Marine vessel trajectory association method, device, equipment and medium

Abstract

The application provides a marine ship track association method, device, equipment and medium, according to the method of the application, including: using OpenCV to extract marine ship images for pretreatment, and then using edge detection and contour processing to obtain the size characteristics of the marine ship; extracting the contour barycenter position of the marine ship based on the six-parameter affine transformation matrix of GDAL, and extracting the target latitude and longitude coordinates; obtaining the multi-modal characteristics of the marine ship, and performing weighted fusion on the multi-modal characteristics, taking the N targets with the highest probability as a candidate set; calculating the predicted target position after obtaining the heading and speed of the marine ship based on the dynamic behavior characteristics; matching the target latitude and longitude coordinates in the candidate set with the predicted target position, selecting the predicted position with the minimum distance as the association point through the Euclidean distance calculation, generating a continuous track and forming target situation information. The technical scheme provided by the application can improve the target recognition accuracy and form a continuous target situation.

Description

Technical Field

[0001] This document relates to the field of ocean vessel trajectory association technology, and in particular to a method, device, equipment and medium for ocean vessel trajectory association. Background Technology

[0002] In the field of marine target monitoring, existing technologies face the following main problems in the process of identifying unknown marine target slices in images and associating them with target trajectories to form a continuous situation:

[0003] First, TIFF images of marine targets are often affected by wave noise and have low resolution, resulting in blurred ship outlines. Traditional image segmentation algorithms, such as thresholding, struggle to accurately extract the length and width features of ships. For example, under wave interference, thresholding may misidentify some waves as ship outlines or miss the true ship edges, leading to subsequent misidentification of ship type.

[0004] Secondly, existing methods rely solely on single features such as length, width, or positional information, lacking a multi-dimensional verification mechanism, which can easily lead to confusion between ships of similar size. For example, two ships of different types but similar sizes are difficult to distinguish accurately based on length and width features alone, resulting in identification errors.

[0005] Third, due to the hourly delay in target location reporting, the reconnaissance data is not synchronized with the trajectory database, causing trajectory breaks and impairing the continuity of the target's situation. This makes it impossible to form a complete and continuous trajectory when analyzing the ship's movement, affecting the judgment and prediction of ship behavior.

[0006] These problems seriously affect the accuracy and real-time performance of marine target monitoring, requiring a new technological approach to solve them. Summary of the Invention

[0007] According to embodiments of the present invention, a method, apparatus, device, and medium for associating ocean vessel trajectories are provided, aiming to solve the above-mentioned problems.

[0008] According to an embodiment of the present invention, a method for associating the trajectories of ocean-going vessels is provided, comprising:

[0009] S1. After preprocessing the images of ocean ships extracted using OpenCV, edge detection and contour processing are used to obtain the size features of the ocean ships.

[0010] S2. Extract the centroid position of the ocean vessel's outline based on the six-parameter affine transformation matrix of GDAL, and extract the target's latitude and longitude coordinates based on the centroid position.

[0011] S3. Obtain the physical attributes, geographic semantic features, and dynamic behavior features of ocean ships, and perform weighted fusion of the physical attributes, geographic semantic features, and dynamic behavior features to select the N targets with the highest probability as a candidate set.

[0012] S4. Calculate and predict the target position after obtaining the heading and speed of the ocean vessel based on dynamic behavior characteristics;

[0013] S5. Match the latitude and longitude coordinates of the target in the candidate set with the predicted target location, select the predicted location with the smallest distance through Euclidean distance calculation as the associated point, generate a continuous trajectory and form target situation information.

[0014] According to an embodiment of the present invention, a device for associating the trajectory of ocean-going ships is provided, comprising:

[0015] The size feature acquisition module uses OpenCV to extract images of ocean ships, performs preprocessing, and then uses edge detection and contour processing to obtain the size features of the ocean ships.

[0016] The latitude and longitude coordinate acquisition module extracts the centroid position of the outline of the ocean vessel based on the six-parameter affine transformation matrix of GDAL, and extracts the latitude and longitude coordinates of the target based on the centroid position of the outline.

[0017] The candidate set module acquires the physical attributes, geographic semantic features, and dynamic behavior features of ocean ships, and performs weighted fusion of the physical attributes, geographic semantic features, and dynamic behavior features to select the N targets with the highest probability as the candidate set.

