A ship in and out of port trajectory acquisition method and system based on Beidou positioning

Abstract

The application discloses a ship in and out port track acquisition method and system based on Beidou positioning, relates to the ship in and out port track monitoring technical field, and the method comprises the steps of: obtaining ship navigation data, and pre-processing the ship navigation data to obtain pre-processed ship navigation data, and extracting ship navigation features; the ship navigation features are subjected to feature screening to obtain navigation track features, and ship track prediction is carried out according to the navigation track features to obtain a ship predicted track; a sea area image is generated based on pre-acquired navigation sea area data, and the ship predicted track is divided to obtain a ship sub-track; the ship sub-track is subjected to track mode recognition to obtain a track behavior mode, and the ship sub-track is evaluated, and the ship sub-track is subjected to track correction based on the evaluation result. Through semantic division of the sea area image, the application realizes the transition from geometric track fitting to navigation intention.

Description

Technical Field

[0001] This invention relates to the field of ship entry and exit trajectory monitoring technology, and more specifically, to a method and system for acquiring ship entry and exit trajectories based on BeiDou positioning. Background Technology

[0002] Ship entry and exit from ports is a crucial link in maritime transportation, characterized by narrow channels, dense traffic, and frequent maneuvers, demanding extremely high navigation accuracy and management capabilities. With the widespread application of the BeiDou Navigation Satellite System, its high-precision, highly reliable, and autonomously controllable positioning capabilities provide strong technical support for dynamic ship monitoring. Ship entry and exit trajectory acquisition based on BeiDou positioning not only allows for real-time monitoring of ship position, speed, and heading, but also integrates electronic charts, navigation environment, and behavioral models to achieve intelligent trajectory prediction and semantic analysis. Therefore, constructing a method and system for acquiring ship entry and exit trajectories that integrates BeiDou positioning and artificial intelligence is of great significance for improving port operational efficiency, ensuring navigation safety, and promoting the development of smart shipping and autonomous vessels.

[0003] However, current technologies for acquiring ship entry and exit trajectories generally rely on simple filtering or linear prediction of raw AIS data. This makes it difficult to accurately capture complex nonlinear behaviors of ships in narrow channels, such as sharp turns, deceleration, and frequent changes of direction. This results in large deviations and poor continuity in predicted trajectories. Furthermore, existing methods remain at the level of geometric trajectory fitting, lacking a deep understanding mechanism based on spatial semantics, behavioral patterns, and environmental constraints. This prevents a leap towards recognizing navigation intentions, causing the trajectory correction process to be detached from the actual navigation context and making it difficult to determine the true motives behind abnormal behavior. In addition, trajectory segmentation often uses fixed time or distance thresholds, lacking an understanding of the navigation environment context. This results in mechanical and rigid segmentation results that cannot accurately reflect the behavioral transitions of ships between different functional areas, thus hindering subsequent behavioral analysis and risk warning.

[0004] There are currently no effective solutions to the problems in the relevant technologies. Summary of the Invention

[0005] To address the problems in related technologies, this invention proposes a method and system for obtaining ship entry and exit trajectories based on BeiDou positioning, in order to overcome the aforementioned technical problems existing in the existing related technologies.

[0006] Therefore, the specific technical solution adopted by the present invention is as follows: According to one aspect of the present invention, a method for obtaining ship entry and exit trajectories based on BeiDou positioning is provided, the method comprising the following steps: S1. Acquire ship navigation data, preprocess the ship navigation data to obtain preprocessed ship navigation data, and extract ship navigation features based on the preprocessed ship navigation data. S2. Perform feature filtering on the ship's navigation characteristics to obtain navigation trajectory features, and predict the ship's trajectory based on the navigation trajectory features to obtain the predicted ship trajectory. S3. Generate a sea area profile based on the pre-acquired navigation sea area data, and divide the predicted ship trajectory using the sea area profile to obtain the ship sub-trajectory; S4. Perform trajectory pattern recognition on the ship sub-trajectory to obtain the trajectory behavior pattern, evaluate the ship sub-trajectory based on the trajectory behavior pattern, and correct the ship sub-trajectory based on the evaluation results.

