A track initiation method based on batch processing

Abstract

The present invention discloses a track initiation method based on batch processing, comprising the following steps: 1: agglomerating measurement points within a scanning cycle; 2: selecting measurement points for calculating characteristic parameters in combination with target motion characteristics; 3: classifying the measurement points according to the target motion characteristics, and creating a characteristic parameter linked list for each of them; 4: transforming the characteristic parameter information of the measurement points filtered in step 2, projecting the information onto a velocity-coordinate parameter space of a reference frame, and serving as point track data in the parameter space; 5: repeating steps 2 to 4, traversing the scanning cycle, and clustering the point track data in the parameter space using a DBSCAN algorithm; 6: obtaining the centroid of the clusters formed by clustering in the parameter space, and serving as the characteristic information of the target corresponding to the cluster; and 7: inferring the coordinate information of the target in each scanning cycle based on the characteristic information of the target, thereby completing track initiation.

Description

Technical Field

[0001] The present invention relates to the field of radar data processing, and in particular to a track initiation method based on batch processing. Background Art

[0002] Track initiation is the primary issue in track management and the foundation of target tracking. If track initiation is incorrect, target tracking and subsequent processing will be impossible. During the target track initiation phase, the target is far from the radar, resulting in poor measurement accuracy and low detection resolution. This is especially true in complex background environments, where clutter can interfere with the target's echo signal. Track initiation is a challenging issue to address.

[0003] Currently, track initiation algorithms are divided into two main categories: sequential processing techniques and batch processing techniques. Sequential processing techniques acquire target tracks by sequentially processing multiple frames of data. These techniques primarily include intuitive methods, logical methods, and related improved algorithms. Batch processing techniques, which primarily involve the Hough transform and its derived improved algorithms, jointly process multiple frames of data.

[0004] Sequential processing is relatively suitable for track initiation in environments with low clutter density. When clutter density is high, in addition to the actual target track, sequential processing will also generate a large number of false tracks, interfering with subsequent target tracking. The widely used batch processing technology based on Hough transform can only determine the direction of the line during detection. The process of inferring the original track point data is computationally intensive and complex.

[0005] Therefore, in order to achieve the target track initiation, it is necessary to meet the optimal starting performance and consider the feasibility of engineering implementation, which is the current technical difficulty in this field. Summary of the Invention

[0006] In order to overcome the deficiencies in the prior art, the present invention provides a track initiation method based on batch processing.

[0007] In order to achieve the above-mentioned purpose of the invention, the technical solutions adopted to solve the technical problems are as follows:

[0008] A batch processing-based track initiation method comprises the following steps:

[0009] Step 1: Get the measurement set D1, D2, ..., D of N scan cycles N , for n measurement points D in each scanning cycle i ={(x1,y1),(x2,y2),…,(x n ,y n )} for cohesion;

[0010] Step 2: For the measurement set D of the i-th scanning cycle i And the measurement set D of the i+1th scanning cycle i+1 , combined with the maximum speed v of the target track max , filter out the measurement points that can be selected to calculate the characteristic parameters;

[0011] Step 3: Classify the measurement points according to the target motion characteristics and create a feature parameter list for each;

[0012] Step 4: For each feature parameter list in step 3, transform the feature parameter information of the measurement points filtered in step 2 and project it into the velocity-coordinate parameter space of a certain reference frame as the point trace data of the parameter space;

[0013] Step 5: Repeat steps 2 to 4, traverse the scanning cycle, and use the DBSCAN algorithm to complete the point data clustering in the parameter space;

[0014] Step 6: Obtain the centroid of the cluster formed after the DBSCAN algorithm clustering is completed in the parameter space as the feature information of the target corresponding to the cluster;

[0015] Step 7: Based on the characteristic information of the target in step 6, the coordinate information of the target in each scanning cycle is reversed to complete the track start.

