Interrupted track interconnection judgment method based on multi-dimensional feature information

Abstract

The present invention relates to a method for determining the interconnection of interrupted tracks based on multi-dimensional feature information, and belongs to the technical field of track interconnection. In order to solve the problems existing in the prior art, the method includes the following steps: 1. Determine the set of new and old tracks, and filter the head and tail points of the new and old tracks; 2. Predict the target prediction state at the middle moment of the new and old tracks according to different motion models; 3. Calculate the spatial distance between the two predicted positions, determine the unified measure of spatial consistency between the two tracks, and define its membership function; 4. Determine the membership function of AD amplitude consistency; 5. Constitute spatial consistency and AD amplitude consistency as a feature judgment basis for determining whether the new and old tracks correspond to the same target; 6. Construct an interconnection relationship matrix that describes the correspondence between the new and old tracks, and solve it. The method realizes a relatively long time, non-one-to-one interrupted track robust association judgment method, and the accuracy of the method is high.

Description

Technical Field

[0001] The invention relates to a track interconnection judgment method, in particular to a method for judging interrupted track interconnection based on multi-dimensional feature information, and belongs to the technical field of track interconnection. Background Art

[0002] Early warning detection radars are affected by many factors such as target maneuvers, electronic interference, detection distance limitations, speed blind spots, and non-uniform detection backgrounds. Track interruptions often occur, resulting in many track fragments. Track interruptions result in multiple short tracks for the same target, which seriously affects the quality of early warning detection intelligence. Therefore, in the subsequent intelligence compilation and analysis process, it is necessary to analyze and judge the interrupted tracks to determine which track fragments belong to the same target, and then associate them together to form a continuous observation of the target.

[0003] During the track interruption, due to the lack of observation information, the unknown and difficult description of the target motion model in the tracking system increases dramatically, and even a combination of multiple motion models may occur, which will inevitably make track prediction more difficult and seriously affect the subsequent track association judgment. In view of the above technical difficulties, it is necessary to conduct in-depth research on the target motion model description method, track prediction method and association judgment method for longer interruption time to solve the problem of difficult track association.

[0004] The interactive multi-model interrupted track continuation association algorithm uses a one-step prediction based on the interactive multi-model to predict backward from the new track until the last update point of the old track, and then perform a chi-square test, using the log-likelihood function as the cost function for the final two-dimensional optimal matching. Using the interactive multi-model-based method, the new track is predicted backward while the old track is predicted forward. After the two are predicted to the same time, the predicted values ​​are tested by a chi-square test, and a cost function matrix is ​​constructed for two-dimensional optimal allocation to obtain associated track pairs. The above two methods will result in large prediction errors for the target in the presence of measurement errors and the limitations of existing filtering algorithms for track tracking, leading to mis-association and missed association.

[0005] The algorithm for connecting interrupted tracks with multiple hypothesis motion models based on prior information is based on the idea of ​​multiple hypothesis, establishes multiple possible target motion models and implements track prediction, and describes the fuzzy matching relationship between the predicted track and the new starting track based on the fuzzy correlation function of position and velocity information. The association effect of this algorithm decreases significantly as the interruption interval becomes larger and larger, and because multiple hypotheses are required, in order to more accurately associate, selecting too many models will also increase the amount of calculation of the algorithm.

[0006] A statistical double threshold-based interrupted track association algorithm introduces the statistical double threshold principle and increases the number of associated samples for chi-square distribution threshold detection. When the target undergoes large maneuvers, the algorithm cannot accurately describe the maneuvering motion in the interrupted interval, and the association effect deteriorates seriously. Summary of the invention

[0007] The purpose of the present invention is to solve the practical engineering problems such as the availability of multi-dimensional information in actual engineering scenes, long track interruption time, and fast change of target motion mode, and to provide a method for determining the interconnection of interrupted tracks based on multi-dimensional feature information. The method makes full use of multi-dimensional information such as spatial position, motion state, and AD amplitude, and adopts technical methods such as fuzzy membership function, global nearest neighbor, and bidirectional filtering prediction based on the state information of the first and last points and multiple models to achieve a robust association determination method for interrupted tracks for a long time and not one-to-one.

