A ship track association fusion method

Abstract

The present application relates to a kind of ship track association fusion method, belong to ship track fusion technical field.The present application is first obtained ship track data, and it is preprocessed;Then based on nearest neighbor distance method, the same source track data is associated with interrupted track, and the different source track data is associated;Finally, the associated interrupted track pair is spliced and fused, and the associated different source track pair is weighted and fused.In order to facilitate visual analysis, the initial track data, the track data after preprocessing, the interrupted track association relationship, the time relationship of the track data of the interrupted track association relationship are visually displayed, and the fusion association result is checked and adjusted in real time through interactive interface, to further achieve the purpose of improving the efficiency and accuracy of association fusion.

Description

Technical Field

[0001] This invention relates to a method for associating and fusing ship tracks, belonging to the field of ship track fusion technology. Background Technology

[0002] Radar sensors can be used to detect ships at sea and provide their position and attribute information within their detection range. However, due to external or internal factors such as target obstruction and limited sensor detection probability during data collection, omissions can occur when updating ship information, leading to track interruptions. These interruptions not only increase the scale and difficulty of data management but also cause discrepancies between the observed and actual number of targets. Multi-source observation refers to collaborative observation among sensors of the same or different types within the detection range, with multiple sensors simultaneously recording ship tracks in the scene. Under multi-source observation, the track information of the same target may be recorded by multiple sensors. However, due to varying degrees of detection error between sensors, direct fusion analysis of multi-source tracks is difficult. Therefore, to obtain more accurate and comprehensive ship track information, it is necessary to correlate interrupted tracks and multi-source tracks and perform fusion analysis.

[0003] Currently, the main methods for solving the problem of track correlation and fusion are based on traditional data statistical analysis. These methods start with the data, perform statistical analysis on the track data to mine its features, and then use these features to correlate and fuse the tracks. However, current correlation and fusion methods have relatively complex statistical processes, low efficiency, lack real-time performance, and are not very accurate. Summary of the Invention

[0004] The purpose of this invention is to provide a method for ship track association and fusion to solve the problems of low efficiency and poor accuracy in the current ship track association and fusion process.

[0005] To solve the above-mentioned technical problems, the present invention provides a method for ship track association and fusion, which includes the following steps:

[0006] 1) Acquire ship track data and preprocess it, wherein the track data includes track data from different sources; the preprocessing includes track resampling processing to make track data from different sources have the same sampling period;

[0007] 2) Based on the nearest neighbor distance method, perform interrupted track association on track data from the same source and associate track data from different sources;

[0008] 3) Perform splicing and fusion on the associated interrupted track pairs, and perform weighted fusion on the associated track pairs from different sources.

[0009] This invention improves the accuracy of trajectory association by acquiring trajectory data from different sources, using the nearest neighbor distance method to associate interrupted trajectories of trajectory data from the same source, and associating trajectory data from different sources, thereby improving the accuracy of subsequent fusion.

[0010] Furthermore, the process of interrupting track association is as follows:

[0011] Iterate through the trajectory data of the source to be associated and determine whether the time difference between any two trajectory data is greater than the interruption time threshold. The time difference between two trajectory data refers to the time difference between the end time of one trajectory and the start time of another trajectory.

[0012] If it is greater than the threshold, the correlation distance between the two tracks is calculated. The correlation distance between the two tracks refers to the correlation distance between the track point at the end of one track and the track point at the start of another track. The correlation distance is a weighted sum of the spatial Euclidean distance, the heading angle distance, and the speed numerical distance between the two tracks.

[0013] A distance matrix is ​​constructed based on the correlation distance between each pair of tracks, and the correlation result matrix is ​​determined by using the position of the smaller distance value in the distance matrix.

[0014] This invention, when associating interrupted tracks, calculates the distance between two interrupted tracks and selects the track pair with the smaller distance as the associated track. This association method does not require a large amount of statistical data; the association relationship can be determined based on distance. The method is simple, easy to implement, and highly accurate.