[0016] Furthermore, in step 1, the steps for condensing the n measurement points in each scanning cycle are as follows:

[0017] Step 11: Traverse the traces within the scanning period to determine whether any traces meet the condensation condition;

[0018] Step 12: In the set of points that meet the aggregation conditions, perform equal-weight fusion aggregation on the measurement values ​​of the points;

[0019] Step 13: Add the condensed measurement values ​​to the measurement set D of the scan cycle i ;

[0020] Step 14: Delete measurement collection D i The measured values ​​of the condensed traces in .

[0021] Furthermore, in step 2, the rules for selecting measurement points that can be used to calculate characteristic parameters are as follows:

[0022] Considering that the target track has the maximum moving speed v max , assuming (x i ,y i ) is the point of the i-th frame, (x j ,y j) is the point of the jth frame, and the scanning period is T. Therefore, the premise for calculating the characteristic parameters should meet the following conditions:

[0023]

[0024] Among them, (x i ,y i ) is the measured value of the point trace in the x-axis and y-axis directions of the i-th frame, (x j ,y j ) is the measured value of the j-th frame trace in the x-axis and y-axis directions, v max is the maximum speed of the target track, T is the scanning period, i and j are the measurement points (x i ,y i ) and (x j ,y j ) corresponds to the frame count.

[0025] Furthermore, in step 3, the measurement points are classified according to the target motion characteristics, and the method of creating a feature parameter list is as follows:

[0026] Taking two-dimensional space as an example, let (x i ,y i ) is the coordinate of the point in the i-th frame, (x j ,y j ) is the coordinate of the point in the jth frame, where j>i, (v x ,v y ) is the velocity calculated from the coordinates of the two points, v Ref If the reference speed is set, the calculated speed will produce the following 5 results:

[0027]

[0028] Create 5 feature parameter lists corresponding to the above 5 results;

[0029] Calculate the speed of each pair of points that meet the conditions in step 2, and store the characteristic parameters in different linked lists according to the different speeds.

[0030] Furthermore, in step 4, the steps of projecting to the velocity-coordinate parameter space are as follows:

[0031] Step 41: Select the parameter projection frame as the Ref frame;

[0032] Step 42: Get the measurement points (x i ,y i ) and (x i+1 ,y i+1 )Calculated speed:

[0033]

[0034] Step 43: Calculate the measurement point of the i-th frame with the velocity (v x ,v y ) is projected onto the coordinates (x Ref,i ,y Ref,i ):

[0035] x Ref,i =x i +v x *(Ref-i)*T

[0036] y Ref,i =y i +v y *(Ref-i)*T

[0037] Among them, (x i ,y i ) is the measured value of the i-th frame trace, (v x ,v y ) is the speed value calculated in step 42, (x Ref,i ,y Ref,i ) is the measuring point (x i ,y i ) is projected onto the projection coordinate value of the Ref frame, Ref is the frame count corresponding to the reference frame, i is the measurement point (x i ,y i ) corresponds to the frame count.

[0038] Furthermore, in step 5, the steps for implementing the clustering of trace data in the parameter space using the DBSCAN algorithm are as follows:

[0039] Step 51: Determine the clustering parameters and minimum number of points m for the DBSCAN algorithm pts ;

[0040] Step 52: Transform the point data (x Ref,i ,y Ref,i ,v x ,v y ) and the existing trace data in the parameter space For comparison, if (x Ref,i ,y Ref,i ,v x ,v y )exist In the neighborhood conditional on the clustering parameter, (x Ref,i ,y Ref,i ,v x ,v y )join in The corresponding data set Φ m In the data set, the number of data in the data set is increased by one;

[0041] Step 53: If (x Ref,i ,y Ref,i ,v x ,v y ) does not belong to any In the neighborhood of the center, then judge (x Ref,i ,y Ref,i ,v x ,v y ) belongs to Φ m The data in the neighborhood is centered; if it belongs to, the point data (x Ref,i ,y Ref,i ,v x ,v y ) Add Φ m , the number of data in the data set increases by one; otherwise, (x Ref,i ,y Ref,i ,v x ,v y ) is regarded as the new point data in the parameter space, and the number of data in the corresponding data set is assigned to one;

[0042] Step 54: After traversing the scanning cycle, the number of data in the data set is greater than or equal to m pts The point data of is marked as the core point, and all the data contained in the data set are marked as a cluster to complete the clustering.

