A multi-source information fusion method based on acoustic data

Abstract

The present invention discloses a multi-source information fusion method based on acoustic data, which is applicable to the application scenario of underwater target passive detection. In the underwater full-passive acoustic detection scenario, real-time underwater acoustic detection data obtained by a multi-source heterogeneous acoustic detection array is utilized to obtain the data distribution characteristics, a specific underwater acoustic data fusion method is studied, a multi-array data fusion platform is built, and underwater targets are fused. By means of methods such as detection, association, correlation, and combination, the target information of various acoustic detection types such as passive sonar array detection data, passive vector array detection data, and matched field processing data is comprehensively processed, so as to obtain a more accurate estimation of the state and identity of underwater targets, obtain more reliable and accurate target information within a lower time complexity, improve the usage efficiency of detection equipment, and lay a technical foundation for expanding the application scope of related equipment; the highlight of this method is that the process is scalable and suitable for fusing three types of acoustic detection data in multiple dimensions.

Description

Technical Field

[0001] The present invention belongs to the technical field of information fusion, and mainly aims at the underwater target detection data fusion under the background of acoustic joint detection. Specifically, it relates to a multi-source information fusion method based on acoustic data. Background Art

[0002] According to the different ways of obtaining target detection information by sonar, underwater target detection data is divided into active sonar detection data and passive sonar detection data. Active sonar detection data can include relatively accurate target detection information such as azimuth angle and target distance. However, the data acquisition process is an active detection, which has characteristics such as large energy consumption and easy exposure of one's own position.

[0003] The positioning effect of the passive sonar detection scheme on underwater targets is poor. And with the rapid development of modern technologies such as "acoustic stealth", the radiated noise signal of the target is increasingly low. At the same time, due to the influence of human activities such as the development of marine resources, the underwater environmental background noise is increasing day by day, resulting in low accuracy of passive sonar detection data generated by traditional single sonar platforms, high time complexity of data processing methods, and large measurement errors.

[0004] To solve the problem of low target tracking accuracy of passive sonar arrays, the patent document "A pure azimuth underwater target tracking algorithm for multi-level information fusion, publication number CN107192995B" realizes underwater positioning and tracking of targets by using a data fusion method with multiple detection devices carried on a UUV cooperative system, achieving high-precision underwater targets in a large range of space. However, this method only focuses on the detection data fusion method of a single sonar platform, and has a high complexity due to involving partial derivative operations. The patent document "A multi-passive sonar non-cooperative target line spectrum information fusion method, acceptance number CN20190469360.9" provides a multi-passive sonar multi-target information fusion method, which analyzes the target time-series characteristics from the perspective of spectral information to realize target tracking, but does not reflect the data heterogeneous information of multi-passive sonar detection arrays. The patent document "A sonar integrated target recognition method for multi-source information fusion, publication number CN110488301A" combines AIS information, radar information and sonar detection information for joint target recognition. However, the use of AIS and radar data is to distinguish whether the target is an underwater target, and does not participate in the actual fusion process of underwater sonar targets. Summary of the Invention

[0005] In order to solve the deficiencies and defects in the above-mentioned prior art, the applicant has developed and designed a method for real-time analysis and data fusion processing of the acquired detection data in a passive acoustic integrated detection array, and finally forms a new method for fusion target trajectory information, which is suitable for underwater target detection and tracking in scenarios containing sonar linear array detection data, vector array detection data, and matching field processing data. It aims to study the characteristics of the output data of acoustic detection devices with different characteristics in the acoustic sensor network, and explore how to integrate, analyze and fuse the output data of multiple acoustic detection devices, so that the final fusion detection efficiency is better than the detection efficiency of a single detection device.

[0006] The method provided by the present invention is: a multi-source information fusion method based on acoustic data, which is used to fuse three types of acoustic detection equipment output data: an even-dimensional passive sonar detection data set, an even-dimensional vector detection data set, and a one-dimensional matching field data set; it includes the following steps S:

[0007] Step S1, real-time acquisition of three types of acoustic detection data and sampling intervals, and storage of detection information in a local server after synchronization of timestamps; calculating the minimum sampling interval of multi-source data, and when the sampling data meets the set threshold value, interpolating the multi-source detection data whose sampling interval is greater than the above minimum sampling interval;

[0008] Step S2: performing initial information preprocessing on multiple types of acoustic detection information: sonar linear array detection data, vector array detection data, and matching field detection data, and removing null values ​​and abnormal values;

[0009] Step S3: performing azimuth positioning and candidate target screening on each type of data in step S2 frame by frame to obtain a set of candidate target positions;

[0010] Step S4: Fusing the two types of acoustic detection information candidate target position sets obtained in step S3 with the target orientation set after matching field processing, and forming associated point targets frame by frame;

[0011] Step S5: Associating the tracks of the fused point targets in the time series dimension and forming a fused track of the underwater target;

[0012] Step S6: Repeat steps S2 to S5 to finally obtain the detected target track after association fusion.

