Multi-platform passive data fusion method based on underwater vehicle azimuth information compression

Abstract

The application provides a multi-platform passive data fusion method based on underwater vehicle azimuth information compression, comprising the following steps: step 1, obtaining frame data of each detection platform in a target tracking process, and performing compression processing on the data frame to obtain a measurement value of each detection platform; step 2, performing data association between any two measurement values of each detection platform at the current moment, and positioning the measurement values according to the association result; calculating the positioning information of the underwater vehicle through the azimuth information; and step 3, constructing a multi-sensor multi-target tracking scene according to the azimuth information, and sending the positioning information of the plurality of underwater vehicles to a fusion center for data fusion processing to obtain a positioning result.

Description

Technical Field

[0001] This invention relates to an underwater acoustic signal fusion technology, specifically a multi-platform passive data fusion method based on underwater vehicle azimuth information compression. Background Technology

[0002] Marine resources are a precious resource of the Earth. With the increasing scarcity of land resources, vigorously developing marine industries and scientifically utilizing marine resources has become one of the main tasks of the world's major coastal powers. However, whether it is to safeguard the rights to develop marine resources or to defend against maritime threats, long-range detection of maritime targets is essential. Due to the unique propagation characteristics of seawater, other common propagation media (such as light, electricity, and magnetism) cannot achieve long-distance signal transmission except for sound waves. Therefore, sonar technology based on sound wave transmission has become the key to this technology.

[0003] Since sonar detection may be mixed with a large amount of interfering target information, the measurement data of the sonar is usually filtered. Multi-target data fusion tracking method has been widely used due to its excellent filtering performance. Literature (Dai Mingzhen, Cui Shaohui. Multi-target tracking of sonar. Journal of Nanjing Aeronautical Institute, 1991(4):130-134.), dissertation (Liu Wei. Networked sonar target tracking technology. Harbin Engineering University, 2017.) and patent (Hu Peng, Feng Jinlu, Li Ranwei. An active sonar single-frequency tracking method. Application No.: CN201811069753) etc. utilize target position information to realize the detection of underwater targets through multi-target tracking method. Literature (Clark DE, Bell J. Bayesian multiple target tracking in forward scan sonar images using the PHD filter. IEE Proceedings-Radar, Sonar and Navigation, 2005, 152(5):327-0.), patents (Wang Xingmei, Wang Guoqiang et al. A forward-looking sonar underwater target tracking method based on adaptive particle swarm optimization. Application No.: CN201810757443), and patents (Wang Xingmei, Duan Binghua, Wang Guoqiang. An improved kernel correlation filter underwater target tracking method based on forward-looking sonar. Application No.: CN201810870281) have been used to detect targets in forward-looking (or side-scan) sonar images, realizing the detection and tracking of multiple targets at close range. Although the above methods have achieved good results, these methods are all based on active sonar detection mode. However, in long-range detection, active sonar is usually limited by its disadvantages such as easily exposing its own sonar position and short detection range.

[0004] Passive sonar, with its advantages of good concealment and long detection range, is widely used in target surveillance in large-scale scenarios. However, because passive sonar cannot obtain as many reliable non-cooperative target features as active sonar, many traditional active sonar data tracking and fusion methods are difficult to use. Literature (Li Juwei, Xu Yicheng, Sun Mingtai. Target motion analysis of passive directional sonar buoys. Electro-optics and Control, 2011, 18(12).), patents (Quan Hengheng, Zhou Bin, et al. A passive target tracking method based on interactive multi-model. Application No.: CN201810697477) and (Li Zheng, Huang Haining, Li Yu. An underwater acoustic multi-target autonomous detection and azimuth tracking method. Application No.: CN201310703672) utilize target azimuth information to perform data fusion tracking of targets, thereby achieving detection functionality.