[0043] Furthermore, in step 7, the coordinate information of the target in each scanning cycle is inferred based on the characteristic information of the target in step 6, and the implementation method is as follows:

[0044] Assume that the characteristic information of the target corresponding to the cluster formed by clustering is (x Ref ,y Ref ,v xRef ,v yRef ), then the coordinates of the measurement point (x k ,y k )for:

[0045] x k =x Ref +v xRef *(k-Ref)*T

[0046] y k =y Ref +v yRef *(k-Ref)*T

[0047] Among them, k is the frame count corresponding to the target track during the backpropagation process, (x k ,y k) is the coordinate value of the target in the kth frame obtained by reverse calculation, (x Ref ,y Ref ,v xRef ,v yRef ) is the centroid of the cluster obtained by parameter space clustering, including the coordinate values ​​of the target in the x-axis and y-axis directions (x Ref ,y Ref ) and the components of the target velocity in the x-axis and y-axis directions (v xRef ,v yRef ).

[0048] Due to the adoption of the above technical solution, the present invention has the following advantages and positive effects compared with the prior art:

[0049] This method is suitable for target track initiation in dense clutter environments. Batch processing can effectively suppress clutter. After obtaining the target feature information corresponding to the cluster, the track point coordinate information can be directly inferred, accurately initiating the target track and avoiding the initiation of false tracks. It has low computational complexity, easy implementation, and high engineering application value. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following is a brief introduction to the drawings required for describing the embodiments. 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 inventive work. In the drawings:

[0051] Figure 1 A flow chart of the method proposed in the present invention;

[0052] Figure 2 The schematic diagram of the tracks formed by four simulated targets in four radar scanning cycles;

[0053] Figure 3 The distribution of all measurement points obtained by the radar 4 scans;

[0054] Figure 4 is the track starting result obtained by the method proposed in the present invention;

[0055] Figure 5 is the track starting result obtained by using the 3 / 4 logic method. DETAILED DESCRIPTION

[0056] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0057] refer to Figure 1 The present invention discloses a batch processing-based track initiation method, which includes the following steps: step 1: agglomerating the measurement points in each scanning cycle; step 2: selecting the measurement points that can be selected for calculating the feature parameters in combination with the target motion characteristics; step 3: classifying the measurement points according to the target motion characteristics, and creating feature parameter linked lists respectively; step 4: transforming the feature parameter information of the measurement points after being screened in step 2 for each feature parameter linked list in step 3, and projecting it to the speed-coordinate parameter space of a certain reference frame as the point track data in the parameter space; step 5: repeating steps 2 to 4, traversing the scanning cycle, and using the DBSCAN algorithm to complete the point track data clustering in the parameter space; step 6: obtaining the centroid of the cluster formed after the DBSCAN algorithm clustering in the parameter space as the feature information of the target corresponding to the cluster; step 7: inferring the coordinate information of the target in each scanning cycle based on the feature information of the target in step 6 to complete the track initiation.

[0058] The above steps are described in detail as follows:

[0059] Step 1: Get the measurement set D1, D2, ..., D of N scan cycles N , for n measurement points D in each scanning cycle i ={(x1,y1),(x2,y2),…,(x n ,y n )} for cohesion;

[0060] Furthermore, in step 1, the steps for condensing the n measurement points in each scanning cycle are as follows:

[0061] Step 11: Traverse the traces within the scanning period to determine whether any traces meet the condensation condition;

[0062] Step 12: In the set of points that meet the aggregation conditions, perform equal-weight fusion aggregation on the measurement values ​​of the points;

[0063] Step 13: Add the condensed measurement values ​​to the measurement set D of the scan cycle i ;

[0064] Step 14: Delete measurement collection D i The measured values ​​of the condensed traces in .