[0015] Furthermore, the process of determining the association between different source tracks in step 2) is as follows:

[0016] a. Construct a distance matrix to represent the distance between two tracks from different sources, and initialize the elements in the distance matrix to positive infinity;

[0017] b. Traverse all track data under two different sources. For two tracks under different sources, if the ratio of the larger to the smaller number of sampling points of the track is less than a set ratio, calculate the distance between the two tracks and update it to the corresponding position in the distance matrix.

[0018] c. Search the updated distance matrix to find the position of the minimum value. This position represents the correlation between the tracks of the two different sources. Record this position and set the elements of the row and column containing this position in the matrix to positive infinity.

[0019] d. Repeat step c until all elements in the distance matrix are positive infinity, and the track pairs corresponding to each minimum position are associated tracks.

[0020] Furthermore, the formula used to calculate the correlation distance between two different source tracks is as follows:

[0021]

[0022] Where dis(i,j) represents the distance between tracks i and j from different sources, and dis(1,1) represents the distance between two first tracks from different sources.

[0023] This invention, when performing multi-source track association, calculates the distance between two multi-source tracks and selects the track pair with the smaller distance as the associated track. This association method does not require a large amount of statistical data; the association relationship can be determined based on distance. The method is simple, easy to implement, and highly accurate.

[0024] Furthermore, in step 3), an interpolation-based fitting method is used to stitch together the associated interrupted tracks.

[0025] Furthermore, the weighted fusion process of different source tracks in step 3) is as follows:

[0026] The stability of each source sensor is evaluated based on the stability index of the track sampling point interval and the track noise index, and the track fusion weight of each source is obtained.

[0027] In the time-overlapping region of the multi-source sensors, the multi-source tracks to be associated are weighted and fused according to the fusion weight of each source track; in the time-non-overlapping region of the multi-source sensors, the last few track points of the weighted fused track part are linearly extrapolated, and the last few track points and the linearly extrapolated track points are weighted and fused.

[0028] The present invention fully considers the stability of sampling and the impact of noise in the fusion weight determined by the stability index of track sampling point interval and the track noise index. In view of the difference in detection range and detection probability of different sensors, the track of the same target under different sensors is difficult to completely overlap in time. The fusion is divided into time overlap region and time non-overlap region, which solves the problem of incomplete time overlap of multi-source track.

[0029] Furthermore, the stability index of the track sampling point interval uses the variance of the track sampling point interval, and the track noise index uses the ratio of track distance before and after track smoothing as the evaluation standard.

[0030] Furthermore, this method visualizes the temporal relationships between the initial track data, the preprocessed track data, the interrupted track correlations, and the track data with interrupted track correlations.

[0031] This invention visualizes the temporal relationship between initial track data, preprocessed track data, interrupted track association, and track data with interrupted track association, allowing people to participate in the track association and fusion process.

[0032] Furthermore, the method also includes human-computer interactive management of the ship track association and fusion process, including track data management, interrupted track association and fusion processing, and multi-source track association and fusion processing;

[0033] Track data management is used to specify the latitude and longitude fields, sensor source fields, and batch number fields of the imported data. Based on the specified fields, data grouping and scene track positioning are realized. Finally, the scene track is rendered to the map window, and the current scene is added to the layer list on the left. The display color of the track can be changed through the layer list.

[0034] Interrupted track association and fusion processing is used to select the source of information based on the selected layer to be processed, specify the name of the new layer, display the association and fusion process of the interrupted track, and store the processed track as a new layer to display the association and fusion result;

[0035] Multi-source track association and fusion processing is used to select the layer to be processed, specify the name of the new layer, display the multi-source track association and fusion process, and store the processed track as a new layer to display the association and fusion result.

[0036] This invention enables people to participate in the trajectory association and fusion process through human-computer interaction, and allows them to check and adjust the results in real time through an interactive interface, thereby further improving the efficiency and accuracy of the association and fusion. Attached Figure Description

[0037] Figure 1 This is a flowchart of the ship track association and fusion method of the present invention;

[0038] Figure 2 This is a schematic diagram of the track resampling process in the ship track association and fusion method of the present invention;

[0039] Figure 3 This is a schematic diagram of the weighted moving average method in the ship track correlation and fusion method of the present invention;

[0040] Figure 4 This is a flowchart of the distance matrix update process of the interrupted track association algorithm used in the ship track association fusion method of the present invention;

[0041] Figure 5 This is a flowchart illustrating the distance matrix update process using the multi-source track association algorithm in the ship track association and fusion method of the present invention.