[0007] Furthermore, feature filtering is performed on the ship's navigation characteristics to obtain navigation trajectory features, and ship trajectory prediction is performed based on these features to obtain the predicted ship trajectory, which includes: S21. Use the Pearson coefficient to perform correlation analysis on the ship's navigation characteristics, and based on the correlation analysis results, remove redundant features from the ship's navigation characteristics to obtain the initial navigation trajectory characteristics. S22. Perform a statistical significance test on the initial navigation trajectory features, and filter the initial navigation trajectory features based on the significance test results and a preset threshold to obtain the navigation trajectory features; S23. Based on the characteristics of the navigation trajectory and the preset correction gain coefficient, the trajectory is predicted and corrected to obtain the predicted trajectory of the ship.

[0008] Furthermore, based on the navigation trajectory characteristics and preset correction gain coefficients, trajectory prediction and correction are performed to obtain the ship's predicted trajectory, including: S231. Based on the characteristics of the navigation trajectory, perform preliminary trajectory prediction to obtain preliminary trajectory prediction results; S232. Calculate the trajectory error based on the preliminary trajectory prediction results, and correct the preliminary trajectory prediction results by combining the preset correction gain coefficient to obtain the ship prediction trajectory.

[0009] Furthermore, a sea area profile is generated based on pre-acquired navigation sea area data, and the predicted ship trajectory is divided using the sea area profile to obtain ship sub-trajectories, including: S31. Based on the pre-acquired navigation area data, the navigation environment constraints and navigation environment parameters are extracted using BeiDou positioning technology, and navigation area environment data is constructed based on the extracted navigation environment constraints and navigation environment parameters. S32. Convert the navigation area environmental data into a vector layer, and combine it with the pre-acquired ship morphology parameters to perform navigation adaptability filtering on the vector layer to obtain the navigation area layer. S33. Extract the starting point of the route from the navigation area layer, construct the route network skeleton based on the starting point of the route, perform semantic parsing on the route network skeleton, and generate a sea area image based on the semantic parsing results. S34. Utilize sea area images to perform route spatial topology analysis, and based on the results of the route spatial topology analysis, perform spatial semantic segmentation on the predicted ship trajectory to obtain the ship sub-trajectory.

[0010] Furthermore, the starting points of the shipping routes are extracted from the navigation area layer, and a route network skeleton is constructed based on these starting points. Semantic parsing is then performed on the route network skeleton, and a sea area profile is generated based on the semantic parsing results, including: S331. Extract the starting point of the route based on the navigation area layer, and use the starting point of the route as the initial source of route growth. S332. Set the direction of the channel light source, and use the direction of the channel light source to guide the initial route growth source to extend along the optimal navigation path, so as to obtain the preliminary optimized trunk channel layout. S333. Utilize the apex dominance principle to prioritize the expansion of the main routes in the initially optimized main channel layout to obtain the main route network. Then, extend the branch route network according to the main route network to obtain the route network skeleton. S334. Perform functional area analysis on the route network skeleton, and perform semantic annotation based on the functional area analysis results to generate a sea area profile.

[0011] Furthermore, a route spatial topology analysis is performed using marine imagery. Based on the results of this analysis, the predicted ship trajectories are spatially semantically segmented, resulting in ship sub-trajectories including: S341. Extract the spatial geometry and functional zones of air routes based on the sea area image, and perform air route topology modeling based on the spatial geometry and functional zones of air routes to obtain the air route topology network. S342. Perform route spatial topology analysis using route topology network, set route distance thresholds based on route spatial topology analysis results, and perform spatial proximity analysis on ship predicted trajectories based on route distance thresholds. S343. Based on the spatial proximity analysis results, the initial predicted ship trajectory is divided into preliminary ship sub-trajectories, and the preliminary ship sub-trajectories are spatially semantically labeled using the route functional area to obtain the ship sub-trajectories.

[0012] Furthermore, the expression for calculating the route distance threshold is as follows: ; In the formula, t Indicates the route distance threshold; e Represents the set of edges in the route topology network; N Indicates the total number of route segments; eij Indicates the connection node i and j The route segment; d ij Indicates the length of the route segment; m d This represents the average length of all flight segments; c Indicates the first weighting coefficient; s (·) represents an sigmoid function; C env ( e ij This represents the overall marine environmental cost of the voyage segment; This represents the global average marine environmental cost; l Indicates the second weighting coefficient; Δ i ij Indicates the connection node i Outbound heading angle and connecting nodes j The change in heading between the incoming heading angles; β This represents the third weighting coefficient; f ij Indicates the frequency of vessel traffic on a particular section of the route; f max This indicates the maximum frequency of traffic across all air routes.

