target array are updated and stored using a cache eviction algorithm to obtain the low-frequency period target array and the medium-frequency period target array, respectively, including the following steps:
[0071] S131. Determine the maximum number of the initial low-frequency periodic target array and the initial high-frequency periodic target array;
[0072] S132. When the number of initial low-frequency periodic target arrays and initial high-frequency periodic target arrays reaches the maximum number, determine the target data that has not been updated for the longest time.
[0073] S133. Replace the oldest unupdated target data with the new target data, update and store the initial low-frequency period target array and the initial high-frequency period target array, and obtain the low-frequency period target array and the medium-term high-frequency period target array respectively.
[0074] It should be explained that the algorithm for updating the "target array" is based on the longest unused time. For example... Figure 3 As shown, the algorithm for the longest unused period is applied in the specific method of this invention. The maximum number N in the "target array" is fixed. When the number of targets in the "target array" reaches N, a new target needs to find the target that has not been updated for the longest time in the static array, such as... Figure 3 In the process, the rear pointer node of the circular queue with target number 4 has a time of 5500ms, and it is determined to be the target that has not been updated for the longest time. The new target N+1 replaces 4, and the circular queue corresponding to the original target 4 is reinitialized and allocated to the point track data of the new target N+1.
[0075] S2. Perform initial matching on the low-frequency periodic target array and the high-frequency periodic target array, and determine whether the initial matching is successful based on the initial matching conditions.
[0076] It should be explained that the point track matching task is scheduled periodically, traversing all slow-cycle targets and performing initial matching.
[0077] In this optional embodiment, the initial matching of the low-frequency periodic target array and the high-frequency periodic target array, and the determination of whether the initial matching is successful based on the initial matching conditions, includes the following steps:
[0078] S21. Calculate the Euclidean distance between the latest track points in the low-frequency periodic target array and the high-frequency periodic target array;
[0079] S22. Based on the Euclidean distance between the track points at the latest time and combined with the initial matching conditions, determine whether the initial matching was successful.
[0080] It should be explained that the Euclidean distance between the latest track points of the slow-period and fast-period targets is calculated using the following formula:
[0081]
[0082] In the formula, {x2, y2} and {x1, y1} represent two coordinate points, corresponding to latitude, longitude, and altitude in three-dimensional space. The distance between the two points is:
[0083]
[0084] In the formula, {x2, y2, z2} and {x1, y1, z1} are the latitude, longitude, and altitude data of the point tracks provided by the two sensors at the same time. The distance error is within a certain range of sensor accuracy error, thus satisfying the initial matching condition. The distance error calculation formula is as follows:
[0085] Distance error = Sensor accuracy error * Empirical value multiple
[0086] The sensor measurement accuracy error is provided by the manufacturer, and the empirical value multiplier can be used as a parameter for adjustment in actual engineering applications.
[0087] In this optional embodiment, determining whether the initial match was successful, based on the Euclidean distance between the waypoints at the latest time and in conjunction with the initial matching conditions, includes:
[0088] The initial radius is determined by using the latest track point in the low-frequency periodic target array as the center. If the Euclidean distance between the latest track point in the high-frequency periodic target array and the latest track point in the low-frequency periodic target array is greater than or equal to the initial radius, the initial match is considered unsuccessful. If the Euclidean distance between the latest track point in the high-frequency periodic target array and the latest track point in the low-frequency periodic target array is less than the initial radius, the initial match is considered successful.
[0089] It needs to be explained that, for example Figure 4As shown, fast-cycle targets within the radius of the dashed line are considered to be targets with successful initial matching, while fast-cycle targets outside the radius of the dashed line are considered to be targets with unsuccessful initial matching.
[0090] S3. If the initial matching fails, return to step S2; if the initial matching succeeds, backtrack the low-frequency periodic target array and the high-frequency periodic target array after the initial matching succeeds, and determine whether the backtracking matching is successful based on the backtracking matching conditions.
[0091] It should be noted that historical track matching can only be performed if the initial matching conditions are met.
[0092] In this optional embodiment, if the initial matching is successful, a backtracking match is performed on the low-frequency periodic target array and the high-frequency periodic target array after the initial matching is successful. The determination of whether the backtracking match is successful based on the backtracking match conditions includes the following steps:
[0093] A motion model is constructed using historical points from a low-frequency periodic target array; and based on the motion model, the average distance error for the corresponding time period in the high-frequency periodic target array is calculated.
[0094] Based on the average distance error of the corresponding time period in the high-frequency periodic target array, and combined with the backtracking matching conditions, it is determined whether the backtracking matching is successful.