[0042] Figure 6 This is a schematic diagram of the shared and unique parts of track fusion in an embodiment of the present invention;

[0043] Figure 7 This is a schematic diagram illustrating the fusion of weighted fused tracks and single-sensor tracks in an embodiment of the present invention;

[0044] Figure 8 These are the relationship network diagram and relationship Gantt diagram in the embodiments of the present invention;

[0045] Figure 9 This is a system interface diagram of the trajectory data association and fusion system in an embodiment of the present invention;

[0046] Figure 10 This is a schematic diagram of the track data management function in an embodiment of the present invention;

[0047] Figure 11 This is a schematic diagram of the track data anomaly handling function in an embodiment of the present invention;

[0048] Figure 12 This is a schematic diagram of the track filtering processing function in an embodiment of the present invention;

[0049] Figure 13 This is a schematic diagram of the track resampling processing function in an embodiment of the present invention;

[0050] Figure 14 This is a schematic diagram of the trajectory smoothing function in an embodiment of the present invention;

[0051] Figure 15 This is a schematic diagram illustrating the interrupted track association and fusion processing function in an embodiment of the present invention;

[0052] Figure 16 This is a schematic diagram of the multi-source track association and fusion processing function in an embodiment of the present invention. Detailed Implementation

[0053] The specific embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0054] This invention first acquires and preprocesses ship track data; then, it uses the nearest neighbor distance method to correlate interrupted tracks from the same source and correlates tracks from different sources; finally, it splices and fuses the correlated interrupted track pairs and performs weighted fusion on the correlated tracks from different sources. To facilitate visualization and analysis, the initial track data, preprocessed track data, interrupted track correlation relationships, and the temporal relationships of track data within these correlations are visualized. An interactive interface allows for real-time checking and adjustment of the fusion correlation results, further improving the efficiency and accuracy of the correlation fusion. The implementation process of this method is as follows: Figure 1 As shown below, a detailed explanation will follow.

[0055] 1. Acquire track data and preprocess it.

[0056] The track data acquired in this embodiment includes multi-source data, i.e., data collected by different sensors. The preprocessing process includes outlier handling, track filtering, track resampling, and track smoothing. Each process is described below.

[0057] 1) Outlier handling

[0058] Track data anomalies refer to records where certain attribute values ​​in the attribute or location information received by sensors deviate significantly from other values ​​due to equipment failure, system errors, or other reasons. This includes attribute value anomalies and track outliers. For attribute anomalies, this invention uses a box plot method for detection. Considering the relative stability of the ship's navigation state, the mean or mode of non-outlier values ​​can be used to fill in the gaps. The formula for detecting track data attribute anomalies using the box plot method is as follows:

[0059] IQR = Q 0.75 -Q 0.25

[0060] threshold = Q 0.75 +1.5*IQR

[0061]

[0062] Q 0.25 Q 0.75 These are the first and last quartiles of the current track speed value, respectively.

[0063] For isolated points in the track, we use the difference method and box plot method to filter out outliers and delete them; the method is as follows:

[0064] By performing a difference operation on the coordinate information of waypoints, the coordinate difference components between adjacent waypoints are obtained. The sum of these coordinate difference components is then used to determine the distance between adjacent waypoints. The formula is as follows:

[0065] lat_diff=lat next -lat pre

[0066] lon_diff = lon next -lon pre

[0067] lat pre and lon pre These are the latitude and longitude of the previous flight path data, respectively. next and lon nextThese are the latitude and longitude of the next flight path data, respectively.

[0068] Outliers in the distance between adjacent waypoints are detected using box plot detection, and the detected waypoints are deleted.

[0069] 2) Track filtering processing

[0070] This step primarily addresses the filtering of single-point tracks and excessively short tracks. A single-point track refers to a track consisting of only one track point due to accidental factors in the equipment's detection environment; it lacks inherent track descriptive capabilities. An excessively short track refers to a track with an extremely short lifespan (track lifespan is the time interval between the end and start of track recording). Data from excessively short tracks contains significant noise and errors, reducing the accuracy of track matching and increasing the complexity of subsequent track association.