[0008] In order to solve the above problems, the present application is implemented through the following technical solutions:

[0009] A method for determining interrupted track interconnection based on multi-dimensional feature information, which is special in that it includes the following steps:

[0010] Step 1: determine the set of new and old tracks, and filter the head and tail points of the new and old tracks based on the head and tail point motion information;

[0011] Step 2: predict the target prediction state at the middle moment of the new and old tracks based on the information of the starting and ending points of the new and old tracks obtained in step 1 and different motion models;

[0012] Step 3: According to the target prediction state at the middle moment of the new and old tracks obtained in step 2, the spatial distance between the two predicted positions is calculated, a unified measure of spatial consistency between the two tracks is determined, and its membership function is defined;

[0013] Step 4, determine the membership function of AD amplitude consistency;

[0014] Step 5, the spatial consistency in step 3 and the AD amplitude consistency in step 4 are used as a feature judgment basis for judging whether the new and old tracks correspond to the same target;

[0015] Step 6: Construct an interconnection relationship matrix describing the correspondence between the new and old tracks and solve it.

[0016] Furthermore, the step 1 specifically includes the following steps:

[0017] Step 1.1: Determine the set of new and old tracks: When compiling and analyzing early warning detection intelligence, the set of all tracks is ,in is the track segment, ,

[0018] in Representative The starting time of the track, Indicates The time when the track ends;

[0019] set up hour ,definition is the earliest end time of the track segment, is the latest start time of the track segment;

[0020] Then the old track set is , ; The new track set is , , ;

[0021] Step 1.2, filtering the beginning and end points of the new and old tracks: using the motion state information of the tracks at the front and rear ends, perform positive and reverse filtering from the start time of the old track and the end time of the new track to the interruption time respectively;

[0022] For a pair of new and old tracks , the old track is Forward filtering to Moment, get ;

[0023] New track Inverse filtering to Moment, get .

[0024] Furthermore, the step 2 comprises the following steps:

[0025] Step 2.1: Determine the target prediction state of the two tracks at the intermediate time:

[0026] Based on the filtering state of the first and last points of the new and old tracks, that is, Update status of old tracks at all times and Update status of new tracks at all times , towards the middle moment Predict the target state of the new and old tracks, that is,

[0027] , ,

[0028] Where: and is the state transfer matrix, , Respectively represent the target prediction states of the old track and the new track at the intermediate moments;

[0029] Step 2.2, select different motion models, determine the state transfer matrix, and obtain the target prediction state of the two tracks at the intermediate time;

[0030] The target moves in three-dimensional space, and the state of the target is predicted using the state transfer matrices of three motion modes: uniform straight line, uniformly accelerated straight line, and uniform turning.

[0031] For uniform linear and uniformly accelerated linear motion modes, since there is no coupling relationship between different vector coordinates, the state transfer matrix of the target used at this time is It consists of three one-dimensional motion state transfer matrices, namely ,

[0032] Where: is a one-dimensional motion state transfer matrix. When uniform straight line prediction is performed , when performing uniform acceleration straight line prediction ;

[0033] , ,

[0034] For the uniform turning motion mode, due to the coupling relationship between different vector coordinates, the state transfer matrix of the target used at this time is It consists of a two-dimensional uniform turning motion state transfer matrix and a one-dimensional uniform linear motion state transfer matrix, that is, ,

[0035] In the formula and They represent the state transfer matrix corresponding to the target making a uniform turning motion in the xy plane and the state transfer matrix of the target making a uniform motion on the z axis respectively;

[0036] , ;

[0037] According to different motion models, the corresponding state transfer matrix is ​​selected, and the state transfer matrix at the intermediate moment between the new and old tracks is obtained respectively, so as to obtain the target predicted state of the two tracks at the intermediate moment.