[0013] Introduction to the working principle and beneficial effects of the present invention: In order to improve the accuracy of underwater target detection and reduce the time complexity of the algorithm, this method is applicable to the application scenario of underwater target passive detection. In the underwater full-passive acoustic detection scenario, real-time underwater acoustic detection data obtained by a multi-source heterogeneous acoustic detection array is used to obtain the data distribution characteristics, study specific underwater acoustic data fusion methods, build a multi-array data fusion platform, and finally provide a new solution for actual engineering applications to improve the efficiency of underwater acoustic detection and target tracking. The use of this method makes the system have the characteristics of complex data input types, large data volume, and hierarchical information. Based on a hierarchical model, multi-source information is comprehensively processed, and each processing level reflects different degrees of abstraction of the original information, from the evaluation of states and attributes at a lower level to the estimation of the situation at a higher level. It can comprehensively utilize the different characteristics of multiple sensors, use spatio-temporal information, and obtain different attribute information of the target in all directions to improve the detection performance. The present invention successfully comprehensively processes target information of various acoustic detection types such as passive sonar array detection data, passive vector array detection data, and matched field processing data through methods such as detection, association, correlation, and combination, obtains a more accurate state estimation and identity estimation of underwater targets, and obtains more reliable and accurate target information with a lower time complexity, thereby improving the use efficiency of detection equipment and laying a technical foundation for expanding the application scope of related equipment. The highlight of this method is that the process is scalable and suitable for fusing three types of acoustic detection data in multiple dimensions. Brief Description of the Drawings

[0014] Figure 1 It is a schematic diagram of the overall process of this method;

[0015] Figure 2 It is a flowchart of single-frame point target fusion, mainly describing the method of fusing detection data from different detection devices at the same timestamp;

[0016] Figure 3 It is a flowchart of sequential point target trajectory fusion, mainly describing the process of trajectory association and fusion of the fused single-frame point targets. Detailed Description of the Specific Embodiment

[0017] To make the purpose, technical solution, and advantages of the present invention clearer, the present invention will be further described in detail below in combination with specific embodiments and with reference to the accompanying drawings. It should be understood that these descriptions are exemplary and are not intended to limit the scope of the present invention. In addition, in the following description, the descriptions of well-known structures and technologies are omitted to avoid unnecessarily confusing the concepts of the present invention.

[0018] The method involved in the present invention is mainly used in underwater multi-target acoustic detection scenarios, where underwater targets include torpedoes, underwater submersibles, surface ships, etc. Considering the scalability of this information fusion method, the present invention is also applicable to surface and underwater cross-domain acoustic target detection and fusion scenarios. The multi-source sonar data involved in this embodiment are: even-dimensional passive sonar array detection data, even-dimensional passive vector array detection data, and 1-dimensional matching field terminal processing data, where each dimension of data comes from an acoustic detection (processing) data output device, and the passive sonar detection data and vector detection data are processed in the same way in the present invention.

[0019] This method has certain scalability and is suitable for integrating the output data of three types of acoustic detection equipment that meet the following requirements:

[0020] a. Passive sonar detection data set of even dimensions;

[0021] b. Vector detection dataset of even dimension;

[0022] c. 1-dimensional matching field dataset;

[0023] In order to clearly describe the acoustic data fusion method with more than 2 dimensions, this scheme assumes that the passive sonar detection data has 4 dimensions, the passive vector detection data has 2 dimensions, and the matching field detection data has 1 dimension.

[0024] On the basis of the above, the technical process steps of the present invention are as follows:

[0025] Step 1: Acquire three types of acoustic detection data and sampling intervals in real time, store the detection information in the local server after synchronizing the timestamp; calculate the minimum sampling interval (period) of multi-source data, and when the sampling data meets a certain threshold value, interpolate the multi-source detection data whose sampling interval is greater than the above minimum sampling interval;

[0026] Step 2: Perform initial information preprocessing on multiple types of acoustic detection information (sonar linear array detection data, vector array detection data, and matching field detection data) to remove null values ​​and abnormal values;

[0027] Step 3: Perform azimuth positioning and candidate target screening on the data in step 2 frame by frame to obtain a set of candidate target positions;

[0028] Step 4: Fuse the candidate target position set of the two types of acoustic detection information obtained in step 3 with the target orientation set after matching field processing, and form associated point targets frame by frame;

[0029] Step 5: Associate the tracks of the fused point targets in the time dimension and form a fused track of the underwater target;

[0030] Step 6: Repeat steps 2 to 5 to finally obtain the detected target track after association fusion.