[0013] Furthermore, trajectory pattern recognition is performed on the ship's sub-trajectory to obtain trajectory behavior patterns, and the ship's sub-trajectory is evaluated based on the trajectory behavior patterns. Based on the evaluation results, trajectory correction is performed on the ship's sub-trajectory, including: S41. Extract the ship behavior feature vector based on the ship sub-trajectory, and use Euclidean distance to perform trajectory pattern matching on the ship behavior feature vector to obtain the trajectory behavior pattern. S42. Based on the trajectory behavior pattern and the pre-acquired navigation rule base, the ship sub-trajectory is evaluated, and the evaluation results are used to judge the abnormal navigation sub-trajectory to obtain the abnormal ship sub-trajectory. S43. Use trajectory behavior patterns to correct the abnormal ship sub-trajectories, obtain the corrected sub-trajectories, and perform trajectory splicing based on the corrected sub-trajectories to obtain the optimized ship prediction trajectory.

[0014] Furthermore, trajectory behavior patterns are used to correct the abnormal vessel sub-trajectories, resulting in corrected sub-trajectories. These corrected sub-trajectories are then stitched together to obtain optimized predicted vessel trajectories, including: S431. Extract the spatiotemporal state data of ship trajectory points based on abnormal ship sub-trajectories, and construct boundary constraints for correcting the trajectory by combining the trajectory behavior pattern. S432. Based on the boundary constraints of the corrected trajectory and combined with the trajectory cost function, perform global trajectory reconstruction on the abnormal ship sub-trajectories to generate a candidate corrected trajectory set; S433. Calculate the trajectory deviation through candidate corrected trajectories, compare the trajectory deviation with the preset deviation threshold, use the comparison results to perform optimal screening of candidate corrected trajectories, obtain the corrected sub-trajectories, and perform trajectory splicing on the corrected sub-trajectories to obtain the optimized ship prediction trajectory.

[0015] According to another aspect of the present invention, a ship entry and exit trajectory acquisition system based on BeiDou positioning is provided. The ship entry and exit trajectory acquisition system based on BeiDou positioning includes: a ship navigation feature extraction module, a ship trajectory prediction module, a ship trajectory division module, and a ship trajectory correction module. The ship navigation feature extraction module is used to acquire ship navigation data, preprocess the ship navigation data to obtain preprocessed ship navigation data, and extract ship navigation features based on the preprocessed ship navigation data. The ship trajectory prediction module is used to filter ship navigation features to obtain navigation trajectory features, and predict ship trajectory based on navigation trajectory features to obtain the predicted ship trajectory. The ship trajectory segmentation module is used to generate a sea area profile based on pre-acquired navigation sea area data, and to segment the predicted ship trajectory using the sea area profile to obtain ship sub-trajectories; The ship trajectory correction module is used to perform trajectory pattern recognition on the ship sub-trajectory, obtain the trajectory behavior pattern, evaluate the ship sub-trajectory based on the trajectory behavior pattern, and correct the ship sub-trajectory based on the evaluation result.

[0016] The beneficial effects of this invention are as follows: 1. This invention achieves the generation of predicted ship trajectories through ship trajectory prediction. At the same time, by introducing spatial semantic segmentation driven by marine imagery and a closed-loop correction mechanism based on trajectory behavior patterns, it realizes a leap from geometric trajectory fitting to understanding navigation intentions. This significantly improves the usability and decision support capabilities of trajectory data in scenarios such as intelligent pilotage, maritime supervision, navigation safety early warning, and port scheduling optimization. It makes trajectory segmentation no longer limited to geometric distance or time segments, but logically segmented according to the actual navigation context of the ship, thus enhancing the functional interpretability of sub-trajectories.

[0017] 2. This invention, through marine profiling, can accurately identify key nodes of a trajectory, enabling refined segmentation of the trajectory and ensuring that each sub-trajectory corresponds to a relatively independent and consistent navigation phase. It effectively solves the problems of strong subjectivity and high threshold dependence in trajectory segmentation in traditional methods, improves the comparability of trajectory structures between different ships and different voyages, lays a solid foundation for subsequent behavior pattern mining and anomaly detection, and significantly improves the rationality of ship navigation planning in complex nearshore waters.