[0038] Furthermore, the step 3 specifically includes the following steps:

[0039] Step 3.1, calculate the spatial distance between the two predicted positions:

[0040] Step 3.1.1: For the three-dimensional uncoupled uniform motion model CV, the spatial distance between the two predicted positions is for: ;

[0041] Step 3.1.2: For the three-dimensional uncoupled uniform acceleration motion model CA, the spatial distance between the two predicted positions is for: ;

[0042] Step 3.1.3: Based on the heading angles of the first and last points of the new and old tracks and , calculate the mean turning rate , respectively according to and , the spatial distance between the two predicted positions for: ;

[0043] in: ;

[0044] Step 3.1.4: Based on the average turning rate at the end point of the old track and its previous time t , based on the average turning rate at the first point of the new track and its subsequent time t , respectively according to and

[0045] , the spatial distance between the two predicted positions for: ;

[0046] in: , ,

[0047] Step 3.2: The maximum spatial distance between the new and old tracks is defined as a unified measure of spatial consistency between two tracks, and its membership function is defined as:

[0048] in: ,

[0049] is the minimum spatial distance, that is: ,

[0050] The membership function is: .

[0051] Furthermore, the step 4 is specifically as follows:

[0052] The maximum AD amplitude in the track is selected as its standard value, and the corresponding membership function is defined as: , ,

[0053] Where: and is the AD amplitude value corresponding to the new and old tracks, The statistical variance of the AD amplitude of the same track is obtained by statistically analyzing historical data or taking the maximum AD amplitude value between the new and old tracks.

[0054] Further, the step 5 is as follows: the association decision value after integrating the membership function of the spatial consistency in step 3 and the AD amplitude consistency in step 4 with their corresponding weights As a basis for judgment:

[0055] ,in is the weight.

[0056] Furthermore, the step 6 specifically includes the following steps:

[0057] Step 6.1: Construct the interconnection relationship matrix and the association decision matrix describing the correspondence between the new and old tracks:

[0058] definition The association decision matrix To describe the associated decision value of the new and old tracks, ,

[0059] Where: Indicates Old track and The association decision value of the new track, m and n represent the number of old tracks and the number of new tracks involved in the association decision respectively;

[0060] definition The interconnection matrix To describe the correspondence between the new and old tracks, ,

[0061] Indicates the judgment Old track and The new tracks are not associated, Indicates the judgment Old track and New track associations;

[0062] Since there is a non-one-to-one correspondence between the new and old tracks, the interconnection matrix satisfies , ;

[0063] Step 6.2, solving the interconnection relationship matrix of step 6.1, specifically comprising the following steps:

[0064] Step 6.2.1, initial timing judgment:

[0065] Based on the end time of the old track Less than the start time of the new track Constraint relationship, set the matrix elements that do not meet this condition in step 4.1 to zero, that is: ,

[0066] The interruption time is greater than a certain threshold The track of is no longer associated with judgment, that is ;

[0067] Step 6.2.2: Preliminary judgment of motion relationship:

[0068] At maximum speed , Maximum acceleration , minimum speed and maximum turning rate The non-zero elements in the interconnection relationship matrix are screened and updated, and the element values ​​corresponding to the track pairs that meet the following conditions are set to 1, and the rest are set to 0:

[0069] Calculate the Euclidean distance between the end point of the old track and the start point of the new track in the track pair , and the speed constraint is imposed on the track pair according to the speed obtained by dividing the distance by the interruption time:

[0070] , ,

[0071] According to the speed at the end of the old track and the speed at the start of the new track Divide the acceleration of the track pair by the interruption time to limit the acceleration of the track pair: ,

[0072] According to the previously calculated mean turning rate To limit the turning rate of the track pair: ,

[0073] Among them, the maximum speed Selected based on historical track data of air and surface targets; maximum acceleration Set based on air and surface target characteristics or historical data; minimum speed Set based on historical data of air and surface targets;

[0074] Maximum turning rate of surface targets Set according to the target characteristics of the surface target; the maximum turning rate of the air target It is set according to the characteristics of the air target or according to the maximum turning rate of the non-interrupted track multiplied by the magnification factor. , is the magnification factor, is the maximum turning rate without interruption of the track, It is obtained by traversing the track segment;

[0075] Step 6.2.3, global optimal decision:

[0076] Select the interconnection matrix The element with the largest association decision value corresponding to all the remaining non-zero elements in the , determines that the corresponding new and old tracks are successfully associated, and then the row and column elements of the element are All elements of the interconnection relationship matrix are set to zero, and the judgment is iterated in sequence, and all elements are set to zero, so that the judgment results of the association between all new and old tracks are obtained.

