Underwater sound weak target and cross trajectory extraction method based on azimuth history diagram

Abstract

The invention discloses an underwater sound weak target and cross trajectory extraction method based on an azimuth history diagram, and belongs to the technical field of underwater sound detection target azimuth history diagram extraction. The technical problems that weak target azimuth trajectories are difficult to extract and target azimuth trajectory association is prone to errors are solved. The method comprises the following steps: firstly, carrying out interference azimuth elimination on an azimuth history graph in an input period of time, carrying out bisection segmentation on detectable azimuth, carrying out adaptive threshold segmentation after data difference stretching, and carrying out trajectory region communication and trajectory extraction based on morphological operation; and a track correlation method for setting a threshold value based on a track end and overall angle characteristic, a straight line fitting characteristic and a target maneuvering characteristic. The method can effectively solve the problems of trajectory intersection, weak target trajectory disconnection, target maneuvering curve disconnection and the like in the azimuth history graph, and can quickly process data within a period of time of long-term sonar return to sea observation.

Description

Technical Field

[0001] The present invention belongs to the technical field of extracting the azimuth history diagram of underwater acoustic detection targets, and particularly relates to a method for extracting weak underwater acoustic targets and cross trajectories based on the azimuth history diagram. Background Art

[0002] At present, the demand for detecting underwater targets in the ocean is increasing day by day. The main technical means is to detect the radiated noise of underwater targets through sonar to achieve target directional tracking. Conducting target azimuth history trajectory tracking based on the underwater acoustic target azimuth history diagram is a commonly used and effective method for underwater targets. The principle is that the acoustic noise intensity in the azimuth where there is an underwater acoustic target is stronger than that in the azimuth where there is no underwater acoustic target. By combining beamforming means to compensate the signal in the time series, a stronger signal in the target azimuth can be obtained. If the target persists, a stable trajectory will appear on the azimuth history diagram, and this trajectory will bend, tilt, etc. according to the change of the target azimuth.

[0003] In the actual application process, for targets with strong acoustic signals, trajectories with strong contrast and brightness can be obtained. However, for some targets with weak signals, due to ocean environmental noise, interference from nearby strong target signals, etc., bright azimuth history trajectories cannot be obtained, which makes it difficult for common trajectory extraction methods to effectively extract the azimuth trajectories of weak targets. At the same time, there are also situations where multiple targets cross in the azimuth trajectory. It is difficult to judge whether the target trajectory bends here during the trajectory association process, which brings great difficulty to the target trajectory association and causes errors in the target azimuth trajectory association. Summary of the Invention

[0004] (1) Invention Objectives

[0005] The objective of the present invention is to propose a method for extracting weak underwater acoustic targets and cross trajectories based on the azimuth history diagram, so as to solve the technical problems in the prior art that it is difficult for the trajectory extraction method to effectively extract the azimuth trajectories of weak targets, and there are also situations where multiple targets cross in the azimuth trajectory, and it is easy to make mistakes in the target azimuth trajectory association.

[0006] (2) Technical Solutions

[0007] To achieve the above objective and solve the above technical problems, the technical solution of the present invention is as follows:

[0008] The present invention is mainly applicable to extracting target trajectories from the azimuth history diagram within a period of time, and can provide effective information for the azimuth distribution and situation estimation of underwater targets. To avoid the interference of self-noise on the extraction of the underwater acoustic target azimuth history trajectory, the method of the present invention combines the self-detection azimuth ability of the sonar array to eliminate the azimuth angle information that cannot be detected.

