Multi-hypothesis tracking processing method based on point track clustering division and program product

Abstract

The invention relates to the technical field of radar data processing, in particular to a multi-hypothesis tracking processing method based on point track clustering division and a program product. Receiving radar detection plots from radar front-end equipment, taking a batch of plots as a processing beat, and executing the following steps on each processing beat; establishing a one-dimensional point track association table for each trace point, and combining the association tables of all trace points to obtain a point track correlation matrix; according to the point track correlation matrix, all tracks are clustered to obtain preliminary clustering results, the preliminary clustering results of the latest q processing beats are read and combined to obtain final clustering results, and for each cluster of the final clustering results, the tracks are updated by using a multi-hypothesis tracking algorithm process; and finally, outputting all the updated tracks. According to the method, the problem of calculation force saturation of a multi-hypothesis tracking algorithm caused by multiple false detection trace points in a complex environment is effectively prevented, and engineering application of multi-hypothesis tracking is facilitated.

Description

Technical Field

[0001] The present invention relates to the technical field of radar data processing, and in particular to a multi-hypothesis tracking processing method based on point track clustering and a program product. Background Art

[0002] The Multiple Hypothesis Tracking (MHT) algorithm was proposed by DB Reid, a scientist at Lockheed Martin in 1976. It is an algorithm for tracking multiple targets in a high clutter density background. When the multi-target tracking algorithm receives each frame of point track measurement data, it needs to perform data association based on the measurement at the current moment and the point track at the previous moment or the existing track, based on the measurement uncertainty. The core idea of ​​the multi-hypothesis tracking algorithm is delayed decision logic (non-delayed output), that is, when a conflict occurs in data association, track hypothesis branches are performed to maintain the possibility of various associations, and multiple hypotheses are formed on this basis. As the target detection data accumulates, the multi-hypothesis tracking algorithm will continuously pass on the multiple hypothesis branches of the target generated, avoiding the decision point association pairing at the current single moment. Instead, it will jointly resolve this uncertainty through subsequent measurement data, use multiple frames of data to make the best decision on the hypothesis branches, and more accurately select the most likely association sequence of the target, avoiding the randomness brought by single-frame decision-making, so as to improve the continuity of target tracking. The schematic diagram of track analysis is shown below. Figure 1 shown.

[0003] However, since the multi-hypothesis tracking algorithm needs to maintain all the possibilities of correlations at multiple consecutive detection moments, the track branch hypothesis will encounter the problem of correlation combination explosion. Especially in strong clutter and strong interference environments, as the number of false points increases, the computational complexity of the multi-hypothesis tracking algorithm increases exponentially. For a long time, the computational complexity has become a bottleneck in the engineering application of the multi-hypothesis tracking algorithm. In order to solve the problem of computational saturation of the multi-hypothesis tracking algorithm, scholars have proposed many optimization methods. For example, the multiple wave gate method, the N-frame pruning algorithm, the likelihood probability evaluation method for a single track, the track hypothesis branch restriction method, etc. These methods have achieved some results in improving computational efficiency, but most of the above methods are at the expense of sacrificing some performance of the algorithm. From the perspective of public technology, there are few methods for optimizing the design of the software architecture for implementing the algorithm. Summary of the Invention

[0004] This paper addresses the computational saturation problem of multi-hypothesis tracking algorithms in complex environments by providing a multi-core parallel processing architecture for multi-hypothesis tracking based on point track clustering. Through the parallel multi-core design of the software architecture, the processing efficiency of the multi-hypothesis tracking algorithm is linearly improved, solving the engineering application challenges of the algorithm.