[0065] Step 2: For the measurement set D of the i-th scanning cycle i And the measurement set D of the i+1th scanning cycle i+1 , combined with the maximum speed v of the target track max , filter out the measurement points that can be selected to calculate the characteristic parameters;

[0066] Furthermore, in step 2, the rules for selecting measurement points that can be used to calculate characteristic parameters are as follows:

[0067] Considering that the target track has the maximum moving speed v max =500m / s, assuming (x i ,y i ) is the point of the i-th frame, (x j ,y j ) is the point of the jth frame, and the scanning period is T = 1s. Therefore, the premise for calculating the characteristic parameters should meet the following conditions:

[0068]

[0069] Among them, (x i ,y i ) is the measured value of the point trace in the x-axis and y-axis directions of the i-th frame, (x j ,y j ) is the measured value of the j-th frame trace in the x-axis and y-axis directions, v max is the maximum speed of the target track, T is the scanning period, i and j are the measurement points (x i ,y i ) and (x j ,y j ) corresponds to the frame count.

[0070] Step 3: Classify the measurement points according to the target motion characteristics and create a feature parameter list for each;

[0071] Furthermore, in step 3, the measurement points are classified according to the target motion characteristics, and the method of creating a feature parameter list is as follows:

[0072] Taking two-dimensional space as an example, let (x i ,y i ) is the coordinate of the point in the i-th frame, (x j ,y j ) is the coordinate of the point in the jth frame, where j>i, (v x ,v y ) is the velocity calculated from the coordinates of the two points, v Ref If the reference speed is set, the calculated speed will produce the following 5 results:

[0073]

[0074] Create 5 feature parameter lists corresponding to the above 5 results;

[0075] Calculate the speed of each pair of points that meet the conditions in step 2, and store the characteristic parameters in different linked lists according to the different speeds.

[0076] Step 4: For each feature parameter list in step 3, transform the feature parameter information of the measurement points filtered in step 2 and project it into the velocity-coordinate parameter space of a certain reference frame as the point trace data of the parameter space;

[0077] Furthermore, in step 4, the steps of projecting to the velocity-coordinate parameter space are as follows:

[0078] Step 41: Select the parameter projection frame as the Ref frame;

[0079] Step 42: Get the measurement points (x i ,y i ) and (x i+1 ,y i+1 )Calculated speed:

[0080]

[0081] Step 43: Calculate the measurement point of the i-th frame with the velocity (v x ,v y ) is projected onto the coordinates (x Ref,i ,y Ref,i ):

[0082] x Ref,i =x i +v x *(Ref-i)*T

[0083] y Ref,i =y i +v y *(Ref-i)*T

[0084] Among them, (x i ,y i ) is the measured value of the i-th frame trace, (v x ,v y ) is the speed value calculated in step 42, (x Ref,i ,y Ref,i ) is the measuring point (x i ,y i ) is projected onto the projection coordinate value of the Ref frame, Ref is the frame count corresponding to the reference frame, i is the measurement point (x i ,y i ) corresponds to the frame count.