[0077] This application draws on effective technical measures such as two-way filtering prediction and global optimization, makes full use of multi-dimensional information such as spatial position, motion state and AD amplitude, and adopts technical methods such as fuzzy membership function, global nearest neighbor and two-way filtering prediction based on the state information of the first and last points and multiple models to realize a long-term, non-one-to-one interrupted track robust association judgment method with high accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 is a flow chart of the present invention;

[0079] Figure 2 This is the first track association diagram;

[0080] Figure 3 It is a diagram of the association effect of the corresponding methods for different interruption durations;

[0081] Figure 4 It is the correlation effect diagram of the corresponding methods with different numbers of tracks;

[0082] Figure 5 This is the correlation effect diagram of the corresponding methods for different track missing rates. DETAILED DESCRIPTION

[0083] The following is a specific implementation of the present invention with reference to the accompanying drawings to further illustrate the composition of the present invention.

[0084] Embodiment 1. A method for determining interrupted track interconnection based on multi-dimensional feature information, such as Figure 1 As shown, the following steps are included:

[0085] Step 1: determine the set of new and old tracks, and filter the head and tail points of the new and old tracks based on the head and tail point motion information;

[0086] Step 2: predict the target prediction state at the middle moment of the new and old tracks based on the information of the first points of the new and old tracks obtained in step 1 and different motion models;

[0087] Step 3: According to the target prediction state at the middle moment of the new and old tracks obtained in step 2, the spatial distance between the two predicted positions is calculated, a unified measure of spatial consistency between the two tracks is determined, and its membership function is defined;

[0088] Step 4, determine the membership function of AD amplitude consistency;

[0089] Step 5, the spatial consistency in step 3 and the AD amplitude consistency in step 4 are used as a feature judgment basis for judging whether the new and old tracks correspond to the same target;

[0090] Step 6: Construct an interconnection relationship matrix describing the correspondence between the new and old tracks and solve it.

[0091] Furthermore, the step 1 specifically includes the following steps:

[0092] Step 1.1: Determine the set of new and old tracks: When compiling and analyzing early warning detection intelligence, the set of all tracks is ,in is the track segment, ,

[0093] in Representative The starting time of the track, Indicates The time when the track ends;

[0094] set up hour ,definition is the earliest end time of the track segment, is the latest start time of the track segment;

[0095] Then the old track set is , ;

[0096] The new track set is , , .

[0097] Step 1.2: Filter the beginning and end points of the new and old tracks:

[0098] Using the motion state information of the front and rear tracks, positive and reverse filtering are performed from the start time of the old track and the end time of the new track to the interruption time respectively;

[0099] For a pair of new and old tracks , the old track is Forward filtering to Moment, get ;

[0100] New track Inverse filtering to Moment, get .

[0101] Furthermore, the step 2 comprises the following steps:

[0102] Step 2.1: Determine the target prediction state of the two tracks at the intermediate time:

[0103] Based on the filtering state of the first and last points of the new and old tracks, that is, Update status of old tracks at all times and Update status of new tracks at all times , towards the middle moment Predict the target state of the new and old tracks, that is,

[0104] , ,

[0105] Where: and is the state transfer matrix, , Respectively represent the target prediction states of the old track and the new track at the intermediate moments;

[0106] Step 2.2, select different motion models, determine the state transfer matrix, and obtain the target prediction state of the two tracks at the intermediate time;

[0107] The target moves in three-dimensional space, and the state of the target is predicted using the state transfer matrices of three motion modes: uniform straight line, uniformly accelerated straight line, and uniform turning.