[0005] In the actual detection environment, the spatial distribution of the point track data detected by the radar is not continuous in the entire airspace, but presents clustering characteristics, and there are "isolation zones" between clusters, such as Figure 2 When processing radar detection data, the existence of these isolation zones can divide the point tracks into different, independent clusters for processing. Through the analysis of measured data, the number of independent point track clusters in each processing cycle ranges from 20 to 100, as shown in Figure 2. Figure 3 As shown. Based on the above data characteristics, in each processing, if the point track is divided into multiple independent clusters, and the idea of ​​core processing is adopted for each cluster, the multi-core processing resources of the hardware are fully utilized, and the computational efficiency of the algorithm can be greatly improved. Based on this, the present invention designs a multi-hypothesis tracking multi-core parallel processing architecture based on point track clustering. Through the parallel multi-core design of the software architecture, the processing efficiency of the multi-hypothesis tracking algorithm is linearly improved. The problem of computing power saturation of the multi-hypothesis tracking algorithm in actual application is solved, and the engineering application of the algorithm is promoted.

[0006] The present invention provides a multi-hypothesis tracking processing method based on point track clustering. Radar detection point tracks are received from a radar front-end device. Whenever the cumulative time of reception reaches a time threshold or the number of received point tracks reaches a quantity threshold, this batch of point tracks is regarded as a processing beat. The following steps are performed for each processing beat:

[0007] S1: For each point track in the current processing beat, establish a one-dimensional point track association table containing K elements; the one-dimensional point track association table is used to indicate which tracks the corresponding point track is associated with; where K is the number of tracks;

[0008] S2: Merge the one-dimensional point track association table of each point track in the current processing beat to obtain an N×K point track correlation matrix; where N is the number of points tracks in the current processing beat;

[0009] S3: Cluster all tracks according to the point track correlation matrix obtained by the current processing beat to obtain preliminary clustering results. The clustering rules are:

[0010] Tracks with common points are related to each other and divided into a cluster;

[0011] Read the preliminary clustering results of the most recent q processing beats and merge them with the preliminary clustering results obtained for the current processing beat to obtain the final clustering result of the current processing beat. The merging rules are:

[0012] Clusters with common tracks are merged into one cluster, where q is the window number threshold;

[0013] S4: For each cluster of the final clustering result of the current processing beat, all tracks are updated using the multi-hypothesis tracking algorithm process; finally, all updated tracks are output.

[0014] Preferably, in step S1, the process of establishing a one-dimensional point track association table for each point track is processed in parallel.

[0015] Preferably, in step S4, the process of updating the track of each cluster is processed in parallel.

[0016] Preferably, the value of q is 3.

[0017] The present invention also provides a computer program product, comprising a computer program, which implements the above method when executed by a processor.

[0018] Beneficial effect: The present invention designs a multi-hypothesis tracking multi-core parallel processing architecture based on point track clustering, which linearly improves the processing efficiency of the multi-hypothesis tracking algorithm, effectively prevents the computing power saturation problem of the multi-hypothesis tracking algorithm caused by a large number of false detection points in complex environments, and is conducive to the engineering application of multi-hypothesis tracking. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 Schematic diagram of the track hypothesis branches of the multi-hypothesis tracking algorithm;

[0020] Figure 2 This is a schematic diagram of the “cluster” distribution of midpoint tracks in the detected airspace;

[0021] Figure 3 This is a graph of point track clustering based on the measured data;

[0022] Figure 4 This is a flowchart of multi-core parallel processing based on point track clustering; DETAILED DESCRIPTION

[0023] In order to make the purposes, features, and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0024] like Figure 4As shown in the figure, a multi-hypothesis tracking processing method based on point track clustering receives radar detection points from the radar front-end device. Whenever the cumulative time of reception reaches a time threshold or the number of received points reaches a quantity threshold, this batch of points is regarded as a processing beat. The following steps are performed for each processing beat:

[0025] S1: For each point track in the current processing beat, establish a one-dimensional point track association table containing K elements; the one-dimensional point track association table is used to indicate which tracks the corresponding point track is associated with; wherein K is the number of tracks; wherein, the process of establishing the one-dimensional point track association table for each point track is processed in parallel.