[0071] 3) Track resampling: Track data generated by radar scanning has a fixed scanning period, but some track data from other sources (such as AIS data) do not have similar periodic characteristics. Therefore, this invention needs to use a resampling method to make track data from different sources have the same periodic characteristics. In this embodiment, track resampling is performed based on linear interpolation, such as... Figure 2 As shown, the process is as follows:

[0072] Based on the recording time and other information of the track points in the original track data, the original track data is continuously and smoothly fitted; according to the set sampling period, a new time sequence with equal intervals is generated according to the start time, end time and sampling period of the track data; the generated time sequence is substituted into the fitting function to obtain the track point information corresponding to the new time, and the resampling is completed.

[0073] 4) Track smoothing

[0074] Because the original track data contains noise and data bias, visualizing the original track directly through simple line connections would distort its shape. Therefore, smoothing is necessary. This embodiment uses a weighted moving average method to smooth the original track. Figure 3 As shown, the process of this method is as follows:

[0075] A fixed-length window is set as a sliding window, which is gradually slid along the starting point to the ending point of the track. All track data points within the window are recorded at each position. Weighting coefficients are assigned to the data points in the window. This invention adopts a weighting coefficient allocation strategy that follows a standard normal distribution, where data points closer to the center of the window are assigned larger weights. The data points within each window are weighted and averaged based on the assigned weights to obtain the weighted moving average of each window, and then the weighted moving average of the entire track is obtained, thus achieving smoothing of the track data.

[0076] 2. Use the preprocessed track data to perform track association processing.

[0077] This invention mainly targets the correlation of interrupted tracks and multi-source tracks. Interrupted tracks refer to the situation where tracks exist in the same source (collected by a single sensor). The two types of correlation are explained below.

[0078] 1) Disruption of track association

[0079] For interrupted tracks, this invention employs an association method based on improved nearest neighbor distance. This method iterates through the track data of the source to be associated, determines if the time difference between any two track data is greater than an interruption time threshold, and if so, calculates the association distance between the two tracks. A distance matrix is ​​constructed based on the association distances between each pair of tracks, and the position of the smaller distance value in this distance matrix is ​​used to determine the association result matrix. Figure 4 As shown below, the association process will be described in detail.

[0080] a. Construct a distance matrix D of size N×N, initialize the elements of matrix D to positive infinity inf, where N is the number of existing and recorded tracks under the current sensor, and set the interruption time threshold T; construct an association matrix A of size N×N, initialize the elements of the matrix to 0;

[0081] b. Iterate through the tracks recorded by the sensors to be associated, setting the current track index as i. Then iterate through the tracks recorded by the sensors to be associated, setting the current track index as j. If the time between the end time of track i and the start time of track j is greater than the set time threshold T, then calculate the association distance d between track i and track j. i，j Update the value at position (i, j) in the distance matrix D to d. i，j :

[0082]

[0083]

[0084]

[0085]

[0086]

[0087]

[0088]

[0089] Among them lon i (e), lat i (e), cou i (e), vel i (e) represents the longitude, latitude, heading, and speed of the last track point after track i is sorted in ascending order by time; lon j (s), lat j (s), cou j (s), vel j (s) represent the longitude, latitude, heading, and speed of the first track point after track j is sorted in ascending order by time; These represent the spatial Euclidean distance, heading angle distance, and speed numerical distance between track i and track j, respectively. α, β, and γ are the normalized spatial Euclidean distance, heading angle distance, and speed numerical distance for track i and track j, respectively; α, β, and γ are the weights assigned to the spatial Euclidean distance, heading angle distance, and speed numerical distance, respectively.

[0090] c. Take the position (i, j) of the minimum value in D, update the corresponding value in A to 1, and update the element in the i-th row and j-th column of D to inf;

[0091] d. Repeat step c until all elements in matrix D are inf. The resulting A is the correlation result matrix of pairwise correlation of the interrupted tracks under the current sensor.