[0031] Example:

[0032] Step 1: Obtain three types of acoustic detection data and sampling intervals in real time. After synchronizing the timestamps, store the detection information in the local server; calculate the minimum sampling interval (period) of the multi-source data. When the sampled data meets a certain threshold, interpolate the multi-source detection data with a sampling interval greater than the above minimum sampling interval.

[0033] As Figures 1-3 shown. In the above Step 1, three types of acoustic detection data are obtained in real time. S a 、S b 、S c 、S d represent the acoustic detection data sets, V 1 、V 2 represent the vector detection data sets, and M represents the matched field processing data set. The above data sets respectively correspond to the data measured by 4 sets of passive sonar detection equipment terminals (PS1, PS2, PS3, PS4), 2 sets of vector array detection terminals (PV1, PV2), and 1 set of matched field (PM1) detection terminals. The location of the data fusion center is O(0,0).

[0034] Therefore, the dimensions of the above three types of acoustic detection data are 4, 2, and 1 respectively. The dimension represents the number of array elements of a single type of acoustic detection device, that is, the number of output terminals of the detection data.

[0035] In an actual scenario, the sampling interval of the sonar array detection device is 1 s, the data sampling interval of the vector array detection device is 2 s, and the sampling interval of the matched field detection data is 1 s. That is, the sampling intervals of the three types of real-time detection data are: T s = 1 s, T v = 2 s, T m = 1 s, and the minimum time interval is min(T s , T v , T m ) = 1 s. When storing the detection data, always use the current time of the local central server as the local time, and use this as the synchronization method for the timestamps of the multi-source data.

[0036] The minimum sampling interval described in the above Step 1 is min(T s , T v , T m ) = 1 s, that is, interpolation should be performed on the data with a sampling interval greater than 1 s. For V iThe (i=1,2) data set is interpolated, and the passive detection data structure is: {T, targno, azi} (T represents the detection timestamp, targno represents the detection target number of the current detection array element, and azi represents the target azimuth), that is, the azimuth and detection time are interpolated based on the target motion trajectory.

[0037] Step 2: Perform initial information preprocessing on multiple types of acoustic detection information (sonar linear array detection data, vector array detection data, and matching field detection data) to remove null values ​​and abnormal values;

[0038] The null value in step 2 refers to the angle information in the acquired acoustic detection data being a "null" value or other value that cannot be converted into a numerical type, and the preprocessing operation for the null value is to discard the data;

[0039] The abnormal value in step 2 refers to the azimuth angle value in the acquired acoustic detection data not being in the interval [0°, 360°), and the preprocessing operation for the abnormal value is to discard the data;

[0040] Step 3: Perform azimuth positioning and candidate target screening on the data in step 2 frame by frame to obtain a set of candidate target positions;

[0041] The azimuth positioning described in step 3 is the process of intersecting the angle values ​​of the target detected by different terminals of the same type of equipment at the same time and obtaining a set of candidate target positions. Generally, two detection terminal output data angle systems are selected for intersection. When the output data exceeds 2 dimensions, the 2-dimensional data with the largest variance of the detection data angle values ​​is selected for azimuth positioning each time.

[0042] The above two types of multi-terminal detection data specifically refer to sonar array and vector array. The candidate target position set obtained after the intersection and positioning of the target detection angle values ​​of two elements in the four elements of the sonar array is S 1 , S 2 ; After the two array elements of the vector array detect the target angle values, the candidate target position set V is obtained;

[0043] For the aforementioned S 1 , S 2 The distance matrix is ​​calculated for the candidate targets in S. If the values ​​in the same row or column of the matrix are all greater than 10m, then the candidate target corresponding to the row or column label is in S 1 , S 2 Eliminate from.

[0044] Considering the expansion situation in the previous steps, if the number of sonar array elements is 6, the last two array elements will intersect and locate to form a set S 3 , then we need to construct S 1 , S 3 , S2 , S 3 The distance matrix formed is then eliminated according to the above method. 1 , S 2 , S 3 Candidate targets in .