[0009] The detectable angle range of the sonar array is split to effectively ensure the contrast intensity of the target azimuth trajectory within the relative block azimuth, providing favorable conditions for the segmentation of the target azimuth trajectory;

[0010] For the split azimuth history diagram, the background noise is estimated to determine the target azimuth trajectory segmentation threshold, and the morphological algorithm is used to process the segmented azimuth history diagram to achieve a large amount of background noise removal and refine the target azimuth trajectory to ensure the azimuth accuracy;

[0011] The processed split azimuth history diagrams are fused and the target trajectory blocks are extracted, and the target trajectory boundary, the target trajectory and the trajectory intersection point are detected in combination with the original azimuth information;

[0012] For the case where the target azimuth trajectory is interrupted, the signal correlation and the overall trend change of the trajectory are carried out to calculate the multi-trajectory correlation, so as to realize the association of the relevant trajectories and compensate the trajectory in the interrupted area;

[0013] For the large maneuver curve association, in the end, the present invention calculates the correlation of the terminal trends and the mutual distances of each section of the azimuth trajectory to realize the curve trajectory association;

[0014] Marking different trajectories with different colors can effectively display the azimuth trajectory intersection process of different targets;

[0015] For the weak target azimuth trajectory, the present invention realizes effective trajectory extraction and association, solves the problem that it is difficult to detect and associate the weak target trajectory by conventional methods, and has strong reliability.

[0016] (III) Effective benefits

[0017] 1. By dividing the azimuth of the azimuth history diagram, stretching the data histogram to enhance the contrast, combining the adaptive threshold calculation and the binarization processing, the present invention can effectively extract the weak trajectory area, and can effectively realize the connection and segmentation of the trajectory area through the morphological opening and closing operations.

[0018] 2. For the trajectories inside the connected block, they must be continuous. By analyzing and processing the boundary point features, the present invention can effectively calculate the coordinate information of the boundary intersection points, the trajectory convergence points and the trajectory bifurcation points.

[0019] 3. The present invention performs data fitting on the trajectories between paired intersection points to realize the smooth processing of the trajectories within the region. For the trajectories in the global range, the trajectory association is completed through the judgment of whether there is a time series intersection, the distance judgment and the angle change limitation based on the time series and the maneuver characteristics, so as to effectively solve the problems of trajectory disconnection, trajectory intersection and trajectory maneuver bending in the trajectory association extraction. Description of the drawings

[0020] Figure 1 This is an example of the effect after interpolation of the azimuth history diagram with weak target trajectories and trajectory intersections in the examples of the present invention;

[0021] Figure 2 This is the threshold segmentation effect of the azimuth history diagram in the example of the present invention;

[0022] Figure 3 This is the trajectory region diagram after morphological operation on the azimuth history diagram in the example of the present invention to eliminate background noise;

[0023] Figure 4 These are the trajectory segments extracted from each connected region of the azimuth history diagram in the example of the present invention;

[0024] Figure 5 This is a schematic diagram of associated trajectories within the connected region of the azimuth history diagram in the example of the present invention;

[0025] Figure 6 This is a schematic diagram of the global trajectory association result of the azimuth history diagram in the example of the present invention. Detailed implementation manners

[0026] The present invention will be described in detail below in conjunction with the accompanying drawings and specific implementation manners. However, the protection scope of the present invention is not limited to the following embodiments, and should include all the contents in the claims. Moreover, those skilled in the art can realize all the contents in the claims from the following one embodiment.

[0027] The design idea of the present invention is described as follows. A method for extracting underwater acoustic weak targets and cross - azimuth trajectories based on an azimuth history diagram includes eliminating interference azimuths from the azimuth history diagram within the input time period, bisecting the detectable azimuths, performing adaptive threshold segmentation after data - difference stretching, connecting and extracting the trajectory region based on morphological operations, and a trajectory association method that sets thresholds based on the angle characteristics of the trajectory end and the whole, the linear fitting characteristics, and the target maneuvering characteristics. The method of the present invention can effectively solve problems such as trajectory intersections, disconnections of weak target trajectories, and disconnections of target maneuvering curves in the azimuth history diagram. It can quickly process the data within a short time period of the sonar echo for long - term sea observations, extract effective target trajectory characteristics, and provide useful information for centralized data processing. Among them, in the process of trajectory association, the smaller the angle difference between two trajectories and the closer the end - straight - line parameter characteristics are, the more likely they are the same target. In the case where there are large - maneuvering curves and trajectory disconnections, the maneuvering angle threshold setting determined jointly based on the assumption of the target maneuvering ability within a period, the sampling time - series difference, and the sampling frequency proposed by the present invention can effectively solve this problem.