[0092] 2) Multi-source track association

[0093] This implementation uses dual-source track association as an example, where track data is acquired from two different types of sensors. This invention sets a proportional threshold, allowing only track pairs that pass the threshold check to perform subsequent association distance calculations. A dynamic time planning algorithm is used to adjust the correspondence between sampling points. Figure 5 As shown.

[0094] ① For the two input sequences s1 and s2, initialize a cumulative distance matrix D of size N1×N2, where N1 is the length of sequence s1 and N2 is the length of sequence s2;

[0095] ② Traverse all points in sequence s1, set the index of the current point to i, and calculate the cumulative distance d between the current point and the first point of sequence s2. i，1 Update the value at position (i, 1) in D to d i，1 ;

[0096] ③ Traverse all points in sequence s2, set the index of the current point to j, and calculate the cumulative distance d between the current point and the first point of sequence s1. i，j Update the value at position (1, j) in D to d i，j ;

[0097] ④ Traverse all points in sequence s1, set the index of the current point to i, traverse all points in sequence s2, set the index of the current point to j, and calculate the cumulative distance d between the i-th point in sequence s1 and the j-th point in s2. i，j Update the value at position (i, j) in D to d i，j :

[0098] dis(i, j) = |lon i -lon j |+|lat i -lat j |

[0099]

[0100] Where dis(i, j) is the distance between the i-th point of sequence s1 and the j-th point of sequence s2.

[0101] The basic steps of threshold checking are as follows:

[0102] ① Construct a distance matrix D of size N1×N2, and initialize the elements in the matrix to positive infinity inf, where N1 is the number of tracks recorded by sensor 1 and N2 is the number of tracks recorded by sensor 2;

[0103] ② Construct an N×N correlation matrix A, and initialize the elements of the matrix to 0;

[0104] ③ Iterate through all tracks under sensor 1, with the current track index being i. Then iterate through all tracks under sensor 2, with the current track index being j. If the ratio of the larger to the smaller of the number of sampling points of track i and track j is less than a set threshold, then a dynamic time warping algorithm is used to determine the correlation distance d between track i and track j. i，j Calculate and update the value at position (i, j) in D to d. i，j ;

[0105] ④ Take the position (i, j) of the minimum value in D, set the corresponding position in A to 1, and set the element in the i-th row and j-th column of D to inf;

[0106] ⑤ Repeat step ④ until all elements in D are inf, and the resulting A is the pairwise correlation result of the multi-source tracks.

[0107] 3. Perform track stitching and fusion processing based on the correlation results.

[0108] For splicing interrupted flight tracks, this invention uses an interpolation-based fitting method to splice interrupted flight tracks.

[0109] The algorithm flow is as follows:

[0110] ① For the input associated track pairs, select the last 10 sampling points of the old track and the first 10 sampling points of the new track, and use the linear interpolation method to fit each attribute of the track based on these 20 sampling points.

[0111] ② Generate the sampling time of the interrupted part based on the sampling period of the track, and substitute the generated time into the fitting function corresponding to each attribute to obtain the track attribute value at the corresponding time.

[0112] ③ After merging the associated track pairs with the generated interrupted track, the sampling points are arranged in chronological order to obtain the complete track of the target ship.

[0113] For multi-source track fusion, this invention employs a sensor stability assessment-based method to determine the weights of associated tracks, and then uses a linear extrapolation weighted splicing method to achieve weighted fusion. The sensor stability assessment-based method for determining track fusion weights evaluates sensor stability from two perspectives: track sampling point interval stability assessment and track noise assessment. The track sampling point interval stability uses the variance of the track sampling point interval as the evaluation criterion, and the calculation formula is as follows:

[0114]

[0115]

[0116]

[0117] in Let be the distance between the i-th sampling point and the (i-1)-th sampling point of the l-th track of the information source s. The score for the stability assessment of the l-th track of source s with respect to the track sampling point interval is given. The score is the normalized score of the l-th track of source s after evaluating the stability of the track sampling point interval.

[0118] Track noise assessment uses the ratio of track distance before and after track smoothing as the evaluation criterion, and the final track weight is determined by the following formula:

[0119]

[0120]

[0121] in Let be the normalized score of the l-th track from source s regarding track noise assessment, and α and β be the weights of the track sampling point interval stability assessment and track noise assessment, respectively. Let the weight of the l-th track of source s be , The final weight after normalization of the l-th track of source s.