[0018] The target location acquisition module calculates the predicted target location after acquiring the heading and speed of the ocean vessel based on dynamic behavioral characteristics.

[0019] The target situation acquisition module matches the latitude and longitude coordinates of the target in the candidate set with the predicted target location. It selects the predicted location with the smallest distance through Euclidean distance calculation as the associated point, generates a continuous trajectory, and forms target situation information.

[0020] According to an embodiment of the present invention, an electronic device is provided, comprising:

[0021] Processor; and,

[0022] A memory is configured to store computer-executable instructions, which, when executed, cause the processor to perform the steps of the method embodiments described above.

[0023] According to an embodiment of the present invention, a storage medium is provided for storing computer-executable instructions, which, when executed, implement the steps of the method embodiments described above.

[0024] This invention employs multimodal target feature extraction, combined with a multi-dimensional weighted verification mechanism integrating physical attributes, geographic semantics, and dynamic behavior, to accurately screen key target features. By integrating historical activity heatmaps with real-time speed and heading analysis, interference from single-dimensional misjudgments is avoided, significantly reducing the misjudgment rate of ship types. Compared to traditional methods, the misjudgment rate can be reduced by more than 30%. By predicting the target's displacement position during the reconnaissance delay period using dynamic behavioral features, accurate correlation between target identification and trajectory is achieved, effectively solving the trajectory breakage problem caused by data delay and forming a continuous situational awareness of the target. Compared to existing technologies, trajectory continuity is improved by more than 50%, providing more complete and accurate situational information for marine target monitoring, contributing to improved real-time performance and reliability of marine monitoring, and possessing significant application value in maritime safety, fisheries management, and military reconnaissance. Attached Figure Description

[0025] To more clearly illustrate the technical solutions in one or more embodiments of this specification or in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0026] Figure 1 This is a flowchart of the ocean vessel trajectory association method according to an embodiment of the present invention;

[0027] Figure 2 This is a schematic diagram of a marine ship trajectory association device according to an embodiment of the present invention. Detailed Implementation

[0028] To enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the technical solutions in one or more embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this specification, and not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of this document.

[0029] Method Implementation Examples

[0030] According to embodiments of the present invention, a method for associating the trajectories of ocean-going vessels is provided. Figure 1 This is a flowchart of the marine ship trajectory association method according to an embodiment of the present invention. Figure 1 As shown, the method for associating the trajectory of ocean-going ships in this embodiment of the invention specifically includes:

[0031] S1. After preprocessing the images of ocean ships extracted using OpenCV, edge detection and contour processing are used to obtain the size features of the ocean ships.

[0032] Specifically, S1 includes:

[0033] S11. Use OpenCV to preprocess ocean ship images, including: using median filtering to reduce wave noise and using contrast enhancement algorithms to improve the contrast between the ship and the background.

[0034] OpenCV was used to preprocess the TIFF image, including median filtering for noise reduction to minimize the impact of wave noise. A contrast enhancement algorithm was employed to improve the contrast between the ship and the background, mitigating the blurring caused by low resolution. Median filtering was performed using a 3×3 convolution kernel, calculated as follows:

[0035] g(i,j)=median{f(ik,jl),(k,l∈[-1,1] 2 )};

[0036] Regarding the characteristics of ocean wave noise, when the local variance σ 2 When the value is >25, it will automatically upgrade to 5×5 cores to enhance noise reduction.

[0037] Contrast enhancement employs Adaptive Histogram Equalization (CLAHE), dividing the image into 8×8 sub-blocks, with the grayscale mapping function for each sub-block being:

[0038]

[0039] Where, N b n represents the total number of gray levels in the sub-block. i Where i is the number of pixels with grayscale value i, and N is the total number of pixels in the sub-block. In this invention, the cropping limit is set to 0.01 in hi to avoid noise amplification.

[0040] S12. Use the Canny edge detection algorithm to extract edge information from the ocean ship image, then use the findContours function to obtain all contours as the first contour set, and filter out interfering contours by setting a contour area threshold to obtain the second contour set.

[0041] S12 specifically includes: using the Canny edge detection algorithm to extract edge information from the image, and then using the findContours function to obtain all contours. By setting a contour area threshold, interference contours with too small an area are filtered out, and possible ship contours are retained.