[0026] In this embodiment, the number of points in the current processing cycle is 10, and the value of K is 15; 10 cores are used for parallel processing, that is, the process of establishing a one-dimensional point track association table for each point track is executed in parallel in one core;

[0027] Among them, the one-dimensional point track association table established by the first point track, point track 1, is shown in Table 1:

[0028] Table 1 One-dimensional point track correlation table

[0029]

[0030] As shown in Table 1, 0 and 1 indicate whether or not they are associated. As can be seen from the table, point track 1 is associated with track 1, track 2, and track 15.

[0031] S2: Merge the one-dimensional point track association table of each point track in the current processing beat to obtain an N×K point track correlation matrix; where N is the number of points tracks in the current processing beat;

[0032] In this embodiment, the value of N is 10, and the point track correlation matrix is ​​shown in Table 2:

[0033] Table 2 N×K point track correlation matrix

[0034]

[0035] S3: Cluster all tracks according to the point track correlation matrix obtained by the current processing beat to obtain preliminary clustering results. The clustering rules are:

[0036] Tracks with common points are related to each other and divided into a cluster;

[0037] For example, if the point set of track 1 is {point 1, point 2}; the point set of track 2 is {point 2, point 3}; the point set of track 3 is {point 3, point 4}, then track 1, track 2, and track 3 are related to each other and divided into one cluster;

[0038] Read the preliminary clustering results of the most recent q processing beats and merge them with the preliminary clustering results obtained for the current processing beat to obtain the final clustering result of the current processing beat. The merging rules are:

[0039] Clusters with common tracks are merged into one cluster, where q is the window number threshold;

[0040] For example, there are the following four clusters, and their track sets are:

[0041] The track set of cluster 1: {track 1, track 2};

[0042] The track set of cluster 2: {track 2, track 3};

[0043] The track set of cluster 3: {track 3, track 4};

[0044] The track set of cluster 4: {track 5, track 6};

[0045] Cluster 1, cluster 2, and cluster 3 are merged to form a cluster {track 1, track 2, track 3, track 4}; cluster 4 continues to be an independent cluster.

[0046] S4: For each cluster of the final clustering result of the current processing beat, all tracks are updated using the multi-hypothesis tracking algorithm process; finally, all updated tracks are output. The process of updating the tracks of each cluster is processed in parallel.

[0047] In this embodiment, the final clustering result has 6 clusters in total, and 6 cores are used for parallel processing, that is, the track update process of each cluster is processed in parallel in an independent core.

[0048] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present invention. These improvements and modifications should also be regarded as within the scope of protection of the present invention.

Claims

1. A multi-hypothesis tracking processing method based on point track clustering, characterized in that: Receive radar detection traces from the radar front-end device. Whenever the cumulative time of reception reaches a time threshold or the number of received traces reaches a quantity threshold, treat this batch of traces as a processing beat and perform the following steps for each processing beat. S1: For each point track in the current processing beat, establish a one-dimensional point track association table containing K elements; the one-dimensional point track association table is used to indicate which tracks the corresponding point track is associated with; where K is the number of tracks; S2: Merge the one-dimensional point track association table of each point track in the current processing beat to obtain an N×K point track correlation matrix; where N is the number of points tracks in the current processing beat; S3: Cluster all tracks according to the point track correlation matrix obtained by the current processing beat to obtain preliminary clustering results. The clustering rules are: Tracks with common points are related to each other and divided into a cluster; Read the preliminary clustering results of the most recent q processing beats and merge them with the preliminary clustering results obtained for the current processing beat to obtain the final clustering result of the current processing beat. The merging rules are: Clusters with common tracks are merged into one cluster, where q is the window number threshold; S4: For each cluster of the final clustering result of the current processing beat, all tracks are updated using the multi-hypothesis tracking algorithm process; finally, all updated tracks are output.

2. The multi-hypothesis tracking processing method based on point track clustering according to claim 1, characterized in that: In step S1, the process of establishing a one-dimensional point track association table for each point track is processed in parallel.

3. The multi-hypothesis tracking processing method based on point track clustering according to claim 1, characterized in that: In step S4, the process of updating the track of each cluster is processed in parallel.

4. The multi-hypothesis tracking processing method based on point track clustering according to claim 1, characterized in that: The value of q is 3.

5. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images