[0122] The linear extrapolation weighted stitching method aims to address the problem of incomplete temporal overlap between multi-source tracks. Due to differences in the detection range and probability of different sensors, it is difficult for the tracks of the same target to completely overlap temporally across different sensors. Figure 6 As shown, the process of multi-source track fusion involves fusing the common parts of the two sensors (referred to as weighted fused tracks) and the parts unique to a single sensor (referred to as single-sensor tracks). The specific processing method is illustrated in the following example:

[0123] Assuming two tracks are to be fused, they are divided into a weighted fused track component and a single-sensor track component. To achieve reasonable fusion between the weighted fused track and the single-sensor track, normal weighted fusion is performed directly in the time-overlapping region (except for the last n points); in the time-dislapping region, to achieve a reasonable transition, the last n track points of the weighted fused track component are linearly extrapolated. The weights of the last n track points of the weighted fused track and the extrapolated n track points are set sequentially from the inside out as (2n-1) / (2n-1), (2n-2) / (2n-1), (2n-3) / (2n-1)...0 / (2n-1). Then, based on these 2n points, the weights of the 2n track points corresponding to the single sensor track at the given time are set sequentially from the outside in as 0 / (2n-1)...(2n-3) / (2n-1), (2n-2) / (2n-1), (2n-1) / (2n-1), thereby achieving weighted fusion between tracks. The value of n can be determined according to the actual situation; in this embodiment, n equals 3. Figure 7 As shown, the last three track points of the weighted fusion track are linearly extrapolated. The weights of the last three track points of the weighted fusion track and the extrapolated three track points are set to 5 / 5, 4 / 5, 3 / 5, 2 / 5, 1 / 5, and 0 / 5 from the inside out. Then, based on these six points, the weights of the six track points corresponding to the single sensor track at the same time are set to 0 / 5, 1 / 5, 2 / 5, 3 / 5, 4 / 5, and 5 / 5 from the outside in, thereby realizing the weighted fusion between tracks.

[0124] 4. Perform visual mapping.

[0125] This invention can visualize and map various types of data that appear during the trajectory association and fusion process. The visualization in this embodiment includes:

[0126] Visualization of initial track data: Map track data to corresponding points on a map and use different colors to display tracks from different sources;

[0127] The visualization of the processed track data maps the track data to lines at corresponding locations on a map, and uses different colors to display tracks from different information sources;

[0128] The relationship between different batch numbers under the same information source is as follows: Figure 8 As shown, a ring network diagram is used to map each batch number to a node in the network. The size of the node represents the survival time of the track. The relationship between the nodes is represented by lines with arrows inside the network.

[0129] The temporal relationship of related batch numbers under the same information source, such as... Figure 8 As shown, a Gantt chart is used, with the horizontal axis representing time and the vertical axis representing the batch number. The horizontal rectangles on the chart represent the track survival time of that batch number, and the color of the rectangles distinguishes the main track from the associated track.

[0130] 5. Human-computer interaction management.

[0131] This invention allows for the entire trajectory association and fusion process to be controlled through human-computer interaction, increasing human participation and enabling subjective adjustment of the association and fusion results. For example... Figure 9 As shown, the interactive and interconnected fusion system is divided into a top function option, a left-side layer list, a central map window, and a right-side auxiliary panel, specifically including:

[0132] Track data management, such as Figure 10 As shown, clicking the "Import Scene" button allows you to select a local file and specify the latitude and longitude fields, sensor source fields, and batch number fields for the imported data. Based on these fields, the system groups the data and locates the scene's trajectory. Finally, the scene trajectory is rendered onto the map window, and the current scene is added to the left-hand layer list. The trajectory's display color can be changed through the layer list. Simultaneously, by selecting a source of interest in the layer list, a relational network diagram and a relational Gantt chart of the unprocessed raw trajectory data are automatically generated in the right-hand auxiliary panel to assist the user in initial decision-making.