[0042] Canny edge detection specifically includes:

[0043] The following formula is used for adaptive calculation of dual thresholds:

[0044]

[0045] Where k = 1.5, the gradient magnitude Calculated using the Sobel operator.

[0046] Pixel gradient greater than T high Time is defined as the edge, between T low and T high It is only retained when it is connected to a high threshold edge.

[0047] Contour extraction and filtering specifically include:

[0048] Use the findContours function to obtain the contour, with the area threshold set to S. min =500 pixels, the calculation formula is:

[0049] Preserve outline = {C|area(C)≥S} min};

[0050] Perform convex hull defect detection on the preserved contour and repair the broken edges.

[0051] S13. For each contour in the second contour set, calculate its bounding rectangle. The length and width of the bounding rectangle are the length and width features of the ship.

[0052] S13 specifically includes: for each retained contour, calculating its bounding rectangle, the length and width of which represent the ship's length and width features. To improve accuracy, a minimum bounding rectangle algorithm can be used, which better matches the actual shape of the ship.

[0053] The steps for calculating the minimum bounding rectangle (MRER) include: calculating the minimum bounding rectangle for the contour C, with parameters including the center point (x, y), width and height (w, h), and rotation angle θ. The calculation formula is as follows:

[0054]

[0055] The length and width characteristics are:

[0056] Length=max(w,h), Width=min(w,h);

[0057] Rotation angle correction:

[0058] When the rotation angle θ∈(45°,135°), the length and width values ​​are interchanged to ensure that the length is always the major axis:

[0059] ifθ>45°andθ<135°then(w,h)=(h,w).

[0060] S2. Extract the centroid position of the ocean vessel's outline based on the six-parameter affine transformation matrix of GDAL, and extract the target's latitude and longitude coordinates based on the centroid position.

[0061] S21. Obtain image geographic information, specifically including:

[0062] Open a TIFF image using the GDAL library and read its georeferenced information, including coordinate system and geographic transformation parameters. Read the six-parameter affine transformation matrix from GDAL:

[0063] x geo =GT[0]+col·GT[1]+row·GT[2]y geo =GT[3]+col·GT[4]+row·GT[5]\end{cases};

[0064] Where (col,row) are pixel coordinates, GT[0] and GT[3] are the top left corner geographic coordinates, GT[1] and GT[5] are pixel resolutions, and GT[2] and GT[4] are rotation parameters (ideally 0).

[0065] Projected coordinate system transformation:

[0066] If it is a UTM projection, the formula for converting to WGS84 latitude and longitude is:

[0067]

[0068] Where E and N are the UTM eastward and northward pseudo-offsets, k0 = 0.9996 is the scaling factor, and M is the meridian arc length.

[0069] S22. Target location coordinate extraction

[0070] Based on the extracted center of gravity of the ship's outline, and using the geographic transformation function provided by GDAL, the image pixel coordinates are converted into latitude and longitude coordinates to obtain the target's geographical location.

[0071] Contour centroid calculation:

[0072] Geometric centroid: (x i ,y i () represents the pixel coordinates of the outline.

[0073] Weighted center of gravity (considering pixel grayscale):

[0074]

[0075] Where I(x) i ,y i () represents the pixel grayscale value.

[0076] Latitude and longitude conversion:

[0077] Substituting the centroid pixel coordinates (col, row) into the geographic transformation formula, we obtain the geodetic coordinates (lon, lat):

[0078]

[0079]

[0080] S3. Obtain the physical attributes, geographic semantic features, and dynamic behavior features of ocean-going vessels, and perform weighted fusion of these features, selecting the N targets with the highest probabilities as a candidate set; S3 specifically includes:

[0081] S31. Multi-feature acquisition:

[0082] Physical attribute features: Obtain the ship type and related attributes that match the extracted length and width features from the ship-level physical attribute library.

[0083] Ship-level physical attribute library storage format: Ship type: [length range, width range, displacement], such as frigate = [(80m, 120m), (10m, 20m), (2000t, 4000t)];

[0084] The matching degree is calculated as follows:

[0085] Where (l,w) represents the extracted features, (l0,w0) represents the standard values ​​in the database, and σ l =5m,σ w =2m.

[0086] Geographic semantic features: Obtain historical activity distribution data of the target location and analyze the common ship types in the area.

[0087] Historical activity probability calculation:

[0088]

[0089] The area is defined as a circular area with a radius of 50 nautical miles centered on the target location.