[0018] 3. This invention achieves refined and semantic modeling of the ship's entry and exit from port by performing spatial semantic segmentation on the predicted trajectory based on the marine image, which significantly improves the understanding depth of the trajectory data. Compared with traditional methods that often use fixed time intervals or distance thresholds to mechanically segment the trajectory, spatial semantic segmentation can better reflect the actual behavioral stage changes of the ship in complex waters. Each segment of the trajectory corresponds to an independent behavior with a clear navigation intention and environmental constraints, ensuring that the segmentation results are highly consistent with the actual navigation context. This not only achieves refined semantic deconstruction of the ship's navigation process, but also significantly improves the efficiency of subsequent trajectory correction. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a flowchart of a method for obtaining ship entry and exit trajectories based on BeiDou positioning according to an embodiment of the present invention; Figure 2 This is a schematic diagram of a ship entry and exit trajectory acquisition system based on BeiDou positioning according to an embodiment of the present invention.

[0021] In the picture: 1. Ship navigation feature extraction module; 2. Ship trajectory prediction module; 3. Ship trajectory division module; 4. Ship trajectory correction module. Detailed Implementation

[0022] To further illustrate the various embodiments, the present invention provides accompanying drawings, which are part of the disclosure of the present invention. These drawings are mainly used to illustrate the embodiments and can be used in conjunction with the relevant descriptions in the specification to explain the operating principles of the embodiments. With reference to these drawings, those skilled in the art should be able to understand other possible implementation methods and the advantages of the present invention.

[0023] According to an embodiment of the present invention, a method and system for obtaining ship entry and exit trajectories based on BeiDou positioning are provided.

[0024] The present invention will now be further described in conjunction with the accompanying drawings and specific embodiments, such as... Figure 1 As shown, according to an embodiment of the present invention, a method for obtaining ship entry and exit trajectories based on BeiDou positioning is provided, the method comprising the following steps: S1. Acquire ship navigation data, preprocess the ship navigation data to obtain preprocessed ship navigation data, and extract ship navigation features based on the preprocessed ship navigation data.

[0025] Specifically, ship navigation data is acquired and denoised to obtain cleaned ship navigation data. Linear interpolation is then used to interpolate the cleaned ship navigation data to obtain completed ship navigation data. Coordinate transformation is performed on the completed ship navigation data to obtain transformed ship navigation data. The sampling frequency of the transformed ship navigation data is then aligned using time series to obtain preprocessed ship navigation data. Statistical analysis is used to extract ship navigation features from the preprocessed ship navigation data to obtain initial ship navigation features. These initial ship navigation features are then normalized to obtain the final ship navigation characteristics.