[0045] Step 4: Fuse the multi-terminal passive acoustic detection candidate target position set obtained in step 3 with the matching field detection target orientation set, and form associated post-point targets frame by frame;

[0046] The candidate target position set of multi-terminal passive acoustic detection described in step 4 is V, S 1 , S 2 , a single data structure is {T, targno, x i ,y i}, where T is the timestamp, i∈(v,s 1 ,s 2 ).

[0047] The matching field passive detection target azimuth set M described in the above step 4 has an element data structure of {T, targno, azi}, where targno is taken from the independent target detection space and azi is the azimuth of the detection target.

[0048] The aforementioned targno is the target number, and the targnos in the four sets all belong to independent numbering spaces.

[0049] The "forming associated point targets frame by frame" in step 4 refers to fusing the above three types of targets from different target spaces to form a fused target space;

[0050] Step 4.1 V, S 1 , S 2 Perform coordinate transformation, convert it into Cartesian coordinates with the fusion center O(0,0) as the origin, and calculate the azimuth, that is, {T, targno, x i ,y i} is transformed into {T, targno, x, y, azi} to obtain a new time series target position dataset: V, S 1 , S 2 .

[0051] Step 4.2 After executing step 4.1, V and S 1 , S 2 , detect the data for each frame in the M set and perform target association:

[0052] 1) Get V and S of the current timestamp T 1 , S 2 , a subset S of data in M 1 (T), S2 (T), V(T), M(T), match the positions of the targets in the four subsets respectively. The specific rules are as follows:

[0053] 2) Traverse S 1 and S 2 For the target s 1 in S 1,i , if the target with the closest distance found in S 2 is s 2,j , and the Euclidean distance between the two targets is less than es (es is initialized to 5m), then it is considered that the two targets are successfully matched. Calculate the center point s 1,i of s 2,j and s k , and store it in the new set S; for the points in S 1 and S 2 that have no points to the target, they are all stored in the set S and a new targno is assigned;

[0054] As an extension, for a scenario with 6 sonar array elements, traverse S 1 , S 2 , and S 3 for the target S m,i in S (m ∈ [1, 2, 3], i ∈ [1, len(S m )]), if the target S n,j with the closest distance is found in the remaining sets (m ∈ [1, 2, 3], j ∈ [1, len(S n )]), and the Euclidean distance between the two targets is less than es (es is initialized to 5m), then it is considered that the two targets are successfully matched. Calculate the center point s 1,i of s 2,j and s k , and store it in the new set S; for the points in S 1 , S 2 , and S 3 that have no points to the target, they are all stored in the set S and a new targno is assigned;

[0055] 3) If the distance between the target S(m) in the S set and the target V(n) with the closest distance in the V set is less than es (es is initialized to 5m), then it is considered that S(m) finds a matching target in V(n);

[0056] 4) For the target matching in V and M, if the absolute deviation of the angle between the target V(n) in V and the target M(l) in M is less than ea, then it is considered that V(n) and M(l) are successfully matched; similarly, for the target matching in S and M, it is also the same;

[0057] If the target tg1 in S can find matching targets in both V and M, calculate the position of the target after matching (fusion), and assign a fused target number. The fused target is a type of associated target. If there is a match for the targets in exactly two of the candidate sets S, V, and M, calculate the position of the target after matching (fusion), and assign a fused target number. This type of target is a type II associated target. The fused target numbers after the above fusion form the fused target space as shown in Equation 1.1 below, where w1 represents type I associated targets and w2 represents type II associated targets.

[0058] RN(T,n,x,y,type)(n∈(1,2,...),(x,y)∈{R,R},type={'w1','w2'}) Equation 1.1

[0059] Step 5: Perform track association on the fused point targets tgr, tgr ∈ RN in the time series dimension, and form the fused track of underwater targets.

[0060] The data input in the above Step 5 is the point set TGR in the fused target space in the previous Step 4, and tgr is a point in the set. For the target tn1 at time T, if there was a target tn1 at time T - 1 and tn1 is in the fused track route1, determine whether the target tn1 at time T can be fused into the track route1. If it can be fused, add the target tn1 at time T to the track route1 and update the moving point range at the next time of route1. The specific method of path fusion is detailed in the appendix Figure 3 ;

[0061] Step 6: Repeat Steps 2 - 5 to finally obtain the track of the detected target after association and fusion.

