Passive target information autonomous fusion method based on towed array

Abstract

The invention provides a passive target information autonomous fusion method based on a towed array, and the method takes a multi-target tracking theory and a data fusion method as a core, employs the similarity of time-space-frequency domain features of a target, proposes a local track correlation and fusion processing flow, and achieves the autonomous fusion of passive target information. According to the method, local tracks are clustered to form a current frame local track cluster, and then a fusion target track cluster is formed, so that autonomous association and fusion of feature information of underwater target radiation noise are realized, the number of false alarms is reduced, and an underwater target passive feature information set is established. Through autonomous association and fusion of underwater target feature information, the false alarm problem is solved, the false alarm interference number is greatly reduced, target state feature estimation is formed, and powerful technical support is provided for underwater passive weak target detection and identification based on a towed array.

Description

Technical Field

[0001] The present invention relates to the field of passive sonar target detection, and mainly relates to an autonomous fusion method for passive target information based on a towed array. Background Art

[0002] Underwater target passive detection technology has always been an important research direction in the field of sonar target detection. This technology mainly uses signal processing and other methods to detect the underwater target radiation noise signal from environmental noise, and then realizes underwater target detection. Since there is no need for the sonar to actively transmit detection sound signals, it has good concealment and is widely used on platforms such as surface ships and submarines.

[0003] However, with the continuous development of underwater target acoustic stealth technology, the intensity of underwater target radiation noise signals has been continuously decreasing. At the same time, the increasing human ocean activities have led to a continuous reduction in the detection range of passive sonar and a continuous aggravation of the false alarm interference problem. For this reason, a class of passive detection technologies mainly based on data fusion methods has become one of the hot directions.

[0004] In the patent "A Passive Sonar Non-Cooperative Target Line Spectrum Information Fusion Method" (patent number: CN201910451625.2), for the target line spectrum characteristics, a data association fusion method is used to fuse the line spectrum characteristics, which improves the passive detection ability of underwater non-cooperative targets. However, when performing association fusion in this patent, it is mainly to first fuse two associated local tracks, obtain the bearing angle estimate, and then use this bearing angle estimate to associate and fuse with other local tracks. Through continuous association and bearing angle estimate update, the final association result is obtained. When a local track first associates with a false alarm with a large error, the bearing angle estimate generated by this method may deviate, and in severe cases, it may affect the correctness of the subsequent association results. Summary of the Invention

[0005] The purpose of the present invention is to overcome the deficiencies existing in the prior art and provide an autonomous fusion method for passive target information based on a towed array.

[0006] The purpose of the present invention is achieved through the following technical solutions. An autonomous fusion method for passive target information based on a towed array includes the following steps:

[0007] Step 1, passive target tracking: Filter the bearing angle-frequency characteristic information in the underwater target radiation noise signal through a multi-target tracking filter to form a spatio-temporal-frequency domain track of a suspected target, that is, a local track;

[0008] Step 2, Similarity evaluation of local tracks: For the local tracks formed by the target tracking filter in the current frame, in a pairwise combination manner, any two local tracks are combined into a local track pair, and the similarity of the local track pair is evaluated using the association algorithm;

[0009] Step 3, Generation of local track clusters in the current frame: Using the similarity evaluation results of local track pairs, cluster the local tracks that may originate from the same underwater target to form local track clusters in the current frame;

[0010] Step 4, Association between local track clusters in the current frame and fused target track clusters: Perform data association between the local track clusters in the current frame and the fused target track clusters to support the establishment and update of the fused target track clusters;

[0011] Step 5, Fusion of local track clusters in the current frame and fused target track clusters: According to the association results calculated in Step 4, perform data fusion of the local track clusters in the current frame and the fused target track clusters to update the fused target track clusters;

[0012] Step 6, Evaluation and pruning of fused target track clusters: Based on the association between local tracks and fused target track clusters in multiple frames, determine the fused target track numbers to which the local tracks belong, evaluate the fused target track clusters, output the fused target track clusters suspected of being underwater targets, and prune the fused target track clusters that have not been determined as suspected underwater targets for a long time.