[0005] In recent years, with the continuous increase in marine environmental noise, the detection probability of detection platforms is often very low. Target location information is usually one of the more easily obtained information during target detection. Therefore, this invention mainly considers a multi-platform passive data fusion method based on underwater vehicle location information compression. Summary of the Invention

[0006] In view of this, the present invention provides a data fusion method for multiple passive detection platforms and underwater vehicles based on azimuth information compression, which can solve the target tracking problem of multiple passive detection platforms and multiple underwater vehicles in a large-scale surveillance environment with low detection probability. The method includes:

[0007] Step 1: Acquire frame data of each detection platform during the target tracking process, and compress the data frames to obtain the measurement values ​​of each detection platform;

[0008] Step 2: Correlate any two measurement values ​​from each detection platform at the current moment, and locate the measurement values ​​pairwise based on the correlation results; calculate the positioning information of the underwater vehicle using the azimuth information;

[0009] Step 3: Based on the azimuth information, construct a multi-sensor multi-target tracking scenario, and send the positioning information of multiple underwater vehicles to the fusion center for data fusion processing to obtain the positioning result.

[0010] Specifically, the compression process of the data frame in step 1 includes: extracting the frame data every S frames, performing correlation estimation calculation on the frame data using the nearest neighbor algorithm, and correlating the estimation result at time k with the estimation result at time ks, with the correlation matrix denoted as α. ij If α ij If α ≤ T and is the minimum value, then the estimation results of the two frames are correlated, i.e., αij =min(α) i And α ij =min(α) j Where T is the correlation threshold, and in the above formula, i represents the i-th sonar array, j represents the j-th sonar array, and α i α represents the value of the i-th row of the correlation matrix. j The value of the j-th column of the correlation matrix is ​​used; the root mean square value is used as the measurement value of the current frame when the number of correlated frames is greater than the threshold T.

[0011] This can be described in detail as follows:

[0012]

[0013] Here, α ij This represents the absolute value of the difference between the azimuth angle in the i-th row and the azimuth angle in the j-th column, i.e. Here n i and n j Let represent the number of estimated values ​​at time k and time ks, respectively.

[0014] Specifically, in step 2, the data association between any two measurement values ​​of each detection platform at the current moment, and the pairwise location of the measurement values ​​based on the association result, includes: denoting the measurement value of each detection platform as x. i,1 ,x i,2 ...x i,n Where i is the i-th detection platform and n is the number of measurement values. The measurement values ​​of different detection platforms at the current time are correlated pairwise. According to the nearest neighbor correlation algorithm, if the Euclidean distance between two targets is less than or equal to the distance correlation threshold Td and is the smallest, then the two states can be considered to be correlated and the most correlated. Otherwise, they can be considered to be uncorrelated.

[0015] Specifically, in step 2, calculating the underwater vehicle's positioning information using azimuth information includes: correlating the measurement values ​​from different detection platforms at the current moment pairwise; positioning the correlated measurement values ​​pairwise; and using the target's azimuth angle measurements from the two detection platforms, the target's position at the current moment can be obtained. The specific positioning equation is as follows:

[0016]

[0017] Where, θ T and θ R Let x represent the azimuth angles measured by platforms T and R at the current moment, and y represent the position of the dual-probe platform (x, y). T ,y T ) and (x R ,y RGiven that the target's position coordinates (x, y) can be obtained, we can solve for the position coordinates of the target.

[0018]

[0019] Specifically, in constructing a multi-sensor, multi-target tracking scenario, if the measurement processes of each sensor are independent, and the estimation error covariances generated by each sensor are independent and incoherent, or the cross-covariance between any two sensors is negligible; and the measurements of the two sensors are sent to the fusion center simultaneously, the fusion result can be simplified to...