[0095] It needs to be explained that, for example Figure 5 As shown in the figure (where dn represents the distance difference between the high-frequency target point Bn and the low-frequency target point An), a motion model is constructed using two historical points of the slow cycle, the current latest point (i.e., the latest point) and the previous point (i.e., the previous point). Then, all trajectory points within the corresponding time period are found on the fast cycle sensor, and the constructed motion model is used to verify the degree of conformity of the fast cycle trajectory.
[0096] In this optional embodiment, a motion model is constructed using historical points from a low-frequency periodic target array; and based on the motion model, the average distance error for the corresponding time period in the high-frequency periodic target array is calculated, including the following steps:
[0097] Calculate the average velocity vector from the latest point to the previous point in the low-frequency periodic target array;
[0098] A motion model is constructed based on the average velocity vector from the latest point to the previous point in the low-frequency periodic target array.
[0099] Using a motion model, the predicted position of each trajectory point in the corresponding time period in the high-frequency periodic target array is calculated;
[0100] The distance error is calculated based on the actual position of each trajectory point in the high-frequency periodic target array within the corresponding time period and the corresponding predicted position of each trajectory point within the corresponding time period.
[0101] Based on the distance error, the average distance error of the corresponding time period in the high-frequency periodic target array is calculated.
[0102] It should be explained that for a slow-period sensor target, there are two points: the latest point An (time Tn) and the previous point An-1 (time Tn-1). Calculate the average velocity vector (Vx, Vy) from An to An-1.
[0103] For each point (time Ti, position Bn) in the fast-cycle trajectory, the predicted position within the time period is calculated using the slow-cycle motion model.
[0104] The distance error between each fast-cycle trajectory point and its corresponding predicted location is calculated using the following formula:
[0105] dx = xb - predx;
[0106] dy = yb - predy;
[0107]
[0108] In the formula, dx represents the x-coordinate distance error; dy represents the y-coordinate distance error; xb represents the actual x-coordinate of point Bn; yb represents the actual y-coordinate of point Bn; predx represents the predicted x-coordinate of point Bn; and predy represents the predicted y-coordinate of point Bn.
[0109] The average error over the corresponding time period is calculated.
[0110] Average error: avg_shape_error = shape_error / count
[0111] In the formula, avg_shape_error represents the average error within the corresponding time period; shape_error represents the sum of errors within the corresponding time period; and count represents the number of targets in the high-frequency periodic target array.
[0112] In this optional embodiment, calculating the predicted position of each trajectory point within the corresponding time period in the high-frequency periodic target array using a motion model includes:
[0113] Calculate the time difference based on the starting point and the current time.
[0114] By utilizing the time difference and combining it with the average velocity vector from the latest point to the previous point, the position of motion can be obtained;
[0115] Based on the movement position and the position of the latest point, the predicted position at the current time point is obtained.
[0116] It should be explained that for each point (time Ti, position Bn) in the fast-cycle trajectory, the predicted position within the time interval is calculated using the slow-cycle motion model. The formula is as follows:
[0117] Predicted position pred = An + (Ti – Tn-1) * (Vx, Vy);
[0118] In the formula, Ti represents the time of the current point; An represents the position of the latest point; Tn-1 represents the time of the previous point; (Vx,Vy) represents the average velocity vector from An to An-1.
[0119] In this optional embodiment, determining whether the backtracking match is successful, based on the average distance error of the corresponding time period in the high-frequency periodic target array and combined with the backtracking matching conditions, includes:
[0120] The matching score is calculated based on the average distance error of the corresponding time period in the high-frequency periodic target array;
[0121] When the matching score is less than the preset backtracking matching score, the backtracking matching is considered unsuccessful.
[0122] When the matching score is greater than or equal to the preset backtracking matching score, the backtracking matching is considered successful.
[0123] It should be explained that the formula for calculating the matching score is as follows:
[0124] Score: shape_score=exp(-avg_shape_error / (10+speed_A*0.5));
[0125] In the formula, shape_score represents the matching score, which is normalized by an exponential decay function to constrain the result to the range [0,1]; exp() represents an exponential function with the natural number e as the base; avg_shape_error represents the average error within the corresponding time period; and speed_A represents the speed of the low-frequency target A.
[0126] When the matching score is greater than or equal to 0.5 (i.e., the preset backtracking matching score), it is judged as related; when the matching score is less than 0.5, it is judged as not related.