[0013] The beneficial effects of the present invention are as follows: In the process of passive underwater target detection based on a towed array, two challenges are usually faced. On the one hand, there is the false alarm problem. Due to factors such as the ocean environment and human ocean activities, there are usually many false alarms in the process of underwater passive target detection, seriously interfering with underwater target detection. On the other hand, there is the problem of target state feature estimation. For targets with unknown time-space-frequency domain feature information, due to less prior information, the signal intensities at different frequencies are not only significantly different, but there are also large differences in the accuracy of aspect angle estimation, resulting in difficulties in the autonomous association and fusion of underwater target state features, seriously affecting the underwater target discrimination ability. The present invention takes multi-target tracking theory and data fusion and other methods as the core, uses the similarity of target time-space-frequency domain features to achieve the autonomous association and fusion of the feature information of underwater target radiated noise, reduces the number of false alarms, establishes an underwater target passive feature information set, and improves the detection and discrimination ability of underwater targets. Description of the Drawings

[0014] To more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the accompanying drawings required in the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0015] Figure 1 It is a processing flow chart of a passive target information autonomous fusion method based on a towed array.

[0016] Figure 2 It is a local tracking filter diagram.

[0017] Figure 3 It is a diagram of the fusion processing result of the present invention.

[0018] Figure 4 It is a comparison diagram of the fusion tracking result and the true value. Detailed implementation manners

[0019] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the protection scope of the present invention.

[0020] The present invention is mainly applied to the field of passive sonar target detection, especially for the fusion method of passive detection target information based on a towed line array. This method includes the following steps:

[0021] Step 1: Passive target tracking;

[0022] Generally, after preprocessing such as spatio-temporal frequency domain filtering of the underwater target radiation noise signal, the characteristics such as the aspect angle and frequency corresponding to the line spectrum in the radiation signal can be obtained, which are important characteristic information for passive underwater target detection. For the aspect angle-frequency characteristic information in the underwater target radiation noise signal, a multi-target tracking filter can be used for filtering processing, which can not only reduce the number of false alarm interferences, but also form the spatio-temporal frequency domain tracks of suspected targets (referred to as local tracks in the present invention). Therefore, in the present invention, a Generalized Labeled Multi-Bernoulli (GLMB) filter is mainly used for tracking filtering, and the specific methods and steps can refer to relevant literatures, such as "The labeled multi-Bernoulli filter" published by Vo B-N, etc.;

[0023] Step 2: Similarity evaluation of local tracks;

[0024] For the local tracks formed by the current-frame target tracking filter, any two local tracks can be combined pairwise to form a local track pair, and an association algorithm can be used to evaluate the similarity of the local track pair; by calculating the pairwise association results for all local tracks and then using the track numbers to quickly complete the association, the local tracks suspected to be of the same target can be clustered into a single cluster, and a relatively unbiased estimate of the aspect angle can be obtained based on the local tracks in the cluster.

[0025] Step 2.1: Assume that the current is the k-th frame, and the multi-target tracking filter has formed m local tracks, and the local tracks are all independently and identically distributed. Then these local tracks can form the local track set Z k ={z 1,k ,z 2,k ,…,z m,k} at the k-th frame, where z = [id; t; θ; f; σ 2 ; fid] T is the state of the local track formed by the multi-target tracking filter. Here, id is the track number (unique), θ is the aspect angle estimate in the local track, f is the frequency estimate in the local track, σ 2 is the aspect angle error variance in the local track, and fid is the batch number of the fused target track cluster, which is used to indicate the fused target track cluster to which the local track stably belongs;

[0026] In addition, a set FID is established to record the association relationship between the local track and the fused target track cluster. Here, FID = {…, [k; fcid k ,…}, and fcid k is the batch number of the fused target track cluster most relevant to the local track at the k-th frame. If more than half of the fcid values of the local track are the same within the recent N frames, it means that the local track stably belongs to the fused target track cluster with the batch number fcid, and fid = fcid;

[0027] Step 2.2: Then, for any two different local tracks in the local track set, a local track pair (z i,k , z j,k ) can be formed, where

[0028]

[0029] Step 2.3: For a local track pair (z i,k , z j,k ), the Mahalanobis distance can be used to evaluate the degree of association between the two local tracks in the track pair. If the Mahalanobis distance d m is smaller, it means that the two tracks in the track pair are more likely to originate from the same underwater target.