[0028] The specific implementation process of the present invention is as follows in conjunction with the accompanying drawings:

[0029] Step 1: To better ensure the accuracy of the azimuth history trajectory, perform bilinear interpolation on the input azimuth history map according to prior information, and expand the image to 1800 azimuths, as Figure 1 shown.

[0030] Step 2: Normalize the expanded azimuth history map and eliminate the interference area, and then bisect the remaining part, which eliminates the interference of sonar self-noise on the extraction of the azimuth history map and also ensures the extractable effect of weak targets.

[0031] Step 3: Stretch the histogram of the segmented azimuth history map to enhance the contrast, and further perform adaptive threshold segmentation to remove redundant background information.

[0032] Step 4: Perform two filtering processes on the segmented azimuth history map, using the filtering kernels h 1 =ones(10,1) / 20 and h 2 =ones(5,5) / 25.

[0033] Step 5: Calculate the segmentation threshold for the filtered regional image by the OTSU method and perform binary segmentation. The effect after threshold segmentation is as Figure 2 shown.

[0034] Step 6: Stitch and fuse all regional images into the size of the original azimuth history map, denoted as ImageTotal.

[0035] Step 7: Perform morphological opening and closing operations on ImageTotal to eliminate scattered noise interference points, as Figure 3 shown.

[0036] Step 8: Extract connected regions from ImageTotal after morphological opening and closing processing and remove regions with too small areas. Then extract the boundary coordinates of each connected region and store them separately.

[0037] Step 9: Calculate the number of boundary points for the stored connected regions according to the time series, and calculate the gradient of the boundary number values in the time series. The position of the time point where the gradient changes is the position of the time point where the number of trajectories changes, such as trajectory increase, trajectory intersection, trajectory disappearance, etc. The relevant situations are shown in Figure 3 as follows.

[0038] Step 10: Classify the points where the number of boundaries changes. Taking this point as the center, if the number of boundary points at the previous moment is 1 more and the number of boundary points at the next moment is 1 less, then this point is a lower crossing point; otherwise, it is an upper boundary crossing point. Save the coordinates of this point in the crossing point container within the corresponding connected region. Denote it as JCUP or JCDN = append([t, theta]), where [t, theta] are the time and azimuth angle respectively.

[0039] Step 11: Pair JCUP and JCDN, and separately save the boundary values between the paired points.

[0040] Step 12: For the crossing points that cannot be paired, they are considered convergence points or bifurcation points. Evaluate them. If it is at the estimated uppermost end, denote it as DU type, and if it is at the lowermost end of the trajectory, denote it as DA type.

[0041] Step 13: To ensure the smoothness of the trajectory change, some data between JCUP and JCDN are supplemented by bilinear interpolation and compared and corrected with the original data, effectively solving the trajectory deviation caused by the disappearance of boundaries for multiple trajectories.

[0042] Step 14: Segment the boundary data within the same connected region with the crossing points as the segmentation points and obtain their corresponding central trajectories, getting multiple target trajectories. Multiple trajectories can be viewed as examples from Figure 4 which.

[0043] Step 15: For each trajectory segment, calculate the line segment characteristics of its last port at 4 and 30 time points, including the linear fitting slope K and offset b, and synchronously calculate the distance disAB between every two line segments.

[0044] Four points form a line segment, and 30 points also form a line segment. By selecting line segments of different lengths, the short-time characteristics within 4 times at the end of the trajectory and the slightly longer trajectory characteristics within 30 time points can be described.

[0045] Step 16: Taking the coordinates of each sampled trajectory line segment as vectors, calculate the direction angle differences between the vectors.

[0046] θ = atan((K 2 - K 1 ) / (1 + K 1 K 2 ))

[0047] K1 and K2 are the slopes of two different straight lines.