[0020]

[0021] P i Let P be the covariance matrix of the i-th detection platform. j Let be the covariance matrix of the j-th detection platform. This is the state estimation matrix; This represents the estimation error matrix; when the number of sensors N > 2, and all estimation errors... When all of them are unrelated, we have:

[0022]

[0023] Beneficial effects:

[0024] 1. Under the premise that the underwater vehicle is moving at a constant speed in a straight line, the present invention realizes the estimation of the position and motion status of multiple underwater targets by fusing multiple frames of azimuth data;

[0025] 2. The multi-platform passive data fusion method for compressing underwater vehicle position information in this invention achieves the effect of being easy to implement in engineering and requiring less computation.

[0026] 3. This invention fully utilizes azimuth information and the detection technology of multi-sonar platforms to achieve underwater vehicle data fusion, realizing large-scale underwater target detection and tracking. Attached Figure Description

[0027] Figure 1 This is a flowchart illustrating the data fusion method for multiple passive detection platform underwater vehicles based on azimuth information compression in this invention.

[0028] Figure 2 This is a schematic diagram of the positioning of the dual-detection platform in this invention;

[0029] Figure 3 This is a schematic diagram of the real-time motion status of the target in this invention;

[0030] Figures 4a-4b These are schematic diagrams of the original tracking result and the fused tracking result under low detection probability in this invention. Detailed Implementation

[0031] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0032] This invention provides a data fusion method for multiple passive detection platform underwater vehicles based on azimuth information compression, such as... Figure 1 As shown, the method of the present invention includes the following steps:

[0033] Step 1, data compression, which involves acquiring frame data from each detection platform during target tracking, and compressing the data frames to obtain the measurement values ​​of each detection platform;

[0034] Suppose that N frames of data are generated during a target tracking process, and data is extracted every s frames (i.e., s frames of data are compressed into one frame). Within these s frames, data association is first performed using the nearest neighbor algorithm. The estimation result at time k is then associated with the estimation result at time ks, and the association matrix is ​​denoted as α. ij If α ij If α ≤ T and is the minimum value, then the estimation results of the two frames are correlated, i.e., α ij =min(α) i And α ij =min(α) j Where T is the correlation threshold, and in the above formula, i represents the i-th sonar array, j represents the j-th sonar array, and α i α represents the value of the i-th row of the correlation matrix. j The value of the j-th column of the correlation matrix is ​​used; the root mean square value is used as the measurement value of the current frame when the number of correlated frames is greater than the threshold T.

[0035] This can be described in detail as follows:

[0036]

[0037] Here, α ij This represents the absolute value of the difference between the azimuth angle in the i-th row and the azimuth angle in the j-th column, i.e. Here n i and n j Let represent the number of estimated values ​​at time k and time ks, respectively.

[0038] Step 2, data association and positioning, involves associating any two measurement values ​​from each detection platform at the current moment, and positioning the measurement values ​​pairwise based on the association results; and calculating the positioning information of the underwater vehicle using the azimuth information.

[0039] Let x be the data obtained by each detection platform using the above method. i,1 ,xi,2 ...x i,n Where i is the i-th detection platform and n is the number of measurements. The measurements from different detection platforms at the current time are correlated pairwise. According to the principle of the nearest neighbor correlation algorithm, the correlation state can generally be divided into two types: correlated and uncorrelated. If the Euclidean distance between two targets is less than or equal to the distance correlation threshold Td and is the smallest, then the two states can be considered to be correlated and the most correlated; otherwise, they can be considered uncorrelated.

[0040] d is the calculated Euclidean distance.

[0041] The associated measurement values ​​are located pairwise, assuming that the underwater vehicle and the detection platform are connected as follows: Figure 2 The diagram illustrates a triangular relationship. In the figure, T and R represent two detection platforms at different locations, typically with a significant distance between them. If only azimuth information can be obtained from the detection platforms, then the target can be located using only the azimuth information measured by the two platforms. By measuring the target's azimuth angle using the two detection platforms, the target's current position can be determined. At this point, applying a dual-detection-platform positioning algorithm based on the direction of sound wave arrival allows for the location of the specific target. The specific positioning equation is shown below.