[0030]

[0031] Here, Σ is the covariance matrix used to describe the difference between two local tracks.

[0032] Step 3: Generation of the current-frame local track cluster;

[0033] Using the similarity evaluation results of local tracks, cluster the local tracks that may originate from the same underwater target to form the current-frame local track cluster;

[0034] Step 3.1: According to the track numbers in the local track pairs, calculate the association relationship between two local tracks within the most recent N frames, that is, use (2) to calculate the d m,k-N , d m,k-N+1 , …, d m,k values;

[0035] Step 3.2: For the d m,k-N , d m,k-N+1 , …, d m,k in the local track pairs, if the number of conditions satisfied does not exceed , it is considered that the association degree between the two local tracks in this local track pair is relatively low or cannot be discriminated, and this group of local track pairs is discarded;

[0036] Step 3.3: For the remaining local track pairs, calculate the mean of d m,k-N , d m,k-N+1 , …, d m,k and use it as the association degree of this local track pair, that is,

[0037] Association degree = mean(d m,k-N , d m,k-N+1 , …, d m,k ) (3) Here, is the association threshold, and mean(·) is the mean function;

[0038] Step 3.5: Since the calculation of the association degree between any local track and other tracks has been completed, the current-frame local track cluster can be quickly formed according to the track numbers and the association degree values in the local track pairs;

[0039] Step 3.5.1: Use the association degree values calculated in (3) to sort all local track pairs in ascending order;

[0040] Step 3.5.2: Determine in sequence whether the track numbers in the local track pairs can form a closed loop. If a closed loop can be formed, it is considered that these local tracks are pairwise related, that is, it is considered that these local tracks originate from the same underwater target, and a local track cluster of the current frame is formed. The local track information related to the track number (time, local track number, beam angle, frequency, error variance of the beam angle, etc.) is packed in one cluster. Among them, each local track pair can only appear in one local track cluster of the current frame;

[0041] For example, if there are 10 local tracks (track numbers 1 - 10) in the current frame and 10 groups of local track pairs are formed, namely {(1,2), (2,3), (3,1), (3,4), (4,1), (4,2), (6,7), (6,8), (7,8), (9,10)}, then three local track clusters of the current frame that can be formed are {1,2,3,4}, {6,7,8}, and {9,10};

[0042] Step 3.5.3: For a local track that fails to form an association relationship with any other local track, judge the stability of the tracking of this local track, that is, judge whether the tracking of this local track is stable by the tracking time and the detection situation of the track. If within the recent N frames, this local track can be detected and updated by the target tracking filter more than NT times, it is considered stable tracking, where NT ≤ N;

[0043] For example, in the example of Step 3.5.2, if the local track with track number 5 can be stably detected and tracked, it forms an independent local track cluster of the current frame, that is, in the above example, the local track clusters of the current frame that can be formed are {1,2,3,4}, {6,7,8}, {9,10}, and {5};

[0044] Step 4: Associate the local track clusters of the current frame with the fused target track clusters;

[0045] Perform data association on the local track clusters of the current frame and the fused target track clusters to support the establishment and update of the fused target track clusters;

[0046] Step 4.1: For the local track cluster of the k - th frame of the current frame, according to the beam angle θ of the n local tracks in the local track cluster of the current frame k , estimate the fused beam angle of the local track cluster of the current frame using the weighted fusion method Beam angle error variance and the angle transformation rate

[0047]

[0048]

[0049]

[0050] Among them,

[0051]

[0052]

[0053] here, w is the weight value, σ is the standard deviation of the beam angle error, and t is the time;

[0054] Step 4.2: For a surviving fused target track cluster (usually including information such as batch number, time, beam angle, beam angle error variance, beam angle change rate, local track number, frequency, etc.), the fused target beam angle θ of the (k - 1)-th frame can be used fc,k-1 and the angle change rate to judge the association relationship between the local track cluster of the current frame and the surviving fused target track cluster;

[0055]

[0056]

[0057]

[0058]

[0059] Here, and △t k|k-1 are respectively the predicted angle change rate, predicted beam angle, and time difference of the fused target track cluster from the (k - 1)-th frame to the k-th frame.