[0026] Specifically, taking a vessel operating in a coastal port as an example, its raw navigation data is obtained through a shipborne BeiDou terminal. This raw data includes fields such as timestamp, WGS-84 latitude and longitude, ground speed, and ground heading, with a sampling frequency of 1Hz. Data cleaning based on kinematic constraints is used to eliminate signal transmission noise. Specifically, based on the target vessel's model and its design maneuverability within the port's restricted waters, a set of absolute physical boundary parameters is pre-set, including maximum ground speed, maximum rate of change of ground heading, and maximum acceleration or deceleration. The maximum ground speed is set to 18.0 knots, derived from port regulations and vessel certificates; the maximum rate of change of ground heading is set to 15 degrees / second, based on full-rudder turning test data for this vessel type; and the maximum acceleration or deceleration is set to 0.5 knots / second, calculated based on the main engine and thruster power. These parameters constitute a hard constraint filter for data validity. The raw BeiDou data stream input per second is processed frame by frame. For each frame of data, its key dynamic parameters (ground speed and ground heading) are immediately compared with the aforementioned preset constraints. If the speed exceeds 18.0 knots, or the absolute value of the calculated rate of change of heading between two adjacent frames exceeds 15 degrees / second, or the absolute value of the calculated acceleration exceeds 0.5 knots / second, the data frame is immediately marked as a suspected noise point. For each marked point, two consecutive data points before and after the suspected point are extracted to form an analysis window. The median movement of speed and heading for all unmarked points within this window is calculated. The absolute deviation between the measured value of the suspected point and the corresponding median movement is calculated. If this deviation exceeds a secondary threshold preset based on sensor accuracy, for example, a speed deviation exceeding 3 knots or a heading deviation exceeding 10 degrees, the point is ultimately determined to be a noise point requiring repair. For data points confirmed as noise, linear interpolation of the preceding and following valid points is used for repair. That is, using the two nearest valid data points before and after the noise point, linear interpolation is performed on latitude, longitude, speed, and heading in the time dimension to generate a replacement value. For example, if a ship's speed at a certain moment is recorded as 25 knots due to a signal jump, while the speeds at the preceding and following valid points are 10.1 knots and 10.3 knots respectively, the speed at that moment is corrected to approximately 10.2 knots after interpolation. The entire cleaning process is based entirely on publicly available ship kinematic equations and defined threshold comparisons, median calculations, and linear interpolation operations, ensuring the objectivity and traceability of the processing results. It effectively separates and corrects physically impossible noise data, while fully preserving real navigation behavior data that conforms to ship maneuvering patterns and may represent emergency avoidance or sharp turns, laying a high-fidelity data foundation for subsequent analysis. Using coordinate system one, the WGS-84 coordinates are converted to UTM Zone 51N plane projection coordinates. For example, after conversion, the east coordinate is 340123.5 meters and the north coordinate is 3467890.2 meters, and all data are resampled and aligned to a 1Hz equally spaced time series.Based on this, statistical analysis is used to extract ship navigation characteristics from the processed trajectory sequence, calculating initial feature vectors including average speed, standard deviation of heading, average acceleration, and trajectory curvature. Using a preset sliding time window of 60 seconds, deterministic statistical calculations are performed on the UTM coordinate sequence and corresponding ground speed and ground heading data within the window, generating a multi-dimensional initial feature vector characterizing the ship's motion state during that time period. For each window, the average ground speed is calculated: the arithmetic mean of the ground speed values ​​of all sampled points within the window is taken, reflecting the overall speed of the ship during that time period. The standard deviation of heading is calculated: the standard deviation of the ground heading values ​​of all sampled points within the window is calculated to quantify the stability or degree of oscillation of the ship's heading; a larger standard deviation indicates more frequent heading adjustments. The average acceleration is calculated: by performing a first-order difference on the ground speed sequence within the window, a second-by-second instantaneous acceleration sequence is obtained, and then the arithmetic mean of this sequence is taken to characterize the overall acceleration or deceleration trend of the ship. Then, the trajectory curvature is calculated: based on continuous UTM coordinate points within the window, the reciprocal of the instantaneous radius of curvature is calculated point by point using the three-point method. The median of the absolute curvature values ​​of all points within the window is then taken to describe the degree of trajectory curvature; a high curvature value indicates a turn or maneuver. In addition, other statistical features are simultaneously calculated, such as maximum speed, speed variation range, cumulative heading change (i.e., the sum of the absolute values ​​of all COG differences within the window), and total navigation distance within the window (obtained by accumulating the Euclidean distances between adjacent points). This process is executed progressively at each new sampling time, transforming the continuous spatiotemporal trajectory data stream into a series of feature vector sequences with clear physical meaning and statistical representativeness. These feature vectors are then Z-score standardized to obtain normalized ship navigation feature vectors, which serve as input for subsequent models. Features representing the ship's navigation state are extracted from the preprocessed regular trajectory sequence. The standardized ship navigation feature vectors are output for subsequent trajectory prediction and analysis modules.

[0027] S2. Perform feature filtering on the ship's navigation characteristics to obtain navigation trajectory characteristics, and predict the ship's trajectory based on the navigation trajectory characteristics to obtain the predicted ship trajectory.

[0028] Specifically, feature filtering is performed on the ship's navigation characteristics to obtain navigation trajectory features, and ship trajectory prediction is performed based on these features. The predicted ship trajectory includes: S21. Use the Pearson coefficient to perform correlation analysis on the ship's navigation characteristics, and based on the correlation analysis results, remove redundant features from the ship's navigation characteristics to obtain the initial navigation trajectory characteristics. S22. Perform a statistical significance test on the initial navigation trajectory features, and filter the initial navigation trajectory features based on the significance test results and a preset threshold to obtain the navigation trajectory features; S23. Based on the characteristics of the navigation trajectory and the preset correction gain coefficient, the trajectory is predicted and corrected to obtain the predicted trajectory of the ship.