[0090] Dynamic behavioral characteristics: Obtain dynamic information such as the target's heading and speed.

[0091] The heading angle θ and speed v are obtained through AIS data or radar measurements; if missing, they are predicted from historical trajectories.

[0092] v t =v t-1 +a·Δt,θ t =θ t-1+ω·Δt, where a is the acceleration and ω is the angular velocity of the turn, based on historical data statistics.

[0093] S32 Weight Setting and Filtering: Establish a weight model and set weights according to the importance of different features. For example, the weight of physical attribute features is set to 0.4, the weight of geographic semantic features is set to 0.3, and the weight of dynamic behavior features is set to 0.3. Through weighted calculation, the most likely target type is filtered out, reducing the false positive rate.

[0094] The weighted fusion model is as follows:

[0095] S = w1·S phys +w2·S geo +w3·S dyan Where w1 = 0.4 (physical attribute), w2 = 0.3 (geographical semantic), w3 = 0.3 (dynamic behavior), S phys S geo S dyan This is the normalized score for a single feature.

[0096] If S≥0.7, it is determined to be the corresponding ship type; otherwise, multiple hypothesis matching is triggered, and the first 3 high probability types are retained.

[0097] S4. Calculate and predict the target position after obtaining the heading and speed of the ocean-going vessel based on dynamic behavioral characteristics; S4 specifically includes:

[0098] Determine the delay time for reporting the target location, for example, assume the delay time is t hours. Delay time t delay Based on communication log statistics, it follows a normal distribution: t delay ~N(μ,σ 2 ), where μ=0.5h, σ=0.1h.

[0099] Based on the target's heading and speed, a uniform linear motion model is used to predict the displacement within a time delay t. The displacement calculation formula is:

[0100] Displacement distance = speed × delay time

[0101] Displacement direction = heading

[0102] Based on the currently identified target location and the predicted displacement, the predicted location at each time point within the delay period is calculated.

[0103] Calculation of Earthly Distance:

[0104] Considering the Earth's curvature, the Haversine formula is used to calculate the latitude and longitude difference corresponding to the displacement:

[0105]

[0106] Where R = 6371 km is the Earth's radius, φ0 is the current latitude, and θ is the heading angle (0° for true north, positive for clockwise).

[0107] Predicted location iterative calculation:

[0108] The predicted position of the i-th time point (i = 1, 2, ..., n):

[0109]

[0110] Where n = t delay ×2 (one point every 0.5 hours).

[0111] S5. Match the latitude and longitude coordinates of the target in the candidate set with the predicted target location. Select the predicted location with the smallest distance using Euclidean distance calculation as the association point, generate a continuous trajectory, and form target situation information. S5 specifically includes:

[0112] The longitude and latitude of the identified unknown target location are matched with the nearest predicted target location. The Euclidean distance is used to calculate the distance between the target location and each predicted location, and the predicted location with the smallest distance is selected as the association point.

[0113] Target location (lon,lat) and predicted location (lon) i ,lat i ) spherical distance:

[0114] Where, Δφ=lat-lat i Δλ=lon-lon i .

[0115] Related point selection: i * =argmin i d i ,like If a match is found, it is confirmed; otherwise, it is considered a new target.

[0116] Starting from the matched predicted position, subsequent predicted positions are sequentially connected with the actual identified target position to form a continuous trajectory.

[0117] Time series interpolation:

[0118] Linear interpolation is performed between predicted and actual points within the delay period to generate dense trajectory points:

[0119]

[0120] Among them, t i ≤t≤t i+1 .

[0121] Based on the associated trajectory, combined with information such as the target's type, heading, and speed, continuous situational information of the target is generated and displayed in a visual manner.

[0122] Situational parameter calculation:

[0123] Rate of change of heading angle:

[0124] Acceleration:

[0125] Situational Threat Index: Where w θ =0.6, w v =0.4.