[0060] If △θ ≤ T θ and △r ≤ T r are satisfied, it is considered that the local track cluster of the current frame is related to the surviving fused target track cluster; among them, T θ and T r are respectively the beam angle association threshold and the angle change rate threshold, and ||·|| is the absolute value function;

[0061] Step 5: Fusion of the local track cluster of the current frame and the fused target track cluster

[0062] According to the association result calculated in Step 4, perform data fusion of the local track cluster of the current frame and the fused target track cluster to update the fused target track cluster;

[0063] Step 5.1: For the current frame's local track cluster, if there is no associated fused target track cluster, it is considered that the current frame's local track cluster may come from a new underwater target. Therefore, a new fused target track cluster is established, and at the same time, the feature information in the current frame's local track cluster (such as time, beam angle, variance of beam angle error, related local tracks, etc.) is recorded and used as the feature of this fused target track cluster;

[0064] fcid = M fc +1 (13)

[0065] t fc,k = t lc,k (14)

[0066]

[0067]

[0068]

[0069]

[0070] Here, fcid is the batch number of the fused target track cluster, and M fc is the total number of fused target track clusters;

[0071] Step 5.2: For the current frame's track cluster, if there is an associated fused target track cluster and it is the only associated fused target track cluster within the association threshold, directly update the feature data in the fused target track cluster with the feature data in the current frame's local track cluster, such as time, beam angle, variance of beam angle error, related local tracks, etc. The formula is the same as that from (13) to (18);

[0072] Step 5.3: For the current frame's local track cluster, if there is an associated fused target track cluster and there are multiple associated fused target track clusters within the association threshold, first perform an optimal match between the local track information in the local track cluster and the local track information in the associated fused target track clusters, and then perform state fusion update on the associated fused target track clusters according to the matching results;

[0073] Step 5.3.1: For the local tracks in the current frame's local track cluster, it can be determined whether the fcid in this local track is consistent with the fcid of the associated fused target track cluster. If they are consistent, it means that this local track already has a subordinated fused target track cluster, that is, the subordination relationship of this local track in this frame is determined; if they are inconsistent, the association relationship between the current local track and the associated fused target track cluster is calculated using formula (2) to determine the subordination relationship of this local track in this frame;

[0074] Step 5.3.2: For the fused target track cluster, fuse the local tracks of all current frames belonging to this fused target track cluster using (4) to (8), and use the result as the feature of this fused target track cluster.

[0075] Step 6: Evaluation and pruning of fused target track clusters

[0076] Based on the association between the local tracks of multiple frames and the fused target track clusters, determine the fused target track number to which the local track belongs, evaluate the fused target track clusters, output the fused target track clusters suspected of being underwater targets, and prune the fused target track clusters that have not been determined as suspected underwater targets for a long time.

[0077] Step 6.1: For any fused target track cluster, the evaluation of the fused target track cluster can be realized by analyzing the local track information of the fused target track cluster. That is, if there is at least one local track stably belonging to the fused target track cluster, then the fused target track cluster is a suspected underwater target, and the beam angle of the k-th frame of the underwater target is The frequency feature it has is Freq; Freq is the frequency set of all local tracks stably belonging to the fused target track cluster.

[0078] Step 6.2: Realize the pruning process of the fused target track cluster according to the local track feature information of the fused target track cluster.

[0079] Step 6.2.1: For a fused target track cluster that has not been determined as a suspected underwater target, if there is at least one local track stably belonging to the fused target track cluster in its recent N frames, it means that this fused target track cluster is a surviving fused target track cluster, and the features it describes may originate from a suspected underwater target.

[0080] Step 6.2.2: For a fused target track cluster that has not been determined as a suspected underwater target, if there is no local track stably belonging to the fused target track cluster in its recent N frames, it means that this fused target track cluster is suspected of being a false alarm, and pruning processing is performed on this track cluster, that is, deleting this fused target track cluster.