[0048] Step 17: Correlate all the trajectories. Taking trajectory A as the benchmark, analyze the three trajectories with the smallest disAB value within the same connected region. If two trajectories do not have an overlapping region in time and the offset difference between the trajectories meets Condition 1, then the two trajectories are regarded as a pair of candidate targets for the same object.

[0049] Condition 1: |b 1 - b 2 | < B, where B is the set difference threshold between the bs. If the condition is met, it is considered that they may be the same object.

[0050] The analytical formula of a straight line is y = kx + b. There will be two bs for two straight lines, denoted as b1 and b2, and b 1 and b 2 are the offsets of two different straight lines;

[0051] Step 18: Compare the angular differences θ of all pairs of candidate trajectories. If Condition 2 is met, it is considered that they are the same object and the two trajectories are merged.

[0052] Condition 2: |θ 1 - θ 2 | < 5°. Considering the mobility of the object, the object within the same connected region will not have excessive mobility.

[0053] The included angle between the two straight lines obtained by calculation, θ 1 and θ 2 are the angles between the two straight lines and the X-axis in the rectangular coordinate system respectively.

[0054] Step 19: Traverse all connected regions, repeat all the operations from Step 1 to Step 18, save the obtained trajectories, denoted as GlobalTraces. The example diagram of the correlated trajectories after removing the false trajectories caused by sonar self-noise interference is as Figure 5 shown. Compared with Figure 4 , it greatly reduces the number of trajectory segments and significantly reduces the computational workload for global trajectory correlation.

[0055] Step 20: Calculate the features of all the trajectories in GlobalTraces, including the terminal angle θ 4 and the general trend angle θ 20 . If there are less than 20 time series data, calculate the general trend angle of all time series as θ 20 .

[0056] Step 21: Judge whether there is time series overlap between any two trajectories in GlobalTraces. If there is overlap, they are not the same object. Calculate the angular difference disT and the distance disAB for the trajectories without overlapping time series.

[0057] Step 22: Traverse all trajectories. Taking trajectory A as an example, compare the angles of the two trajectories with the closest distance to A. If condition 3 is satisfied, they are considered to be the same target. Satisfying condition 3 can, to a certain extent, solve the problem of associating large maneuvering trajectories of targets.

[0058] Condition 3: disT(A,B)<15*|(t 1 -t 2 )| / f, where |(t 1 -t 2 )| is the terminal time difference of the two trajectories, 15 is prior knowledge, assuming that the maximum maneuvering ability of the target is 15 degrees within a unit cycle time, and f is the sampling frequency of the azimuth history diagram.

[0059] A trajectory is a function of the position corresponding to time. t 1 、t 2 are the times corresponding to the last points of the two trajectories;

[0060] Step 23: Save the associated target trajectories and mark them with lines of different colors on the azimuth history diagram. Display the weak and discontinuous regions of the target trajectories and provide a judgment basis for computer reading. After smoothing the associated trajectories according to the situation, the final example effect is shown in Figure 6 .

[0061] Preferably, the number of columns for eliminating the interference region is 280, and the value for histogram stretching is 1.5.

[0062] Preferably, the gray value for segmenting the low-gray region and the high-gray region in the adaptive segmentation is 200, and the adaptive threshold is 0.25 times the sum of the maximum values of the low-gray region and the high-gray region.

[0063] Preferably, the two filtering processes for segmenting the image use kernels h 1 =ones(10,1) / 20 and h 2 =ones(5,5) / 25.

[0064] Preferably, the morphological opening and closing operations use diamond-shaped and horizontal and vertical linear kernels.

[0065] Preferably, during the trajectory association process, the angle difference threshold within the same connected region is set to 5°, and the maneuvering ability within the target maneuvering period is set to 15°.