[0062] It should be understood that the above specific embodiments of the present invention are only used for exemplary illustration or explanation of the principles of the present invention, and do not constitute a limitation to the present invention. Therefore, any modifications, equivalent replacements, improvements, etc. made without departing from the spirit and scope of the present invention shall be included within the protection scope of the present invention. In addition, the appended claims of the present invention are intended to cover all changes and modification examples that fall within the scope and boundaries of the appended claims, or equivalent forms of such scope and boundaries.

Claims

1. A multi-source information fusion method based on acoustic data, It is characterized in that Used to fuse the output data of three types of acoustic detection equipment: even-dimensional passive sonar detection data set, even-dimensional vector detection data set, and one-dimensional matching field data set; That The process comprises the following steps: Step S1: Acquire three types of acoustic detection data and sampling intervals in real time, and store the detection information in a local server after synchronizing the timestamp; Calculate the minimum sampling interval of multi-source data, and when the sampling data meets the set threshold value, interpolate the multi-source detection data whose sampling interval is greater than the minimum sampling interval; Step S2: performing initial information preprocessing on multiple types of acoustic detection information: sonar linear array detection data, vector array detection data, and matching field detection data, and removing null values ​​and abnormal values; Step S3: performing azimuth positioning and candidate target screening on each type of data in step S2 frame by frame to obtain a set of candidate target positions; Step S4: fusing the sonar linear array detection data, vector array detection data, candidate target position set and target orientation set after matching field processing obtained in step S3, and forming associated point targets frame by frame; Step S5: Associating the tracks of the fused point targets in the time series dimension and forming a fused track of the underwater target; Step S6: Repeat steps S2 to S5 to finally obtain the detected target track after association fusion.

2. The multi-source information fusion method according to claim 1, It is characterized in that The three types of acoustic detection data described in step S1 are: The S corresponding to the terminals PS1, PS2, PS3, and PS4 of the four sets of passive sonar detection devices respectively a , S b , S c , S d , which represents the acoustic detection data set and has a dimension of 4; The V corresponding to the two sets of vector array detection terminals PV1 and PV2 respectively 1 、V 2 to represent the vector detection data set, and its dimension is 2; A set of matching field PM1 detection terminals corresponding to M, to represent the matching field processing data set, and its dimension is 1; The data fusion center position is O(0,0); the dimension represents the number of array elements of a single type of acoustic detection equipment.

3. The multi-source information fusion method according to claim 2, It is characterized in that The sampling interval of the passive sonar array detection equipment is 1s, the data sampling interval of the vector array detection equipment is 2s, and the sampling interval of the matched field detection data is 1s; The sampling intervals of three types of real-time detection data are: T s = 1 s, T v = 2 s, T m = 1 s, and the minimum time interval is min(T s , T v , T m ) = 1 s; When probing data storage, the current time of the local central server is used as the local time, which is used as a way to synchronize the timestamps of multi-source data.

4. The multi-source information fusion method according to claim 3, It is characterized in that Interpolate the data with a sampling interval greater than 1 s for V i Interpolate the data sets for i = 1, 2. The data structure of the passive detection data is: {T, targno, azi}, where: T represents the detection timestamp, targno represents the detection target number of the current detection element, and azi represents the target azimuth angle.

5. The multi-source information fusion method according to claim 2, It is characterized in that In step S2, the null value refers to the angle information in the acquired acoustic detection data being a "null" value or other value that cannot be converted into a numerical type, and the preprocessing operation for the null value is to discard the data; The abnormal value in step S2 refers to the azimuth angle value in the acquired acoustic detection data not being within the interval [0°, 360°], and the preprocessing operation for the abnormal value is to discard the data; The azimuth positioning described in step S3 is a process of intersecting and locating the angle values ​​of the target detected by different terminals of the same type of equipment at the same time and obtaining a set of candidate target positions, and selecting the angle values ​​of the output data of two detection terminals for intersection; when the output data exceeds 2 dimensions, the 2-dimensional data with the largest variance of the detection data angle values ​​is selected each time for azimuth positioning; The set of candidate target positions obtained after intersecting and positioning the target angle values of every two elements among the 4 elements of the sonar array is S 1 and S 2 ; the set of candidate target positions obtained after intersecting and positioning the target angle values of the detection data elements of two elements of the vector array is V; The said S 1 and S 2 calculate the distance matrix of candidate targets. If there are values in the same row or the same column of the matrix that are all greater than 10m, then the candidate target corresponding to the row label or column label is excluded from S 1 and S 2 ; if the number of sonar array elements is 6, then after the intersection positioning of the angles of the last two array elements, a set S 3 is formed. Then, it is necessary to construct S 1 and S 3 , S 2 and S 3 to form a distance matrix. According to the above method, the candidate targets in S 1 and S 2 and S 3 are excluded again.