[0081] Step 6.2.3: For a fused target track cluster that has been determined as a suspected underwater target, if there is no local track stably belonging to the fused target track cluster in its recent N frames, it means that the suspected underwater target has been lost in tracking, and pruning processing is performed on this track cluster, that is, determining that this fused target track cluster has died and will no longer participate in state update.

[0082] As shown in the Figure 1 attachment, the processing flow of the present invention is within the dashed box in the figure;

[0083] As shown in theFigure 2 As shown, each line segment in the figure is a local track obtained by processing the passive target detection information of the towed array using the GLMB tracking filter during a certain sea trial. The total number of such line segments is 892. It can be seen that there are a large number of false alarms in the passive target detection information of the towed array, which increases the difficulty of extracting target feature information and identifying underwater targets.

[0084] As shown in the appendix Figure 3 is the processing result diagram of this method. Each line segment in the figure is an output fused target track cluster (a total of 32). The numbers beside the line segments are the batch numbers of their fused target tracks. By comparing appendix Figure 2 and appendix Figure 3 , it can be seen that a large number of false alarms are filtered out, and the number of tracks is reduced by 96%. The feature information of suspected same underwater targets is packed in the same fused target track cluster. The comparison between the fused target track and the real target track of this sea trial is shown in appendix Figure 4 . In the figure, the solid line is the target track of fused tracking, and the dashed line is the underwater target track recorded. It can be seen the effectiveness of the present invention and the accuracy of state estimation in the sea trial environment.

[0085] The present invention provides a passive target information autonomous fusion method based on a towed array. This method takes multi-target tracking theory and data fusion methods as the core, and uses the similarity of target time-space-frequency domain features to propose a processing flow for the association and fusion of local tracks, that is, local tracks form local track clusters of the current frame from clustering, and then form fused target track clusters, thereby realizing the autonomous association and fusion of the feature information of underwater target radiated noise, reducing the number of false alarms, and establishing a passive feature information set of underwater targets. Through the autonomous association and fusion of underwater target feature information, not only the false alarm problem is solved, the number of false alarm interferences is greatly reduced, but also the target state feature estimation is formed, providing strong technical support for the detection and identification of passive weak targets based on towed arrays underwater.

[0086] In addition, the present invention has the advantages of simple theory, easy engineering implementation, stability and reliability, etc., and the effectiveness of the present invention has been verified by sea trials, meeting the requirements of actual marine application scenarios.

[0087] Explanation of relevant technical terms

[0088] Towed array: In the present invention, the towed array refers to a towed line array sonar, which is a sonar that towes a receiving line array at the stern of a ship.

[0089] As described above, it is only the specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the scope disclosed by the present invention should be covered within the protection scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims described above.

Claims

1. A passive target information autonomous fusion method based on a towed array, characterized in that: It includes the following steps: Step 1, passive target tracking: Filter the bearing-frequency characteristic information in the underwater target radiated noise signal through a multi-target tracking filter to form the spatio-temporal-frequency domain track of the suspected target, that is, the local track; Step 2, similarity evaluation of local tracks: For the local tracks formed by the target tracking filter in the current frame, in a pairwise combination manner, any two local tracks are combined into a local track pair, and the similarity of the local track pair is evaluated using an association algorithm; Step 3, generation of the local track cluster in the current frame: Utilize the similarity evaluation results of the local track pairs to cluster the local tracks that may originate from the same underwater target to form the local track cluster in the current frame; Step 4, association between the local track cluster in the current frame and the fused target track cluster: Perform data association between the local track cluster in the current frame and the fused target track cluster to provide support for the establishment and update of the fused target track cluster; Step 5, fusion of the local track cluster in the current frame and the fused target track cluster: According to the association results calculated in Step 4, perform data fusion of the local track cluster in the current frame and the fused target track cluster to realize the update of the fused target track cluster; Step 6, evaluation and pruning of the fused target track cluster: According to the association situation between the local tracks in multiple frames and the fused target track cluster, determine the fused target track number to which the local track belongs, evaluate the fused target track cluster, output the fused target track cluster of the suspected underwater target, and prune the fused target track cluster that has not been determined as the suspected underwater target for a long time.