[0085] Step 5: Repeat steps 2 to 4, traverse the scanning cycle, and use the DBSCAN algorithm to complete the point data clustering in the parameter space;

[0086] Furthermore, in step 5, the steps for implementing the clustering of trace data in the parameter space using the DBSCAN algorithm are as follows:

[0087] Step 51: Determine the clustering parameters and minimum number of points m for the DBSCAN algorithm pts ;

[0088] Considering that the point data at this time is four-dimensional data (x Ref,i ,y Ref,i ,v x ,v y ), so the clustering parameter is selected as (x th ,y th ,v xth ,v yth ), combined with the target's motion characteristics, we have x th =150,y th =150,v xth =100,v yth =100, when the scan cycle is 4, set m pts =3;

[0089] Step 52: Transform the point data (x Ref,i ,y Ref,i ,v x ,v y ) and the existing trace data in the parameter space For comparison, if (x Ref,i ,y Ref,i ,v x ,v y )exist In the neighborhood conditional on the clustering parameters, it satisfies:

[0090]

[0091]

[0092]

[0093]

[0094] Then (x Ref,i ,y Ref,i ,v x ,v y )join in The corresponding data set Φ m In the data set, the number of data in the data set is increased by one;

[0095] Step 53: If (x Ref,i ,y Ref,i ,v x ,vy ) does not belong to any In the neighborhood of the center, then judge (x Ref,i ,y Ref,i ,v x ,v y ) belongs to Φ m The data in the neighborhood is centered; if it belongs to, the point data (x Ref,i ,y Ref,i ,v x ,v y ) Add Φ m , the number of data in the data set increases by one; otherwise, (x Ref,i ,y Ref,i ,v x ,v y ) is regarded as the new point data in the parameter space, and the number of data in the corresponding data set is assigned to one;

[0096] Step 54: After traversing the scanning cycle, the number of data in the data set is greater than or equal to m pts The point data of is marked as the core point, and all the data contained in the data set are marked as a cluster to complete the clustering.

[0097] Step 6: Obtain the centroid of the cluster formed after the DBSCAN algorithm clustering is completed in the parameter space as the feature information of the target corresponding to the cluster;

[0098] Step 7: Based on the characteristic information of the target in step 6, the coordinate information of the target in each scanning cycle is reversed to complete the track start.

[0099] Furthermore, in step 7, the coordinate information of the target in each scanning cycle is inferred based on the characteristic information of the target in step 6, and the implementation method is as follows:

[0100] Assume that the characteristic information of the target corresponding to the cluster formed by clustering is (x Ref ,y Ref ,v xRef ,v yRef ), then the coordinates of the measurement point (x k ,y k )for:

[0101] x k =x Ref +v xRef *(k-Ref)*T

[0102] y k =y Ref +v yRef *(k-Ref)*T

[0103] Among them, k is the frame count corresponding to the target track during the backpropagation process, (x k ,y k ) is the coordinate value of the target in the kth frame obtained by reverse calculation, (x Ref ,y Ref ,v xRef ,v yRef ) is the centroid of the cluster obtained by parameter space clustering, including the coordinate values ​​of the target in the x-axis and y-axis directions (x Ref ,y Ref ) and the components of the target velocity in the x-axis and y-axis directions (v xRef ,v yRef ).

[0104] The effects of the present invention are verified through the following examples and simulation experiments.

[0105] The simulation conditions are as follows: the simulation sets four targets with initial positions of (8000, 6000), (7000, 2000), (1000, 4000) and (4000, 8000) for uniform linear motion, with velocities of (-200, -200), (-200, 200), (200, 200) and (200, -200) respectively. The targets are tracked using a 2D radar with a scanning period of T = 1s and a direction-finding error of σ. x =σ y =50m. The environmental clutter is generated according to the following rules: The number of clutter in a unit area of ​​radar follows Poisson distribution, given parameter λ = 1 (clutter / km 2 ), first generate a uniformly distributed random number γ in the interval (0,1), and then calculate J using the following formula:

[0106]

[0107] Then J is the number of clutter generated per unit area. After J is determined, the area within each scanning cycle is divided. Each area generates J clutter, which is randomly distributed uniformly within the area. After all areas are processed, the clutter distribution for the scanning cycle is completed.

[0108] Figure 2 Schematic diagram of the tracks formed by four simulated targets during four radar scanning cycles.

[0109] Figure 3 The distribution of all measurement points obtained by the radar after four scans. It can be seen that the environmental clutter is very dense and the target points are submerged in it and difficult to distinguish.