[0042]

[0043] Where, θ T and θ R Let x represent the azimuth angles measured by platforms T and R at the current moment, and y represent the position of the dual-probe platform (x, y). T ,y T ) and (x R ,y R Given. Therefore, we can solve for...

[0044]

[0045] As can be seen from the above formula, by combining the azimuth information of the two detection platform arrays, the target's position coordinates (x, y) can be obtained, that is, the target can be located.

[0046] Step 3, positioning result fusion, that is, based on the azimuth information, a multi-sensor multi-target tracking scenario is constructed, and the positioning information of multiple underwater vehicles is sent to the fusion center for data fusion processing to obtain the positioning result.

[0047] Assuming there are N sets of location results, the data fusion algorithm involved in this patent mainly considers the following two assumptions when fusing these N sets of location results:

[0048] (1) The measurement process of each sensor is independent of each other, and the estimation error covariance generated by each sensor is independent and incoherent or the cross-covariance of any two sensors can be ignored.

[0049] (2) The measurement of any two sensors and the time of sending them to the fusion center are synchronized.

[0050] Based on the two assumptions mentioned above, this paper mainly considers the "simple convex combination" method.

[0051] Suppose that the local estimates and error covariance matrices of sensor i and sensor j for the same target are respectively and Where m = i, j. Then, we have...

[0052]

[0053] in, and Let each be an expression representing the state estimation error matrix of the two sensors. Then, the fusion result can be simplified to:

[0054]

[0055] P j Let be the covariance matrix of the j-th detection platform. Represents the estimation error matrix; This is the state estimation matrix.

[0056] When the number of sensors N > 2, and all estimation errors When all of them are unrelated, we have:

[0057]

[0058] When fusing information from multiple sensors, each sensor is assigned a weight value, and the magnitude of this weight value determines the degree of influence of that sensor in the fusion algorithm. This paper adopts a weighted fusion method, assuming that all sensors have equal weights and allowing for a certain amount of false alarms to ensure the system's maximum detection capability. The main methods for weighted fusion are as follows:

[0059] (1) If the local estimation results of multiple sensors can be correlated, then a simple convex combination fusion method is used for data fusion.

[0060] (2) If the local estimation result of a certain sensor is not correlated with at least one other sensor, then the local estimation result of that sensor may be a false alarm or a real target. If the estimation result of that sensor can be correlated with the estimation results of other sensors, then this part of the sensor results is fused using a simple convex combination method, and the fused result is used as the final estimation result. Otherwise, this part of the result is regarded as clutter.

[0061] Within the monitored area of ​​[-5000, 5000] × [-5000, 5000] (meters), the sonar detection cycle is 1 second, generating a total of 500 frames of simulated data. In each frame, the target's state information is its azimuth angle. In each observation, it is assumed that the passive sonar will be subjected to 0-15 randomly varying interfering targets, with a target detection probability of 0.4.

[0062] In this simulation, there are three passive detection platforms and three moving targets. The target motion status and target positions are shown in the figure below. The passive sonar positions are (-1000m, -500m), (1000m, 0m), and (0m, 1000m). Target 1 starts at (1000m, 300m) and moves at a speed of (-4, 0) (m / s). Target 2 starts at (-1000m, 800m) and moves at a speed of (2, -4) (m / s). Target 3 starts at (1000m, -500m) and moves at a speed of (-5, -1) (m / s). The motion time of targets 1 and 2 is from the first frame to the 500th frame, while target 3 appears from the 100th frame and moves until the 500th frame. The real-time motion status of the targets is shown in the figure below. Figure 3 As shown in the figure, the tracking results obtained under the above simulation conditions are as shown in the figure. The target is extracted once every 10 frames, and the association threshold is 4.