[0066] Embodiment 1

[0067] The present invention proposes a method for extracting underwater acoustic weak targets and cross-bearing trajectories based on an azimuth history diagram, with Figure 1The three azimuth history diagrams in the following are used as examples. After loading a certain azimuth history diagram respectively, first, the size and number of azimuths of the azimuth history diagram are detected, the image is supplemented and expanded, and then the azimuth history diagram is segmented and the data contrast is stretched. Combining with the OTSU adaptive threshold, the trajectory area is segmented and binarized. See Figure 2 , the segmented images are fused and stitched, and further morphological opening and closing processing is adopted, so that the connected areas of the trajectories can be in the same area as much as possible and small scattered noises are removed. See Figure 3 .

[0068] As Figure 4 shown, the boundaries and trajectories of the processed azimuth history diagrams are extracted, the trajectory segments in each connected area are obtained and marked for storage. For multiple trajectories in the same connected area, the distances, directions, and linear fitting parameters are compared to evaluate whether they are trajectories of the same target. To a certain extent, the number of trajectory segments can be reduced, thereby reducing the number of globally associated trajectories and improving the algorithm efficiency. See Figure 5 shown.

[0069] As Figure 6 the results show, finally, the globally associated multiple target azimuth trajectories are obtained, and the target trajectory association is completed by evaluating the azimuth differences, timing differences, etc. between the trajectories. The associated trajectories can be smoothed according to needs, which is more in line with the reality of the smooth movement of the target.

[0070] Therefore, the embodiments of the present invention have been described above in conjunction with the accompanying drawings. However, the present invention is not limited to the above specific embodiments. The above embodiments are merely illustrative and not restrictive. Under the inspiration of the present invention, those of ordinary skill in the art can also make many forms without departing from the spirit and scope protected by the claims of the present invention. These all fall within the protection scope of the present invention. The parts not elaborated in detail in the present invention belong to the well-known technologies of those skilled in the art.