[0108] For uniform linear and uniformly accelerated linear motion modes, since there is no coupling relationship between different vector coordinates, the state transfer matrix of the target used at this time is It consists of three one-dimensional motion state transfer matrices, namely ,

[0109] Where: is a one-dimensional motion state transfer matrix. When uniform straight line prediction is performed , when performing uniform acceleration straight line prediction ;

[0110] , ,

[0111] For the uniform turning motion mode, due to the coupling relationship between different vector coordinates, the state transfer matrix of the target used at this time is It consists of a two-dimensional uniform turning motion state transfer matrix and a one-dimensional uniform linear motion state transfer matrix, that is, ,

[0112] In the formula and They represent the state transfer matrix corresponding to the target making a uniform turning motion in the xy plane and the state transfer matrix of the target making a uniform motion on the z axis respectively;

[0113] , ;

[0114] According to different motion models, the corresponding state transfer matrix is ​​selected, and the state transfer matrix at the intermediate moment between the new and old tracks is obtained respectively, so as to obtain the target predicted state of the two tracks at the intermediate moment.

[0115] Furthermore, the step 3 specifically includes the following steps:

[0116] Step 3.1, calculate the spatial distance between the two predicted positions:

[0117] Step 3.1.1: For the three-dimensional uncoupled uniform motion model CV, the spatial distance between the two predicted positions is for: ;

[0118] Step 3.1.2: For the three-dimensional uncoupled uniform acceleration motion model CA, the spatial distance between the two predicted positions is for: ;

[0119] Step 3.1.3: Based on the heading angles of the first and last points of the new and old tracks and , calculate the mean turning rate , respectively according to and , the spatial distance between the two predicted positions for: ;

[0120] in: ;

[0121] Step 3.1.4: Based on the average turning rate at the end point of the old track and its previous time t , based on the average turning rate at the first point of the new track and its subsequent time t , respectively according to and

[0122] , the spatial distance between the two predicted positions for: ;

[0123] in: , ;

[0124] Step 3.2: The maximum spatial distance between the new and old tracks is defined as a unified measure of spatial consistency between two tracks, and its membership function is defined as:

[0125] in: ,

[0126] is the minimum spatial distance, that is: ,

[0127] The membership function is: .

[0128] Furthermore, the step 4 is specifically as follows:

[0129] The maximum AD amplitude in the track is selected as its standard value, and the corresponding membership function is defined as: , ,

[0130] Where: and is the AD amplitude value corresponding to the new and old tracks, The statistical variance of the AD amplitude of the same track is obtained by statistically analyzing historical data or taking the maximum AD amplitude value between the new and old tracks.

[0131] Further, the step 5 is as follows: the association decision value after integrating the membership function of the spatial consistency in step 3 and the AD amplitude consistency in step 4 with their corresponding weights As a basis for judgment:

[0132] ,in The weight is not fixed. As the interruption time increases, the weight of the spatial consistency attribute will become smaller. For example, 0.7 is selected within 1 minute, 0.5 can be adjusted to 0.5 for 1 to 5 minutes, and 0.3 for more than 5 minutes. At the same time, the spatial consistency weight of the sea surface target can be further increased to 0.8, and it changes little with the interruption time.

[0133] Furthermore, the step 6 specifically includes the following steps:

[0134] Step 6.1: Construct the interconnection relationship matrix and the association decision matrix describing the correspondence between the new and old tracks:

[0135] definition The association decision matrix To describe the associated decision value of the new and old tracks, ,

[0136] Where: Indicates Old track and The association decision value of the new track, m and n represent the number of old tracks and the number of new tracks involved in the association decision respectively;

[0137] definition The interconnection matrix To describe the correspondence between the new and old tracks, ,

[0138] Indicates the judgment Old track and The new tracks are not associated, Indicates the judgment Old track and New track associations;

[0139] Since there is a non-one-to-one correspondence between the new and old tracks, the interconnection matrix satisfies , ;

[0140] Step 6.2, solving the interconnection relationship matrix of step 6.1, specifically comprising the following steps:

[0141] Step 6.2.1, initial timing judgment:

[0142] Based on the end time of the old track Less than the start time of the new track Constraint relationship, set the matrix elements that do not meet this condition in step 4.1 to zero, that is: ,

[0143] The interruption time is greater than a certain threshold The track of is no longer associated with judgment, that is ;

[0144] Step 6.2.2: Preliminary judgment of motion relationship:

[0145] At maximum speed , Maximum acceleration , minimum speed and maximum turning rate The non-zero elements in the interconnection relationship matrix are screened and updated, and the element values ​​corresponding to the track pairs that meet the following conditions are set to 1, and the rest are set to 0:

[0146] Calculate the Euclidean distance between the end point of the old track and the start point of the new track in the track pair , and the speed constraint is imposed on the track pair according to the speed obtained by dividing the distance by the interruption time:

[0147] , ,

[0148] According to the speed at the end of the old track and the speed at the start of the new track Divide the acceleration of the track pair by the interruption time to limit the acceleration of the track pair: ,

[0149] According to the previously calculated mean turning rate To limit the turning rate of the track pair: ,

[0150] Among them, the maximum speed Selected based on historical track data of air and surface targets; maximum acceleration Set based on air and surface target characteristics or historical data; minimum speed Set based on historical data of air and surface targets;

[0151] Maximum turning rate of surface targets Set according to the target characteristics of the surface target; the maximum turning rate of the air target It is set according to the characteristics of the air target or according to the maximum turning rate of the non-interrupted track multiplied by the magnification factor. , is the magnification factor, is the maximum turning rate without interruption of the track, It is obtained by traversing the track segment;

[0152] Step 6.2.3, global optimal decision:

[0153] Select the interconnection matrix The element with the largest association decision value corresponding to all the remaining non-zero elements in the , determines that the corresponding new and old tracks are successfully associated, and then the row and column elements of the element are All elements of the interconnection relationship matrix are set to zero, and the judgment is iterated in sequence, and all elements are set to zero, so that the judgment results of the association between all new and old tracks are obtained.

[0154] Embodiment 2. This embodiment is a simulation of the method described in Embodiment 1.

[0155] The IMM-TSA algorithm is an interrupted track association algorithm based on interactive multi-model;

[0156] The IMMSDP-TSA algorithm is a state-dependent transfer probability interactive multi-model interruption track association algorithm;

[0157] The MHMPI-TSA algorithm is a multi-hypothesis model interruption track association algorithm based on prior information.

[0158] The method provided by the present invention is compared with the IMM-TSA algorithm, the IMMSDP-TSA algorithm and the MMHPI-TSA algorithm through simulation experiments. The average correct rates of the method are higher than those of the three algorithms. The method has better accuracy and effectiveness when processing interrupted track data sets.

[0159] Experiment 1: A radar is used to detect targets. Assuming there are 100 targets, the target switches state 5 times within 600 sampling cycles, and the track is not lost. The performance of this method is compared with the other three algorithms by setting different interruption durations.

[0160] Experiment 2: The environment is the same as above, the target interruption duration is 100s, the track is not missing, and different numbers of tracks are set to compare the performance of this method with the other three algorithms.

[0161] Experiment 3: The environment is the same as above, there are 100 targets, the interruption time of each target is 100s, and different track loss rates are set to compare the performance of this method with the other three algorithms.

[0162] from Figure 3-Figure 5 The experimental results in show that the average association accuracy of the method proposed in this application is higher than that of the other three algorithms in various scenarios. The increase in the interruption time of the track and the increase in the number of tracks will reduce the performance of the algorithm, but the impact on this method is small.