[0063] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

[0064] It will be apparent to those skilled in the art that the embodiments of the present invention are not limited to the details of the exemplary embodiments described above, and that the embodiments of the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the embodiments of the present invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the embodiments of the present invention is defined by the appended claims rather than the foregoing description. Therefore, all variations falling within the meaning and scope of equivalents of the claims are intended to be encompassed within the embodiments of the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims. Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units, modules, or devices recited in the system, apparatus, or terminal claims may also be implemented by the same unit, module, or device through software or hardware. The terms "first," "second," etc., are used to indicate names and do not indicate any particular order.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the embodiments of the present invention and are not intended to limit them. Although the embodiments of the present invention have been described in detail with reference to the above preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions to the technical solutions of the embodiments of the present invention should not depart from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-platform passive data fusion method based on underwater vehicle azimuth information compression, characterized in that, include: Step 1: Acquire frame data for each detection platform during target tracking. Compress the data frames to obtain measurement values ​​for each detection platform. Specifically, this includes: extracting the frame data every S frames, performing association estimation calculations on the frame data using a nearest neighbor algorithm, and then... The time estimation results are respectively and The estimation results at time points are correlated, and the correlation matrix is ​​denoted as follows. ,like And if it is the minimum value, then the estimation results of the two frames are correlated, that is... and in For the correlation threshold, in the above formula, Indicates the first A sonar array Indicates the first A sonar array Represents the first element of the correlation matrix. The value of the row, Represents the first element of the correlation matrix. The value of the column; the number of associated frames is greater than the threshold. The result is used as the root mean square value as the measurement value of the current frame; This can be described in detail as follows: here, Indicates the first The azimuth of the line and the first The absolute value of the difference in azimuth angles of the column, i.e. ,here and Let represent the number of estimates at time k and time ks, respectively; Step 2: Correlate any two measurement values ​​from each detection platform at the current moment, and locate the measurement values ​​pairwise based on the correlation results; calculate the positioning information of the underwater vehicle using the azimuth information; Step 3: Based on the azimuth information, construct a multi-sensor multi-target tracking scenario, and send the positioning information of multiple underwater vehicles to the fusion center for data fusion processing to obtain the positioning result.

2. The multi-platform passive data fusion method based on underwater vehicle position information compression as described in claim 1, characterized in that, In step 2, data correlation is performed between any two compressed measurement values ​​from each detection platform at the current moment. Based on the correlation result, the measurement values ​​are located pairwise, including: recording the estimated value of each detection platform as... ,in, For the first One detection platform, To determine the number of estimated values, pairwise data correlation is performed on the estimated values ​​from different detection platforms at the current moment. Based on the nearest neighbor correlation algorithm, if the Euclidean distance between two targets is less than or equal to the distance correlation threshold... If the first state is the smallest, then the two states can be considered to be related and the most related; otherwise, they can be considered to be unrelated.

3. The multi-platform passive data fusion method based on underwater vehicle position information compression as described in claim 2, characterized in that, Step 2, calculating the underwater vehicle's positioning information using azimuth information, includes: pairwise data correlation of the estimated values ​​from different detection platforms at the current moment; pairwise positioning of the correlated estimated values; and using the target's azimuth angle measurements from the two detection platforms to obtain the target's position at the current moment. The specific positioning equation is as follows: in, and They represent the platforms respectively. and platform The position of the dual-probe platform at the azimuth angle measured at the current moment. and Given; from this, the target's position coordinates can be obtained. ,in 。 4. The multi-platform passive data fusion method based on underwater vehicle position information compression as described in any one of claims 1-3, characterized in that, To construct a multi-target tracking scenario involving multiple underwater vehicles, if the measurement processes of each sensor are independent, and the covariance of the estimation errors generated by each sensor is independent and incoherent, or the cross-covariance between any two sensors is negligible; and if the measurements of the two sensors are sent to the fusion center simultaneously, the fusion result can be simplified to... For the first The covariance matrix of each detection platform; For the first The covariance matrix of each detection platform This is the state estimation matrix; Represents the estimation error matrix; When the number of sensors At that time, and all estimation errors When they are all unrelated, we have: 。

Images