[0126] This invention employs multimodal target feature extraction, combined with a multi-dimensional weighted verification mechanism integrating physical attributes, geographic semantics, and dynamic behavior, to accurately screen key target features. By integrating historical activity heatmaps with real-time speed and heading analysis, interference from single-dimensional misjudgments is avoided, significantly reducing the misjudgment rate of ship types. Compared to traditional methods, the misjudgment rate can be reduced by more than 30%. By predicting the target's displacement position during the reconnaissance delay period using dynamic behavioral features, accurate correlation between target identification and trajectory is achieved, effectively solving the trajectory breakage problem caused by data delay and forming a continuous situational awareness of the target. Compared to existing technologies, trajectory continuity is improved by more than 50%, providing more complete and accurate situational information for marine target monitoring, contributing to improved real-time performance and reliability of marine monitoring, and possessing significant application value in maritime safety, fisheries management, and military reconnaissance.

[0127] Device Example 1

[0128] According to an embodiment of the present invention, a device for associating the trajectory of ocean-going ships is provided. Figure 2 This is a schematic diagram of a marine ship trajectory correlation device according to an embodiment of the present invention. Figure 2 As shown, the marine ship trajectory association device of this invention specifically includes:

[0129] The size feature acquisition module uses OpenCV to extract images of ocean ships, performs preprocessing, and then uses edge detection and contour processing to obtain the size features of the ocean ships.

[0130] The latitude and longitude coordinate acquisition module extracts the centroid position of the outline of the ocean vessel based on the six-parameter affine transformation matrix of GDAL, and extracts the latitude and longitude coordinates of the target based on the centroid position of the outline.

[0131] The candidate set module acquires the physical attributes, geographic semantic features, and dynamic behavior features of ocean ships, and performs weighted fusion of the physical attributes, geographic semantic features, and dynamic behavior features to select the N targets with the highest probability as the candidate set.

[0132] The target location acquisition module calculates the predicted target location after acquiring the heading and speed of the ocean vessel based on dynamic behavioral characteristics.

[0133] The target situation acquisition module matches the latitude and longitude coordinates of the target in the candidate set with the predicted target location. It selects the predicted location with the smallest distance through Euclidean distance calculation as the associated point, generates a continuous trajectory, and forms target situation information.

[0134] According to an embodiment of the present invention, a method is provided.

[0135] Device Example 2

[0136] According to an embodiment of the present invention, an electronic device is provided, comprising:

[0137] Processor; and,

[0138] A memory is configured to store computer-executable instructions, which, when executed, cause the processor to perform the steps of the above-described method for associating ocean vessel trajectories.

[0139] Device Example 3

[0140] According to an embodiment of the present invention, a storage medium is provided for storing computer-executable instructions, which, when executed, implement the steps of the above-described method for associating ocean vessel trajectories.

[0141] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for associating the trajectories of ocean-going ships, characterized in that... include: S1. After preprocessing the TIFF images of ocean ships extracted using OpenCV, edge detection and contour processing are used to obtain the size features of the ocean ships. S2. Extract the centroid position of the ocean vessel's outline based on the six-parameter affine transformation matrix of GDAL, and extract the target's latitude and longitude coordinates based on the centroid position. S3. Obtain the physical attributes, geographic semantic features, and dynamic behavior features of ocean ships, calculate the single feature scores of the physical attributes, geographic semantic features, and dynamic behavior features, normalize the single feature scores to obtain normalized single feature scores, perform weighted fusion on the normalized single feature scores, and take the N targets with the highest weighted fusion scores as the candidate set. S4. Calculate and predict the target position after obtaining the heading and speed of the ocean vessel based on dynamic behavior characteristics; S4 specifically includes: Determine the delay time for target position reporting, and based on the target's heading and speed, use a uniform linear motion model to predict the displacement within the delay time t. Based on the currently identified target position and the predicted displacement, calculate the predicted position at each time point within the delay period. S5. Match the latitude and longitude coordinates of the target in the candidate set with the predicted target location, select the predicted location with the smallest distance through Euclidean distance calculation as the association point, generate a continuous trajectory and form target situation information; The specific steps of matching the latitude and longitude coordinates of the target in the candidate set with the predicted target location include: matching the longitude and latitude of the identified unknown target location with the nearest predicted target location.

2. The method according to claim 1, characterized in that, S1 specifically includes: Preprocessing of ocean ship images using OpenCV includes: using median filtering to reduce wave noise and using a contrast enhancement algorithm to improve the contrast between the ship and the background. The edge information of ocean ship images is extracted using the Canny edge detection algorithm. Then, the findContours function is used to obtain all contours as the first contour set. By setting a contour area threshold, interfering contours are filtered to obtain the second contour set. For each contour in the second contour set, calculate its bounding rectangle. The length and width of the bounding rectangle are the length and width features of the ship.