[0110] Figure 4 The track starting result obtained by the method proposed in this invention is given, where the red line is the formed track.

[0111] Figure 5 The results of track initiation using the 3 / 4 logic method are given, where the red line represents the resulting track.

[0112] Monte Carlo simulations were performed 500 times using the method proposed in the present invention and the traditional Hough transform starting track. The results showed that the average running time of the method proposed in the present invention was 0.56s, and the average running time of the traditional Hough transform starting track was 1.21s.

[0113] From the simulation example, it can be seen that in a dense clutter environment, the 3 / 4 logic method initiates multiple batches of tracks including the real target track, generating a large number of false tracks, which interferes with the subsequent tracking process of the target. The track initiation method of the present invention can effectively suppress clutter and accurately initiate the target track, which is more superior for track initiation in a dense clutter environment. Regarding the step of back-calculating the track points after accumulating the parameter space, the traditional Hough transform track initiation algorithm needs to transform from the ρ-θ parameter space to the xy coordinate space. The track initiation method of the present invention obtains the characteristic information (xy coordinate space) of the target corresponding to the cluster formed by the cluster. Ref ,y Ref ,v xRef ,v yRef ), the track point coordinate information can be directly inferred, with low computational complexity, small storage capacity, and greater engineering application value.

[0114] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in the present invention should be included in the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be based on the scope of protection of the claims.

Claims

1. A track initiation method based on batch processing, characterized in that: The following steps are involved: Step 1: Get the measurement set D1, D2, ..., D of N scan cycles N , for n measurement points D in each scanning cycle i ={(x1,y1),(x2,y2),…,(x n ,y n )} for cohesion; Step 2: For the measurement set D of the i-th scanning cycle i And the measurement set D of the i+1th scanning cycle i+1 , combined with the maximum speed v of the target track max , filter out the measurement points selected for calculating characteristic parameters; In step 2, the rules for selecting the measurement points for calculating the characteristic parameters are as follows: Considering that the target track has the maximum moving speed v max , assuming (x i ,y i ) is the point of the i-th frame, (x j ,y j ) is the point of the jth frame, and the scanning period is T. Therefore, the premise for calculating the feature parameters should meet the following requirements: Among them, (x i ,y i ) is the measured value of the point trace in the x-axis and y-axis directions of the i-th frame, (x j ,y j ) is the measured value of the j-th frame trace in the x-axis and y-axis directions, v max is the maximum speed of the target track, T is the scanning period, i and j are the measurement points (x i ,y i ) and (x j ,y j ) corresponding frame count; Step 3: Classify the measurement points according to the target motion characteristics and create a feature parameter list for each; Step 4: For each feature parameter list in step 3, transform the feature parameter information of the measurement points filtered in step 2 and project it into the velocity-coordinate parameter space of a certain reference frame as the point trace data of the parameter space; Step 5: Repeat steps 2 to 4, traverse the scanning cycle, and use the DBSCAN algorithm to complete the point data clustering in the parameter space; In step 5, the steps for implementing the clustering of point trace data in the parameter space using the DBSCAN algorithm are as follows: Step 51: Determine the clustering parameters and minimum number of points m for the DBSCAN algorithm pts ; Step 52: Transform the point data (x Ref,i ,y Ref,i ,v x ,v y ) and the mth point data in the parameter space For comparison, if (x Ref,i ,y Ref,i ,v x ,v y )exist In the neighborhood conditional on the clustering parameter, (x Ref,i ,y Ref,i ,v x ,v y )join in The corresponding data set Φ m In the data set, the number of data in the data set is increased by one; Step 53: If (x Ref,i ,y Ref,i ,v x ,v y ) does not belong to any In the neighborhood of the center, then judge (x Ref,i ,y Ref,i ,v x ,v y ) belongs to Φ m The data in the neighborhood is centered; if it belongs to, the point data (x Ref,i ,y Ref,i ,v x ,v y ) Add Φ m , the number of data in the data set increases by one; otherwise, (x Ref,i ,y Ref,i ,v x ,v y ) is regarded as the new point data in the parameter space, and the number of data in the corresponding data set is assigned to one; Step 54: After traversing the scanning cycle, the number of data in the data set is greater than or equal to m pts The point data of is marked as the core point, and all the data contained in the data set are marked as a cluster, completing the clustering; Step 6: Obtain the centroid of the cluster formed after the DBSCAN algorithm clustering is completed in the parameter space as the feature information of the target corresponding to the cluster; Step 7: Based on the characteristic information of the target in step 6, the coordinate information of the target in each scanning cycle is reversed to complete the track start.