2. The autonomous fusion method of passive target information based on a towed array according to claim 1, characterized in that: The specific process of Step 2 is as follows: Step 2.1: Assume that the current is the k-th frame. The multi-target tracking filter forms m local tracks, and these local tracks are independent and identically distributed. Then these local tracks constitute the local track set Z of the k-th frame k ={z 1,k ,z 2,k ,…,z m,k}, where z = [id;t;θ;f;σ 2 ;fid] T is the state of the local track formed by the multi-target tracking filter. id is the track number, θ is the estimated beam angle in the local track, f is the estimated frequency in the local track, σ 2 is the variance of the beam angle error in the local track, and fid is the batch number of the fused target track cluster, which is used to indicate the fused target track cluster to which the local track stably belongs; establish a set FID for recording the association relationship between the local track and the fused target track cluster, FID = {…,[k;fcid k ,…}, where fcid k is the batch number of the fused target track cluster most relevant to the local track in the k-th frame. If more than half of the fcid values of the local track are the same in the recent N frames, it means that the local track stably belongs to the fused target track cluster with the batch number fcid, and fid = fcid; Step 2.2: Then, any different local tracks in this set of local tracks can form a set of local track pairs (z i,k , z j,k ), where Step 2.3: For a set of local track pairs (z i,k , z j,k ), use the Mahalanobis distance to evaluate the correlation degree between the two local tracks in the track pair. If the Mahalanobis distance d m is smaller, it means that the two tracks in the track pair are more likely to originate from the same underwater target; Σ is the covariance matrix used to describe the difference between two local tracks.

3. The autonomous fusion method of passive target information based on a towed array according to claim 2, characterized in that: The specific process of Step 3 is as follows: Step 3.1: Calculate the association relationship between two local tracks within the most recent N frames according to the track numbers in the local track pair, that is, calculate the d of the two local tracks in the most recent N frames using formula (2) m,k-N , d m,k-N+1 , …, d m,k value; Step 3.2: For d in the local track pair m,k-N , d m,k-N+1 , …, d m,k , if the number of conditions that meet is no more than pieces, it is considered that the correlation degree between the two local tracks in this local track pair is low or cannot be judged, and this group of local track pairs is discarded; Step 3.3: For the remaining local track pairs, calculate the mean value of d m,k-N , d m,k-N+1 , …, d m,k , and use it as the correlation degree of this local track pair, that is, Degree of association = mean(d m,k-N , d m,k-N+1 , …, d m,k ) (3) Here, is the correlation threshold, and mean(·) is the mean function; Step 3.5: After calculating the degree of association with other tracks for any local track, according to the track numbers and the degree of association values in the local track pairs, quickly cluster to form the local track cluster in the current frame.

4. The autonomous fusion method of passive target information based on a towed array according to claim 3, wherein: The specific process of Step 3.5 is as follows: Step 3.5.1: Sort all the local track pairs in ascending order according to the degree of association values calculated using formula (3); Step 3.5.2: Sequentially determine whether the track numbers in the local track pairs form a closed loop. If a closed loop is formed, it is considered that these local tracks are pairwise related, that is, it is considered that these local tracks originate from the same underwater target, and form the local track cluster in the current frame, and pack the local track information with related track numbers in one cluster; among them, each local track pair can only appear in one local track cluster in the current frame; Step 3.5.3: For the local track that fails to form an association relationship with any other local track, judge the tracking stability of this local track, that is, judge whether the local track tracking is stable by the tracking time and the detection situation of the track; if within the recent N frames, this local track is detected and updated more than NT times by the target tracking filter, it is considered stable tracking, where NT ≤ N.