[0127] S4. If the backtracking match is unsuccessful, return to step S2; if the backtracking match is successful, perform data fusion on the high-frequency periodic target data and low-frequency periodic target data that were successfully backtracked based on the confidence weight to obtain the target point track result.
[0128] It should be explained that for two successfully matched targets, data fusion is performed. The track points of the fast-cycle target and the slow-cycle target are calculated according to the sensor confidence weight ratio, and the calculation formula is as follows:
[0129]
[0130] In the formula, j represents the target number index to be fused; M represents the maximum number of sensors; weight j represents the weight of the j-th target; data j represents the point trajectory data of the j-th target. The confidence weight ratio is determined based on the sensor manufacturer's error accuracy and combined with empirical values from engineering applications, and satisfies the following conditions: The requirement for weight normalization.
[0131] like Figure 2 The diagram shown (Y represents success, N represents failure) illustrates the overall flowchart of this invention. This invention addresses the application scenario of matching multiple targets with varying update cycles (slow and fast). It achieves rapid path matching for multiple targets across two levels: the first level quickly filters targets within a possible distance range; the second level, for the initially filtered targets, calculates the trajectory of the target back to the path n cycles prior to the slow update cycle through time backtracking, determining whether it is the same target's trajectory, where n depends on the ratio of high-frequency time intervals to low-frequency time intervals.
[0132] This invention prioritizes sampling targets with faster update cycles before adding them to a circular queue, significantly improving data processing performance and storage space utilization. The matching process employs a two-stage matching method: first, initial matching is performed on the positions of multiple targets at their last moment. Once the initial matching conditions are met, time backtracking is performed, backtracking four to five cycles according to the slower cycle. The point trajectory paths formed by the fast and slow targets corresponding to these four to five cycles are calculated, and a match is determined based on the similarity of the generated paths.
[0133] This invention has been verified through real-world application scenarios, including applications with radar for rapid target updates and data link for slow target updates, achieving fast and efficient matching verification.
[0134] In summary, leveraging the technical solutions described above, this invention proposes a multi-target trajectory matching method based on temporal backtracking, building upon mature Euclidean distance and nearest neighbor matching algorithms. Through multi-stage matching and reasonable temporal sampling, it eliminates the increased computational resources caused by large differences in fast and slow cycles, achieving efficient and reliable trajectory matching. This invention solves the problem of fast and efficient multi-target matching under conditions where multiple sensors only provide latitude, longitude, and velocity information; it is suitable for scenarios where sensors provide target information frequently with large amounts of data; and it is also suitable for applications with limited CPU computing resources and storage space, but where real-time matching calculations are required.
[0135] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A multi-target trajectory matching method based on temporal backtracking, characterized in that, The method includes the following steps: S1. Update and store the acquired low-frequency periodic target data and high-frequency periodic target data based on the cache eviction algorithm to obtain the low-frequency periodic target array and the high-frequency periodic target array; S2. Perform initial matching on the low-frequency periodic target array and the high-frequency periodic target array, and determine whether the initial matching is successful based on the initial matching conditions; S3. If the initial matching fails, return to step S2; if the initial matching succeeds, backtrack the low-frequency periodic target array and the high-frequency periodic target array after the initial matching succeeds, and determine whether the backtracking matching is successful based on the backtracking matching conditions. S4. If the backtracking match is unsuccessful, return to step S2; if the backtracking match is successful, perform data fusion on the high-frequency periodic target data and low-frequency periodic target data that were successfully backtracked based on the confidence weight to obtain the target point track result.
2. The multi-target trajectory matching method based on time-series backtracking according to claim 1, characterized in that, The process of updating and storing the acquired low-frequency periodic target data and high-frequency periodic target data based on the cache eviction algorithm to obtain the low-frequency periodic target array and the high-frequency periodic target array includes the following steps: S11. Based on the preset low-frequency periodic sensor and high-frequency periodic sensor, acquire low-frequency periodic target data and high-frequency periodic target data; S12. Store target information based on a global static array, store low-frequency periodic target data and high-frequency periodic target data based on a circular queue to obtain an initial low-frequency periodic target array and an initial high-frequency periodic target array; S13. Using a cache eviction algorithm, update and store the initial low-frequency period target array and the initial high-frequency period target array to obtain the low-frequency period target array and the medium-term high-frequency period target array, respectively. S14. Periodically sample the mid-term high-frequency periodic target array to obtain the high-frequency periodic target array.