Claims

1. A method for determining interrupted track interconnection based on multi-dimensional feature information, characterized in that: The following steps are involved: Step 1: determine the set of new and old tracks, and filter the head and tail points of the new and old tracks based on the head and tail point motion information; Step 2: predict the target prediction state at the middle moment of the new and old tracks based on the information of the first points of the new and old tracks obtained in step 1 and different motion models; Step 3: According to the target prediction state at the middle moment of the new and old tracks obtained in step 2, the spatial distance between the two predicted positions is calculated, a unified measure of spatial consistency between the two tracks is determined, and its membership function is defined; Step 4, determine the membership function of AD amplitude consistency; Step 5, the spatial consistency in step 3 and the AD amplitude consistency in step 4 are used as a feature judgment basis for judging whether the new and old tracks correspond to the same target; Step 6: Construct an interconnection relationship matrix describing the correspondence between the new and old tracks and solve it.

2. The method for determining interrupted track interconnection based on multi-dimensional feature information according to claim 1, characterized in that: The step 1 specifically comprises the following steps: Step 1.1: Determine the set of new and old tracks: When compiling and analyzing early warning detection intelligence, the set of all tracks is ,in is the track segment, , in Representative The starting time of the track, Indicates The time when the track ends; set up hour ,definition is the earliest end time of the track segment, is the latest start time of the track segment; Then the old track set is , ; The new track set is , , ; Step 1.2: Filter the beginning and end points of the new and old tracks: Using the motion state information of the front and rear tracks, positive and reverse filtering are performed from the start time of the old track and the end time of the new track to the interruption time respectively; For a pair of new and old tracks , the old track is Forward filtering to Moment, get ; New Track Inverse filtering to Moment, get .

3. A method for determining interrupted track interconnection based on multi-dimensional feature information according to claim 1 or 2, characterized in that: The step 2 comprises the following steps: Step 2.1, determine the target prediction state of the two tracks at the intermediate time; Step 2.2: Select different motion models, determine the state transfer matrix, and obtain the target prediction state of the two tracks at the intermediate moment.

4. The method for determining interrupted track interconnection based on multi-dimensional feature information according to claim 3, characterized in that: The step 2.1 is as follows: Based on the filtering state of the first and last points of the new and old tracks, that is, Update status of old tracks at all times and Update status of new tracks at all times , towards the middle moment Predict the target state of the new and old tracks, that is, ， ， Where: and is the state transfer matrix, , They respectively represent the target prediction states of the old track and the new track at the intermediate moments.

5. The method for determining interrupted track interconnection based on multi-dimensional feature information according to claim 3, characterized in that: The step 2.2 is specifically as follows: The target moves in three-dimensional space, and the state of the target is predicted using the state transfer matrices of three motion modes: uniform straight line, uniformly accelerated straight line, and uniform turning. For uniform linear and uniformly accelerated linear motion modes, since there is no coupling relationship between different vector coordinates, the state transfer matrix of the target used at this time is It consists of three one-dimensional motion state transfer matrices, namely , Where: is a one-dimensional motion state transfer matrix. When uniform straight line prediction is performed , when performing uniform acceleration straight line prediction ; ， ， For the uniform turning motion mode, due to the coupling relationship between different vector coordinates, the state transfer matrix of the target used at this time is It consists of a two-dimensional uniform turning motion state transfer matrix and a one-dimensional uniform linear motion state transfer matrix, that is, , Where: and They represent the state transfer matrix corresponding to the target making a uniform turning motion in the xy plane and the state transfer matrix of the target making a uniform motion on the z axis respectively; ， ； According to different motion models, the corresponding state transfer matrix is ​​selected, and the state transfer matrix at the intermediate moment between the new and old tracks is obtained respectively, so as to obtain the target predicted state of the two tracks at the intermediate moment.

6. The method for determining interrupted track interconnection based on multi-dimensional feature information according to claim 3, characterized in that: The step 3 specifically comprises the following steps: Step 3.1, calculate the spatial distance between the two predicted positions: Step 3.1.1: For the three-dimensional uncoupled uniform motion model CV, the spatial distance between the two predicted positions is for: ; Step 3.1.2: For the three-dimensional uncoupled uniform acceleration motion model CA, the spatial distance between the two predicted positions is for: ; Step 3.1.3: Based on the heading angles of the first and last points of the new and old tracks and , calculate the mean turning rate , respectively according to and , the spatial distance between the two predicted positions for: ; in: ; Step 3.1.4: Based on the average turning rate at the end point of the old track and its previous time t , based on the average turning rate at the first point of the new track and its subsequent time t , respectively according to and , the spatial distance between the two predicted positions for: ; in: , , Step 3.2: The maximum spatial distance between the new and old tracks is defined as a unified measure of spatial consistency between two tracks, and its membership function is defined as: in: , is the minimum spatial distance, that is: , The membership function is .