Claims

1. A method for extracting weak underwater acoustic targets and cross-trajectory based on an azimuth history diagram, characterized in that: The steps include: Step 1: Perform bilinear interpolation on the input orientation history map based on prior information and expand the image to 1800 orientations; Step 2: Normalize the expanded orientation history map and eliminate the interference area, and then divide the remaining part into two equal parts; Step 3: Perform histogram stretching on the segmented orientation history map to enhance the contrast, and further perform adaptive threshold segmentation to remove redundant background information; Step 4: Perform two filtering processes on the segmentation orientation history map; Step 5: Calculate the segmentation threshold of the filtered regional image using the OTSU method and perform binary segmentation; Step 6: All regional images are stitched and fused into the original orientation history map size, recorded as ImageTotal; Step 7: Perform morphological opening and closing operations on ImageTotal to eliminate scattered noise interference points; Step 8: Extract the connected areas of ImageTotal after morphological opening and closing processing and remove the areas with too small areas, then extract the boundary coordinates of each connected area and store them separately; Step 9: Calculate the number of boundary points of the stored connected areas according to the time series, and find the gradient of the boundary value in the time series. The time point where the gradient changes is the time point where the number of trajectories changes, such as the number of trajectories increases, the trajectories cross, and the trajectories disappear; Step 10: Classify the points where the number of boundary points changes. Take the point as the center. If the number of boundary points at the previous moment is 1 more and the number of boundary points at the next moment is 1 less, then the point is the lower boundary intersection point. Otherwise, it is the upper boundary intersection point. The coordinates of the point are saved in the intersection container in the corresponding connected area; recorded as JCUP or JCDN = append([t, theta]), where [t, theta] are time and azimuth respectively; Step 11: Pair JCUP and JCDN, and save the boundary values ​​between the paired points separately; Step 12: For the intersection points that cannot be paired, they are considered to be convergence points or bifurcation points, and they are evaluated. If they are at the top of the estimation, they are recorded as DU type, and if they are at the bottom of the trajectory, they are recorded as DA type; Step 13: To ensure the smoothness of trajectory changes, some data between JCUP and JCDN are supplemented by bilinear interpolation and compared with the original data for correction; Step 14: segment the boundary data in the same connected area using the intersection as the segmentation point and calculate the corresponding center trajectory to obtain multiple target trajectories; Step 15: For each target trajectory, calculate the line segment characteristics of the last 4 and 30 time points of each target trajectory, including its straight line fitting slope K and offset b, and simultaneously calculate the distance disAB between each pair of line segments; Step 16: Using the coordinates of each sampled trajectory segment as a vector, calculate the direction angle difference between each vector; θ=atan((K2-K1) / (1+K1K2)) K1 and K2 are the slopes of two different straight lines; Step 17: Correlate all the trajectories. Taking trajectory A as the reference, analyze the three trajectories with the smallest disAB value within the same connected region. If two trajectories do not have an overlapping region in time and the offset difference between the trajectories meets Condition 1, then consider the two trajectories as a candidate pair for the same target; Condition 1: |b1 - b2| < B, where B is the set difference threshold between bs. If this condition is met, it is considered that they may be the same target; b1 and b2 are the offsets of two different straight lines; Step 18: Compare the angular differences θ of all candidate trajectory pairs; if Condition 2 is met, it is considered the same target and the two trajectories are merged; Condition 2: |θ1 - θ2| < 5°, considering the target mobility, targets within the same connected region will not have excessive mobility; θ1 and θ2 are the angles of two straight lines with the X-axis in the rectangular coordinate system; Step 19: Traverse all connected regions, repeat all operations from Step 1 to Step 18, save the obtained trajectories, denoted as GlobalTraces, and obtain the trajectory association example diagram after removing the false trajectories caused by sonar self-noise interference; Step 20: Calculate the features of all the trajectories in GlobalTraces, including the terminal angle θ4 and the large trend angle θ 20 If there are less than 20 time series data, the trend angle of all time series is calculated as θ 20 ; Step 21: Determine whether there is a temporal overlap between any two trajectories in GlobalTraces. If there is an overlap, they are not the same target; calculate the angular difference disT and the distance disAB for the trajectories without overlapping time sequences; Step 22: Traverse all the trajectories. Taking trajectory A as an example, compare the angles of the two trajectories with the closest distance to A. If Condition 3 is met, it is considered the same target; Condition 3: disT(A,B) < 15 * |(t1 - t2)| / f, where |(t1 - t2)| is the terminal time sequence difference between the two trajectories, 15 is prior knowledge, assuming that the maximum maneuverability of the target is 15 degrees within a unit cycle time, and f is the sampling frequency of the azimuth history diagram; t1 and t2 are the times corresponding to the last points of the two trajectories; Step 23: Save the target trajectories with completed trajectory association, mark them with lines of different colors on the azimuth history diagram, display the weak and disconnected regions of the target trajectories, and provide a judgment basis for computer reading.

2. The method for extracting weak underwater acoustic targets and cross-trajectory based on the azimuth history diagram according to claim 1 is characterized in that: In Step 2, the number of columns for eliminating the interference region is taken as 280, and the histogram stretching value is 1.

5.

3. The method for extracting weak underwater acoustic targets and cross-trajectory based on azimuth history diagram according to claim 1 is characterized in that: In Step 3, the gray value for segmenting the low-gray region and the high-gray region in the adaptive threshold segmentation is 200, and the adaptive threshold is 0.25 times the sum of the maximum values of the low-gray region and the high-gray region.

4. The method for extracting weak underwater acoustic targets and cross-trajectory based on azimuth history diagram according to claim 1 is characterized in that: In Step 4, the two filtering processes for the segmented image use kernels h1 = ones(10,1) / 20 and h2 = ones(5,5) / 25.

5. The method for extracting weak underwater acoustic targets and cross-trajectory based on azimuth history diagram according to claim 1 is characterized in that: In Step 7, the morphological opening and closing operations use diamond-shaped and horizontal and vertical linear kernels.

6. A method for extracting weak underwater acoustic targets and cross-trajectory based on an azimuth history diagram according to any one of claims 1 to 5, characterized in that: During the trajectory association process, the angular difference threshold within the same connected region is set to 5°, and the maneuverability within the target maneuvering period is set to 15°.

Images