3. The multi-target trajectory matching method based on time-series backtracking according to claim 2, characterized in that, The circular queue includes a forward pointer, a backward pointer, and a reserved pointer; The forward pointer marks the beginning of the circular queue, and the backward pointer marks the end of the circular queue. When the forward and backward pointers coincide, the circular queue is empty. When the backward pointer minus the forward pointer equals the reserved pointer, the circular queue is full.
4. The multi-target trajectory matching method based on time-series backtracking according to claim 2, characterized in that, The step of updating and storing the initial low-frequency period target array and the initial high-frequency period target array using a cache eviction algorithm to obtain the low-frequency period target array and the medium-term high-frequency period target array respectively includes the following steps: S131. Determine the maximum number of the initial low-frequency periodic target array and the initial high-frequency periodic target array; S132. When the number of initial low-frequency periodic target arrays and initial high-frequency periodic target arrays reaches the maximum number, determine the target data that has not been updated for the longest time. S133. Replace the oldest unupdated target data with the new target data, update and store the initial low-frequency period target array and the initial high-frequency period target array, and obtain the low-frequency period target array and the medium-term high-frequency period target array respectively.
5. The multi-target trajectory matching method based on time-series backtracking according to claim 1, characterized in that, The initial matching of the low-frequency periodic target array and the high-frequency periodic target array, and the determination of whether the initial matching is successful based on the initial matching conditions, includes the following steps: S21. Calculate the Euclidean distance between the latest track points in the low-frequency periodic target array and the high-frequency periodic target array; S22. Based on the Euclidean distance between the track points at the latest time and combined with the initial matching conditions, determine whether the initial matching was successful.
6. A multi-target trajectory matching method based on time-series backtracking according to claim 5, characterized in that, The determination of whether the initial match is successful, based on the Euclidean distance between the waypoints at the latest time and combined with the initial matching conditions, includes: The initial radius is determined by using the latest track point in the low-frequency periodic target array as the center. If the Euclidean distance between the latest track point in the high-frequency periodic target array and the latest track point in the low-frequency periodic target array is greater than or equal to the initial radius, the initial match is considered unsuccessful. If the Euclidean distance between the latest track point in the high-frequency periodic target array and the latest track point in the low-frequency periodic target array is less than the initial radius, the initial match is considered successful.
7. The multi-target trajectory matching method based on time-series backtracking according to claim 1, characterized in that, If the initial matching is successful, then backtracking matching is performed on the low-frequency periodic target array and the high-frequency periodic target array after the initial matching is successful, and the determination of whether the backtracking matching is successful is based on the backtracking matching conditions, including the following steps: A motion model is constructed using historical points from a low-frequency periodic target array; and based on the motion model, the average distance error for the corresponding time period in the high-frequency periodic target array is calculated. Based on the average distance error of the corresponding time period in the high-frequency periodic target array, and combined with the backtracking matching conditions, it is determined whether the backtracking matching is successful.
8. A multi-target trajectory matching method based on time-series backtracking according to claim 7, characterized in that, The process of constructing a motion model using historical points from a low-frequency periodic target array and calculating the average distance error for the corresponding time period in the high-frequency periodic target array based on the motion model includes the following steps: Calculate the average velocity vector from the latest point to the previous point in the low-frequency periodic target array; A motion model is constructed based on the average velocity vector from the latest point to the previous point in the low-frequency periodic target array. Using a motion model, the predicted position of each trajectory point in the corresponding time period in the high-frequency periodic target array is calculated; The distance error is calculated based on the actual position of each trajectory point in the high-frequency periodic target array within the corresponding time period and the corresponding predicted position of each trajectory point within the corresponding time period. Based on the distance error, the average distance error of the corresponding time period in the high-frequency periodic target array is calculated.
9. A multi-target trajectory matching method based on time-series backtracking according to claim 8, characterized in that, The step of using a motion model to calculate the predicted position of each trajectory point in the corresponding time period in the high-frequency periodic target array includes: Calculate the time difference based on the starting point and the current time. By utilizing the time difference and combining it with the average velocity vector from the latest point to the previous point, the position of motion can be obtained; Based on the movement position and the position of the latest point, the predicted position at the current time point is obtained.
10. A multi-target trajectory matching method based on time-series backtracking according to claim 7, characterized in that, The determination of whether the backtracking match is successful, based on the average distance error of the corresponding time period in the high-frequency periodic target array and combined with the backtracking matching conditions, includes: The matching score is calculated based on the average distance error of the corresponding time period in the high-frequency periodic target array; When the matching score is less than the preset backtracking matching score, the backtracking matching is considered unsuccessful. When the matching score is greater than or equal to the preset backtracking matching score, the backtracking matching is considered successful.