7. The method for determining interrupted track interconnection based on multi-dimensional feature information according to claim 3, characterized in that: The step 4 is specifically as follows: The maximum AD amplitude in the track is selected as its standard value, and the corresponding membership function is defined as , , Where: and is the AD amplitude value corresponding to the new and old tracks, The statistical variance of the AD amplitude of the same track is obtained by statistically analyzing historical data or taking the maximum AD amplitude value between the new and old tracks.

8. The method for determining interrupted track interconnection based on multi-dimensional feature information according to claim 3, characterized in that: The specific step 5 is as follows: the association decision value after integrating the membership function of the spatial consistency in step 3 and the AD amplitude consistency in step 4 with their corresponding weights As a basis for judgment: ,in is the weight.

9. The method for determining interrupted track interconnection based on multi-dimensional feature information according to claim 3, characterized in that: The step 6 specifically comprises the following steps: Step 6.1: Construct the interconnection relationship matrix and the association decision matrix describing the correspondence between the new and old tracks: definition The association decision matrix To describe the associated decision value of the new and old tracks, , Where: Indicates Old track and The association decision value of the new track, m and n represent the number of old tracks and the number of new tracks involved in the association decision respectively; definition The interconnection matrix To describe the correspondence between the new and old tracks, , Indicates the judgment Old track and The new tracks are not associated, Indicates the judgment Old track and New track associations; Since there is a non-one-to-one correspondence between the new and old tracks, the interconnection matrix satisfies , ; Step 6.2, solve the interconnection relationship matrix of step 6.

1.

10. The method for determining interrupted track interconnection based on multi-dimensional feature information according to claim 9, characterized in that: The step 6.2 specifically includes the following steps: Step 6.2.1: Initial judgment of timing: based on the end time of the old track Less than the start time of the new track Constraint relationship, set the matrix elements that do not meet this condition in step 4.1 to zero, that is: , The interruption time is greater than a certain threshold The track of is no longer associated with judgment, that is ; Step 6.2.2: Preliminary judgment of motion relationship: At maximum speed , maximum acceleration , minimum speed and maximum turning rate The non-zero elements in the interconnection relationship matrix are screened and updated, and the corresponding element values ​​of the track pairs that meet the following conditions are set to 1, and the rest are set to 0: Calculate the Euclidean distance between the end point of the old track and the start point of the new track in the track pair , and the speed constraint is imposed on the track pair according to the speed obtained by dividing the distance by the interruption time: ， ， According to the speed at the end of the old track and the speed at the start of the new track Divide the acceleration of the track pair by the interruption time to limit the acceleration of the track pair: , According to the previously calculated mean turning rate To limit the turning rate of the track pair: , Among them, the maximum speed Selected based on historical track data of air and surface targets; maximum acceleration Set based on air and surface target characteristics or historical data; minimum speed Set based on historical data of air and surface targets; Maximum turning rate of surface targets Set according to the target characteristics of the surface target; the maximum turning rate of the air target It is set according to the characteristics of the air target or according to the maximum turning rate of the non-interrupted track multiplied by the magnification factor. , is the magnification factor, is the maximum turning rate without interruption of the track, It is obtained by traversing the track segment; Step 6.2.3, global optimal decision: Select the interconnection matrix The element with the largest association decision value corresponding to all the remaining non-zero elements in the , determines that the corresponding new and old tracks are successfully associated, and then the row and column elements of the element are All elements of the interconnection relationship matrix are set to zero, and the judgment is iterated in sequence, and all elements are set to zero, so that the judgment results of the association between all new and old tracks are obtained.

Images