5. The autonomous fusion method of passive target information based on a towed array according to claim 4, wherein: The specific process of Step 4 is as follows: Step 4.1: For the current frame local track cluster of the k-th frame, according to the bow angles θ of the n local tracks in the current frame local track cluster k , estimate the fused bow angle of the current frame local track cluster using the weighted fusion method Bow angle error variance and angle transformation rate Among them, Among them, w is the weight value, σ is the standard deviation of the bearing error, and t is the time; Step 4.2: For a surviving fused target track cluster, use the fused target beam angle θ of the (k-1)-th frame fc,k-1 and the rate of change of the angle to determine the association relationship between the local track cluster of the current frame and the surviving fused target track cluster; Among them, and △t k|k-1 are respectively the predicted angular rate of change, the predicted relative bearing, and the time difference of the fused target track cluster from the (k - 1)-th frame to the k-th frame; If △θ ≤ T θ and △r ≤ T r are satisfied, it is considered that the current frame local track cluster is related to the surviving fused target track cluster; where T θ and T r are the bearing angle correlation threshold and the angle change rate threshold respectively, and ||·|| is the absolute value function.

6. The autonomous fusion method of passive target information based on a towed array according to claim 5, wherein: The specific process of Step 5 is as follows: Step 5.1: For the current-frame local track cluster, if there is no associated fused target track cluster, it is considered that the current-frame local track cluster may come from a new underwater target. A new fused target track cluster is established, and the feature information in the current-frame local track cluster is recorded and used as the feature of this fused target track cluster; fcid = M fc +1 (13) t fc,k = t lc,k (14) Among them, fcid is the batch number of the fused target track cluster, and M fc is the total number of the fused target track clusters; Step 5.2: For the current-frame track cluster, if there is an associated fused target track cluster and it is the only associated fused target track cluster within the association threshold, the feature data in the current-frame local track cluster is directly used to update the feature data in the fused target track cluster, and the formula is the same as (13) to (18); Step 5.3: For the current-frame local track cluster, if there is an associated fused target track cluster and there are multiple associated fused target track clusters within the association threshold, first perform an optimal match between the local track information in the local track cluster and the local track information in the associated fused target track cluster, and then perform state fusion update on the associated fused target track cluster according to the matching result; Step 5.3.1: For the local track in the current-frame local track cluster, by determining whether the fcid in the local track is consistent with the fcid of the associated fused target track cluster. If it is consistent, it means that the local track already belongs to an associated fused target track cluster, that is, the affiliation of the local track in this frame is determined; if it is not consistent, the association relationship between the current local track and the associated fused target track cluster is calculated using formula (2) to determine the affiliation of the local track in this frame; Step 5.3.2: For the fused target track cluster, the local tracks that belong to this fused target track cluster in the current frame are fused using formulas (4) to (8) and used as the feature of this fused target track cluster.

7. The autonomous fusion method of passive target information based on a towed array according to claim 6, wherein: The specific process of step 6 is as follows: Step 6.1: For any fused target track cluster, evaluate the fused target track cluster by analyzing the local track information of the fused target track cluster. That is, if at least one local track stably belongs to the fused target track cluster, then the fused target track cluster is a suspected underwater target, and the beam angle of the underwater target at the k-th frame is with a frequency characteristic of Freq; Freq is the set of frequencies of all local tracks that stably belong to the fused target track cluster; Step 6.2: Perform pruning processing on the fused target track cluster according to the local track feature information of the fused target track cluster; Step 6.2.1: For a fused target track cluster that has not been determined as a suspected underwater target, if there is at least one local track that stably belongs to this fused target track cluster in its recent N frames, it means that this fused target track cluster is a surviving fused target track cluster, and the features it describes may come from a suspected underwater target; Step 6.2.2: For a fused target track cluster that has not been determined as a suspected underwater target, if there is no local track that stably belongs to this fused target track cluster in its recent N frames, it means that this fused target track cluster is suspected of being a false alarm, and pruning processing is performed on this track cluster, that is, this fused target track cluster is deleted; Step 6.2.3: For a fused target track cluster that has been determined as a suspected underwater target, if there is no local track that stably belongs to this fused target track cluster in its recent N frames, it means that this suspected underwater target has been lost in tracking, and pruning processing is performed on this track cluster, that is, it is determined that this fused target track cluster has died and will no longer participate in state update.

Images