[0133] Track outlier handling, such as Figure 11As shown, click the "Outlier Removal" button, select the layer for outlier processing (e.g., Scene 101), and specify the name of the new layer (e.g., 101_delsingle). Next, outlier sampling points that deviate from the overall track will be removed, and attribute anomalies in the track data will be filled in. The processed track will be stored as a new layer, and the view will be updated in the central map window for easy comparison of the tracks before and after deletion.

[0134] Track filtering, such as Figure 12 As shown, click the "Filter" button, specify the layer to be filtered (e.g., 101_delsingle), set the minimum number of track sampling points n (e.g., 10), and specify the name of the new layer (e.g., 101_gt10). Tracks with fewer than n sampling points in the scene will then be filtered, retaining tracks with longer lifespans for subsequent association processes. The processed tracks will be stored as a new layer, and the view will be updated in the central map window for easy comparison of tracks before and after filtering. If the user is not satisfied with the results, they can choose to adjust the minimum number of track sampling points n and repeat the filtering process until the optimal result is selected.

[0135] Track resampling processing, such as Figure 13 As shown, click the "Resample" button, specify the layer to be resampled (e.g., 101_gt10), set the sampling period T (e.g., 20s), and specify the name of the new layer (e.g., 101_resample). Then, the track sampling points in this scene will be resampled to achieve temporal alignment. The processed track will be stored as a new layer, and the view will be updated in the central map window for easy comparison of the track before and after resampling. If the user is not satisfied with the processing result, they can choose to adjust the sampling period T and repeat the resampling process until the optimal result is selected.

[0136] Track smoothing, such as Figure 14 As shown, click the "Smooth" button, specify the layer to be smoothed (e.g., 101_resample), set the window size S (e.g., 20), and specify the name of the new layer (e.g., 101_smooth). Then, the track noise in the scene will be effectively eliminated. The processed track will be stored as a new layer, and the view will be updated in the central map window for easy comparison of the track before and after smoothing. If the user is not satisfied with the result, they can choose to adjust the window size S and repeat the track smoothing process until the optimal result is found.

[0137] Interrupted track association and fusion processing, such as Figure 15As shown, click the "Interrupt Association" button, select the layer to be processed for interrupted track association fusion (e.g., 101_smooth), select the source, and specify the name of the new layer (e.g., 8_interrupt). Next, the tracks under this source in the scene will undergo interrupted track association fusion processing, and the tracks under each sensor in the scene will be fully presented. The processed tracks will be stored as a new layer, and the view will be updated in the central map window. The association fusion result will be displayed in the right auxiliary panel. Users can click to highlight the association fusion result, check it, participate in the association fusion process, delete unsatisfactory results, and re-associate and fuse them.

[0138] Multi-source track association and fusion processing, such as Figure 16 As shown, click the "Multi-Source Association" button, select the layer to be processed for multi-source track association and fusion (e.g., 8_interrupt), and specify the name of the new layer (e.g., 8_multi). Next, the tracks from each source in the scene will undergo multi-source track association and fusion processing, and the ship target tracks in the scene will be fully presented. The processed tracks will be stored as a new layer, and the view will be updated in the central map window for easy comparison of tracks before and after processing.

[0139] This invention addresses the problems existing in the track association process of nearest neighbor distance-based methods by proposing an improved solution. Additional constraints are added to the original association algorithm, and the calculation method for association distance is improved. This significantly enhances the efficiency and accuracy of the method in interrupted track association, making the fusion weights more objective and effectively reducing the error rate of multi-source track fusion. Furthermore, this invention incorporates an interactive processing method during the association fusion process, allowing users to participate in the process and check and adjust the results in real time through an interactive interface, thereby improving the efficiency and accuracy of the association fusion.

Claims

1. A ship track association fusion method, characterized in that, The fusion method includes the following steps: 1) Acquire ship track data and preprocess it, wherein the track data includes track data from different sources; the preprocessing includes track resampling processing to make track data from different sources have the same sampling period; 2) Based on the nearest neighbor distance method, interrupted track data from the same source are associated, and track data from different sources are associated. For interrupted tracks, if the time difference between two track data is greater than the interruption time threshold, the association distance between the two tracks is calculated. A distance matrix is ​​constructed based on the association distance between each pair of tracks, and the association result of the interrupted track is determined by the position of the smaller distance value in the distance matrix. For two tracks from different sources, if the ratio of the number of sampling points of the larger track to the smaller track is less than a set ratio, the association distance between the two tracks is calculated by a dynamic time planning algorithm, and the association result of the two tracks from different sources is determined based on the position of the minimum value of the association distance. 3) Interpolation fitting is used to stitch together and fuse the associated interrupted track pairs, and weighted fusion is performed on the associated track pairs from different sources.