[0029] Specifically, based on the characteristics of the navigation trajectory and the preset correction gain coefficient, trajectory prediction and correction are performed to obtain the ship's predicted trajectory, including: S231. Based on the characteristics of the navigation trajectory, perform preliminary trajectory prediction to obtain preliminary trajectory prediction results; S232. Calculate the trajectory error based on the preliminary trajectory prediction results, and correct the preliminary trajectory prediction results by combining the preset correction gain coefficient to obtain the ship prediction trajectory.

[0030] Specifically, for the generated ship navigation feature set, including average speed, speed standard deviation, mean heading, standard deviation of heading, lateral acceleration, longitudinal acceleration, trajectory curvature, rate of change of heading, cumulative turning angle, offset from the channel centerline, and distance to the preceding ship, Pearson correlation coefficient matrices are calculated between each pair of these features. By analyzing this matrix, highly linearly correlated redundant features are identified and eliminated. Specifically, the correlation coefficient was found to be 0.92, exceeding the preset redundancy threshold of 0.85. Furthermore, both lateral acceleration and rate of change of heading physically represent the ship's turning maneuver intensity; therefore, based on domain knowledge, the rate of change of heading, with its more direct physical meaning, is retained, while lateral acceleration is eliminated. Similarly, average speed and longitudinal acceleration also show a strong correlation (correlation coefficient 0.78), but considering that average speed is the fundamental core variable describing the motion state, and longitudinal acceleration better reflects the dynamic process of acceleration and deceleration, both are retained. After this round of analysis, three redundant features were eliminated from the initial features, resulting in the initial navigation trajectory feature set. Statistical significance tests were performed on the initial features to screen out the features most influential on the trajectory prediction target, i.e., the prediction of the ship position sequence within the next 60 seconds. The implementation method employed univariate F-test analysis of variance. Historical port entry and exit trajectory data were used as the training set, where each sample contained a feature value and the corresponding true position sequence for the next 60 seconds, with the true position sequence serving as the prediction target. For each feature, the error between its different value intervals and the prediction target was analyzed, i.e., whether there was a significant difference between the mean Euclidean distance between the predicted position and the true position. Each feature was discretized and binned, and the average prediction error of the samples within each bin was calculated. Then, an F-test was used to determine whether the between-group variance was significantly greater than the within-group variance. Specifically, the feature's offset from the channel centerline was analyzed, and its value range was divided into 5 equal intervals. Calculations showed that when the offset was greater than 50 meters, the average prediction error increased significantly, i.e., the significance p-value was less than 0.01, indicating that this feature had a significant impact on the prediction error. Conversely, if the average prediction error of the feature and the distance to the preceding ship is not significantly different across different intervals (i.e., the p-value is greater than 0.1), it indicates that under the current port traffic density, the feature does not significantly contribute to the prediction of a single ship trajectory, and therefore it is discarded. Setting the significance level α to 0.05, features that pass the test are ultimately retained. This screening process significantly reduces the feature dimensionality, improving the efficiency and generalization ability of the subsequent prediction model. After obtaining concise and effective navigation trajectory features, trajectory prediction and real-time correction are performed. This involves inputting the current navigation trajectory feature vector into a pre-trained lightweight machine learning model. This model can use an Extreme Gradient Boosting Tree (XGBoost) regression model, which has been trained based on a large amount of historical port entry and exit trajectory data. It can directly output a sequence of predicted latitude and longitude coordinates for a series of future time points, such as 12 points in the next 60 seconds at 5-second intervals, based on the input features; this is the preliminary trajectory prediction result.A large amount of complete port entry and exit trajectory data, covering different ship types and weather conditions, was selected from historical databases. Each data point underwent preprocessing and feature extraction to form a standardized sample set. This set used standardized navigation trajectory features as input features and the corresponding actual UTM coordinate sequence of ships within a fixed future time period as prediction labels. This sample set was then randomly divided into training, validation, and test sets in a 7:2:1 ratio. Finally, the computationally efficient and interpretable Extreme Gradient Boosting Tree (XGBoost) regression algorithm was selected as the base model, and its hyperparameters were initialized, such as a learning rate of 0.1 and a maximum tree depth of [missing information]. The model was trained using a subsampling ratio of 0.8 and a subsample ratio of 6. The objective function was set to minimize the root mean square error (RMSE) between the predicted and true coordinates. During training, an early stopping method was used to monitor the model. Training was stopped when the validation set loss stopped decreasing after 10 consecutive iterations to prevent overfitting. After training, the model performance was evaluated on an independent test set to ensure that the prediction error met a preset accuracy threshold, for example, the average error of the predicted position in the next 60 seconds should not exceed 50 meters. The trained model parameters were then solidified, packaged, and deployed to an online prediction platform for real-time use. Because ship motion is affected by real-time environmental factors such as sudden changes in wind and current or emergency avoidance, pure model predictions may have instantaneous biases. Therefore, a real-time correction mechanism based on error feedback was introduced. Whenever new BeiDou positioning data is received, for example, after 1 second, the new positioning point is immediately compared with the predicted point at the corresponding time, and the trajectory error vector, including eastward and northward errors, is calculated in meters. A preset correction gain coefficient K is used, with its value determined experimentally to be between 0.2 and 0.5; in this example, K=0.3 is chosen. The correction logic is as follows: the error vector is multiplied by the correction gain coefficient K to obtain the current correction amount, which is then simultaneously superimposed on all future prediction points that have not yet occurred. That is, for all prediction point coordinates after the current moment in the preliminary trajectory prediction results, this correction amount is added to the eastward and northward directions respectively. This operation is physically equivalent to assuming that the currently observed prediction deviation reflects systematic model bias or environmental disturbances, and making a corresponding proportional correction to the future trajectory. This process is repeated at each new observation moment, forming a closed-loop feedback. The output is a more accurate ship prediction trajectory after continuous and dynamic correction. For example, the initial predicted position 60 seconds later, after several such real-time error feedback corrections, has its final output coordinates dynamically adjusted according to the actual navigation situation, thus significantly improving the real-time performance and accuracy of trajectory prediction in complex port entry and exit environments, providing a high-quality data foundation for subsequent trajectory semantic segmentation and behavior analysis.