2. A batch processing based track initiation method according to claim 1, characterized in that: In step 1, the steps for condensing n measurement points in each scanning cycle are as follows: Step 11: Traverse the traces within the scanning period to determine whether any traces meet the condensation condition; Step 12: In the set of points that meet the aggregation conditions, perform equal-weight fusion aggregation on the measurement values ​​of the points; Step 13: Add the condensed measurement values ​​to the measurement set D of the scan cycle i ; Step 14: Delete measurement collection D i The measured values ​​of the condensed traces in .

3. The batch processing-based track initiation method according to claim 1, characterized in that: In step 3, the measurement points are classified according to the target motion characteristics, and the method of creating feature parameter lists is as follows: Taking two-dimensional space as an example, let (x i ,y i ) is the coordinate of the point in the i-th frame, (x j ,y j ) is the coordinate of the point in the jth frame, where j>i, (v x ,v y ) is the velocity calculated from the coordinates of the two points, v Ref If the reference speed is set, the calculated speed will produce the following 5 results: Create 5 feature parameter lists corresponding to the above 5 results; Calculate the speed of each pair of points that meet the conditions in step 2, and store the characteristic parameters in different linked lists according to the different speeds.

4. The batch processing-based track initiation method according to claim 1, characterized in that: In step 4, the steps for projecting into the velocity-coordinate parameter space are as follows: Step 41: Select the parameter projection frame as the Ref frame; Step 42: Get the measurement points (x i ,y i ) and (x i+1 ,y i+1 )Calculated speed: Step 43: Calculate the measurement point of the i-th frame with the velocity (v x ,v y ) is projected onto the coordinates (x Ref,i ,y Ref,i ): x Ref,i =x i +v x *(Ref-i)*T y Ref,i =y i +v y *(Ref-i)*T Among them, (x i ,y i ) is the measured value of the i-th frame trace, (v x ,v y ) is the speed value calculated in step 42, (x Ref,i ,y Ref,i ) is the measuring point (x i ,y i ) is projected onto the projection coordinate value of the Ref frame, Ref is the frame count corresponding to the reference frame, i is the measurement point (x i ,y i ) corresponds to the frame count.

5. The track initiation method based on batch processing according to claim 1, characterized in that: In step 7, the coordinate information of the target in each scanning cycle is inferred based on the characteristic information of the target in step 6. The implementation method is as follows: Assume that the characteristic information of the target corresponding to the cluster formed by clustering is (x Ref ,y Ref ,v xRef ,v yRef ), then the coordinates of the measurement point (x k ,y k )for: x k =x Ref +v xRef *(k-Ref)*T y k =y Ref +v yRef *(k-Ref)*T Among them, k is the frame count corresponding to the target track during the backpropagation process, (x k ,y k ) is the coordinate value of the target in the kth frame obtained by reverse calculation, (x Ref ,y Ref ,v xRef ,v yRef ) is the centroid of the cluster obtained by parameter space clustering, including the coordinate values ​​of the target in the x-axis and y-axis directions (x Ref ,y Ref ) and the components of the target velocity in the x-axis and y-axis directions (v xRef ,v yRef ).

Images