6. The multi-source information fusion method according to claim 3, characterized in that, The multi-terminal passive acoustic detection candidate target location sets are V and S 1 and S 2 , and the single data structure is {T, targno, x i , y i}, where T is the timestamp, and i ∈ (v, s 1 , s 2 ); the set M of passive detection target azimuths in the matched field, the element data structure is {T, targno, azi}, where the value of targno comes from the independent target detection space, and azi is the detection target azimuth angle; the targno is the target number, and the targno in the four sets all belong to the independent number space; the "forming associated point targets frame by frame" refers to fusing the above three types of targets from different target spaces to form a fused target space.

7. The multi-source information fusion method according to claim 6, characterized in that, For the multi-terminal passive acoustic detection candidate target position sets V, S 1 , S 2 Perform coordinate transformation to convert them into Cartesian coordinates with the fusion center O(0, 0) as the coordinate origin, and calculate the azimuth angle, that is, {T, targno, x i , y i} is transformed into {T, targno, x, y, azi}, and a new time-series target position data set is obtained: V, S 1 , S 2 .

8. The multi-source information fusion method according to claim 7, characterized in that, For the new time-series target position data sets: V, S 1 , S 2 and for each frame of data detection data in the M set, perform target association, including the following steps S: Step a, obtain the data subsets S of V, S, 1 , S, 2 , M, S 1 , S(T), 2 , S(T), V(T), M(T) at the current timestamp T, and perform position matching on the targets within the four subsets respectively; 1 and S 2 and the data subset S in M 1 S(T), 2 S(T), V(T), M(T), and perform position matching on the targets within the four subsets respectively; Step b, traverse S 1 and S 2 For the target s 1 in S 1,i , if the target with the closest distance in S 2 is s 2,j , and the Euclidean distance between the two targets is less than es, where es is initialized to 5m, then it is considered that the two targets are successfully matched, and the center point s 1,i of s 2,j and s k is calculated and stored in the new set S; for the points in S 1 and S 2 that have no points to the target, they are all stored in the set S and a new targno is assigned; as an extension, for the scenario with 6 sonar array elements, traverse S 1 and S 2 and S 3 for the target S m，i in S where m ∈ [1, 2, 3] and i ∈ [1, len(S m )], if the target S n，i with the closest distance is found in the remaining sets where m ∈ [1, 2, 3] and j ∈ [1, len(S n )], and the Euclidean distance between the two targets is less than es, where es is initialized to 5m, then it is considered that the two targets are successfully matched, and the center point s 1,i of s 2,j and s k is calculated and stored in the new set S; for the points in S 1 and S 2 and S 3 that have no points to the target, they are all stored in the set S and a new targno is assigned; Step c, if the target S(m) in the S set is less than es from the target V(n) with the closest distance in the V set, and es is initialized to 5m, then it is considered that S(m) finds a matching target in V(n); Step d, for the target matching in V and M, if the absolute angular deviation between the target V(n) in V and the target M(l) in M is less than ea, then it is considered that V(n) and M(l) are successfully matched; similarly for the target matching in S and M.

9. The multi-source information fusion method according to claim 8, characterized in that, in the said step d, if the target tg1 in S can find matching targets in both V and M, calculate the position of the fused target after matching, and assign a fused target number, and the fused target is a type of associated target; if there is and only two candidate sets of targets are matched in S, V, and M, calculate the position of the fused target after matching, and assign a fused target number, and this type of target is a type II associated target; the formula for the fused target space composed of the above fused target numbers is: RN(T, n, x, y, type) n ∈ (1, 2,...), (x, y) ∈ {R, R}, type = {'w1', 'w2'} where w1 represents a type I associated target and w2 represents a type II associated target.

10. The multi-source information fusion method according to claim 7, characterized in that, in the said step S5, perform track association on the fused point targets tgr, tgr ∈ RN in the time sequence dimension, and form a fused track of underwater targets; the data input in the above step S5 is the point set TGR in the fused target space in the previous step S4, and tgr is the point in the set; for the target tn1 at time T, if there is a target tn1 at time T - 1 and tn1 is in the fused track route1, judge whether the target tn1 at time T can be fused into the track route1, and if it can be fused, then add the target tn1 at time T to the track route1 and update the moving point range of route1 at the next moment.

Images