2. The ship track association fusion method according to claim 1, characterized in that, The time difference between two track data during the interruption of track association refers to the time difference between the end time of one track and the start time of the other track. The correlation distance between two tracks refers to the correlation distance between the track point at the end of one track and the track point at the start of another track. The correlation distance is a weighted sum of the spatial Euclidean distance, the heading angle distance, and the speed numerical distance between the two tracks.

3. The ship track association and fusion method according to claim 1, characterized in that, The process of determining the association between different source tracks in step 2) is as follows: a. Construct a distance matrix to represent the distance between two tracks from different sources, and initialize the elements in the distance matrix to positive infinity; b. Traverse all track data under two different sources. For two tracks under different sources, if the ratio of the larger to the smaller number of sampling points of the track is less than a set ratio, calculate the distance between the two tracks and update it to the corresponding position in the distance matrix. c. Search the updated distance matrix to find the position of the minimum value. This position represents the correlation between the tracks of the two different sources. Record this position and set the elements of the row and column containing this position in the matrix to positive infinity. d. Repeat step c until all elements in the distance matrix are positive infinity, and the track pairs corresponding to each minimum position are associated tracks.

4. The ship track association and fusion method according to claim 3, characterized in that, The formula used to calculate the correlation distance between two different source tracks is: Where dis(i,j) represents the distance between the i-th point in track s1 and the j-th point in track s2 from different sources, dis(i,1) represents the distance between the i-th point in track s1 and the 1-th point in track s2 from different sources, and dis(1,j) represents the distance between the 1-th point in track s1 and the j-th point in track s2 from different sources. This represents the distance between the first point in track s1 and the first point in track s2.

5. The ship track association and fusion method according to claim 1, characterized in that, The weighted fusion process of different source tracks in step 3) is as follows: The stability of each source sensor is evaluated based on the stability index of the track sampling point interval and the track noise index, and the track fusion weight of each source is obtained. In the time-overlapping region of the multi-source sensors, the multi-source tracks to be associated are weighted and fused according to the fusion weight of each source track; in the time-non-overlapping region of the multi-source sensors, the last few track points of the weighted fused track part are linearly extrapolated, and the last few track points and the linearly extrapolated track points are weighted and fused.

6. The ship track association and fusion method according to claim 5, characterized in that, The stability index of track sampling point interval uses the variance of track sampling point interval, while the track noise index uses the ratio of track distance before and after track smoothing as the evaluation standard.

7. The ship track association and fusion method according to any one of claims 1-6, characterized in that, This method visualizes the temporal relationships between the initial track data, the preprocessed track data, the interrupted track correlations, and the track data with interrupted track correlations.

8. The ship track association and fusion method according to any one of claims 1-6, characterized in that, The method also includes human-computer interactive management of the ship track association and fusion process, including track data management, interrupted track association and fusion processing, and multi-source track association and fusion processing; Track data management is used to specify the latitude and longitude fields, sensor source fields, and batch number fields of the imported data. Based on the specified fields, data grouping and scene track positioning are realized. Finally, the scene track is rendered to the map window, and the current scene is added to the layer list on the left. The display color of the track can be changed through the layer list. Interrupted track association and fusion processing is used to select the source of information based on the selected layer to be processed, specify the name of the new layer, display the association and fusion process of the interrupted track, and store the processed track as a new layer to display the association and fusion result; Multi-source track association and fusion processing is used to select the layer to be processed, specify the name of the new layer, display the multi-source track association and fusion processing process, and store the processed track as a new layer to display the association and fusion results.

9. The ship track association and fusion method according to claim 8, characterized in that, Human-computer interactive management also includes track outlier handling, track filtering, track resampling, and track smoothing.

Images