[0031] S3. Generate a sea area profile based on the pre-acquired navigation sea area data, and divide the predicted ship trajectory using the sea area profile to obtain the ship sub-trajectory.

[0032] Specifically, a sea area profile is generated based on pre-acquired navigation sea area data, and the predicted ship trajectory is divided using the sea area profile to obtain ship sub-trajectories, including: S31. Based on the pre-acquired navigation area data, the navigation environment constraints and navigation environment parameters are extracted using BeiDou positioning technology, and navigation area environment data is constructed based on the extracted navigation environment constraints and navigation environment parameters.

[0033] S32. Convert the navigation area environmental data into a vector layer, and combine it with the pre-acquired ship morphology parameters to perform navigation adaptability filtering on the vector layer to obtain the navigation area layer.

[0034] S33. Extract the starting point of the route from the navigation area layer, construct the route network skeleton based on the starting point of the route, perform semantic parsing on the route network skeleton, and generate a sea area image based on the semantic parsing results.

[0035] Specifically, the starting points of shipping routes are extracted from the navigation area layer, a route network skeleton is constructed based on these starting points, and semantic parsing is performed on the route network skeleton. The resulting sea area profile is generated using the semantic parsing results, including: S331. Extract the starting point of the route based on the navigation area layer, and use the starting point of the route as the initial source of route growth. S332. Set the direction of the channel light source, and use the direction of the channel light source to guide the initial route growth source to extend along the optimal navigation path, so as to obtain the preliminary optimized trunk channel layout. S333. Utilize the apex dominance principle to prioritize the expansion of the main routes in the initially optimized main channel layout to obtain the main route network. Then, extend the branch route network according to the main route network to obtain the route network skeleton. S334. Perform functional area analysis on the route network skeleton, and perform semantic annotation based on the functional area analysis results to generate a sea area profile.

[0036] S34. Utilize sea area images to perform route spatial topology analysis, and based on the results of the route spatial topology analysis, perform spatial semantic segmentation on the predicted ship trajectory to obtain the ship sub-trajectory.