3. The method according to claim 2, characterized in that, The median filtering denoising uses a 3×3 convolution kernel for median filtering. When the local variance is greater than a preset value, a 5×5 convolution kernel is used for median filtering. The contrast enhancement algorithm uses adaptive histogram equalization, dividing the image into 8×8 sub-blocks. The grayscale mapping function for each sub-block is: ； in, This represents the total number of gray levels in the sub-block. grayscale value The number of pixels, This represents the total number of pixels in the sub-block.

4. The method according to claim 1, characterized in that, S2 specifically includes: Use the GDAL library to read the georeferenced information of TIFF images. The georeferenced information includes the coordinate system and geographic transformation parameters. Based on the centroid pixel coordinates of the contour, and combined with the six-parameter affine transformation matrix, the pixel coordinates are converted into latitude and longitude coordinates. The centroid calculation of the contour uses the weighted centroid method, and the formula is as follows: ； in, This represents the pixel grayscale value.

5. The method according to claim 1, characterized in that, S3 specifically includes: Match the ship's length and width features to the ship's class physical attribute database and calculate the matching degree: ； in, To extract features, The standard value in the library, ; Obtain historical activity distribution data for the target location and calculate the probability of historical activities: ； Combining dynamic behavioral characteristics, a comprehensive score is calculated using a weighted fusion model: ； in, Weights for physical attributes, geographic semantics, and dynamic behavior.

6. The method according to claim 1, characterized in that, S5 specifically includes: Calculate the spherical distance between the latitude and longitude coordinates of each target in the candidate set and the location of each predicted target. The distance formula is: ； Where R is the Earth's radius; lat and lon are the latitude and longitude of the target to be matched; Let the latitude and longitude be the i-th predicted location; ; By comparing the spherical distances, the predicted position with the smallest distance is selected as the association point. If the smallest distance is less than or equal to the preset distance, the association point is confirmed to match the candidate target; if the smallest distance is greater than the preset distance, the candidate target is regarded as a new target. Starting from the matched associated points, time series linear interpolation is performed on the predicted points and the actual identified target locations within the delay period to generate dense trajectory points; Based on the associated trajectory, combined with information on ship type, heading angle, and speed, situational parameters are calculated. The trajectory and situational parameters are integrated to generate continuous situational information of the target and display it in a visual manner.

7. A device for correlating the trajectory of marine vessels, characterized in that, include: The size feature acquisition module uses OpenCV to extract TIFF images of ocean ships, performs preprocessing, and then uses edge detection and contour processing to obtain the size features of the ocean ships. The latitude and longitude coordinate acquisition module extracts the centroid position of the outline of the ocean vessel based on the six-parameter affine transformation matrix of GDAL, and extracts the latitude and longitude coordinates of the target based on the centroid position of the outline. The candidate set module acquires the physical attributes, geographic semantic features, and dynamic behavior features of ocean ships, calculates the single feature scores of the physical attributes, geographic semantic features, and dynamic behavior features, normalizes the single feature scores to obtain normalized single feature scores, performs weighted fusion on the normalized single feature scores, and selects the N targets with the highest weighted fusion scores as the candidate set. The target position prediction module calculates the predicted target position after obtaining the heading and speed of the ocean-going vessel based on dynamic behavioral characteristics; specifically, the target position prediction module is used for: Determine the delay time for target position reporting, and based on the target's heading and speed, use a uniform linear motion model to predict the displacement within the delay time t. Based on the currently identified target position and the predicted displacement, calculate the predicted position at each time point within the delay period. The target situation acquisition module matches the latitude and longitude coordinates of the target in the candidate set with the predicted target location, selects the predicted location with the smallest distance through Euclidean distance calculation as the association point, generates a continuous trajectory and forms target situation information; The specific steps of matching the latitude and longitude coordinates of the target in the candidate set with the predicted target location include: matching the longitude and latitude of the identified unknown target location with the nearest predicted target location.

8. An electronic device, comprising: processor; as well as, A memory configured to store computer-executable instructions, which, when executed, cause the processor to perform the steps of the ocean vessel trajectory association method as described in any one of claims 1-6.

9. A storage medium for storing computer-executable instructions, which, when executed, implement the steps of the marine ship trajectory association method as described in any one of claims 1-6.

Images