[0037] Specifically, a sea area profile is used to perform route spatial topology analysis. Based on the results of the route spatial topology analysis, the predicted ship trajectories are spatially semantically segmented to obtain ship sub-trajectories, including: S341. Extract the spatial geometry and functional zones of air routes based on the sea area image, and perform air route topology modeling based on the spatial geometry and functional zones of air routes to obtain the air route topology network.

[0038] S342. Perform route spatial topology analysis using route topology network, set route distance thresholds based on route spatial topology analysis results, and perform spatial proximity analysis on ship predicted trajectories based on route distance thresholds.

[0039] Specifically, the formula for calculating the route distance threshold is as follows: ; In the formula, t Indicates the route distance threshold; e Represents the set of edges in the route topology network; N Indicates the total number of route segments; e ij Indicates the connection node i and j The route segment; d ij Indicates the length of the route segment; m d This represents the average length of all flight segments; c Indicates the first weighting coefficient; s (·) represents an sigmoid function; C env ( e ij This represents the overall marine environmental cost of the voyage segment; This represents the global average marine environmental cost; l Indicates the second weighting coefficient; Δ i ij Indicates the connection node i Outbound heading angle and connecting nodes j The change in heading between the incoming heading angles; β This represents the third weighting coefficient; f ij Indicates the frequency of vessel traffic on a particular section of the route; f max This indicates the maximum frequency of traffic across all air routes.

[0040] S343. Based on the spatial proximity analysis results, the initial predicted ship trajectory is divided into preliminary ship sub-trajectories, and the preliminary ship sub-trajectories are spatially semantically labeled using the route functional area to obtain the ship sub-trajectories.

[0041] Specifically, the pre-acquired navigation area data can be obtained from pre-accessed electronic nautical chart databases, water depth measurement data with an accuracy of 0.5 meters, channel boundary vectors, tidal current models, and meteorological information. Combined with the ship's morphological parameters (specifically, a total length of 315 meters, a beam of 48 meters, and a maximum draft of 16.8 meters), navigation environmental constraints are extracted. For example, navigation restrictions such as a main channel width ≥ 400 meters, a minimum water depth ≥ 18 meters, and a turning radius ≥ 1.2 kilometers are automatically identified. A navigation area environmental dataset containing 12 types of environmental parameters is constructed, such as a water depth of 20.3 meters, a current speed of 1.8 knots, and a wind direction (NW) of force 6 for a certain route segment. This data is then imported into a GIS platform and converted into multi-layer vector maps, including the channel centerline, prohibited areas, anchorages, and bridge clearance. Subsequently, navigation suitability filtering is performed based on ship dimensions. For example, if a channel is only 320 meters wide, which is less than the ship's safe width, a width of ≥400 meters is recommended, and the channel is automatically marked as unsuitable and removed from the navigable area. Simultaneously, shallow water areas with a depth of less than 18 meters are also filtered out, ultimately generating a navigational area layer containing only the main channel, turning area, and berth front waters. Based on this layer, the route starting point is extracted, using the ship's initial BeiDou position (122.1567°E, 29.9876°N) as the initial growth source. A phototropic mechanism is set, defining the direction of the channel centerline as the light source direction. For example, if the main channel's direction is 85°, the route is guided to extend along this direction, prioritizing the path with the smallest angle to 85°. The principle of apex dominance is applied to path expansion, prioritizing the growth of main channels along the paths with the greatest water depth and least curvature. For example, a main channel extends from the starting point to a port area, approximately 18.5 km long with an average width of 420 meters. At branch points, such as the confluence of waterways, two branch channels extend based on historical vessel track density >50 vessels / day. This ultimately constructs a channel network framework comprising one main channel, three branches, and five key nodes. Functional area analysis is performed on this framework. By overlaying a port operation area layer, the functional attributes of each segment are identified. For instance, assuming a port's inbound channel is 15 km long, it is divided into multiple functional segments: a 6.2 km straight inbound segment, a 2.3 km high-angle turning area with a 68° heading change, and a 6.5 km berthing guidance segment. Semantic annotation is then applied to generate a structured sea area profile. Using this image, topological analysis was performed to extract the route geometry and functional areas of the node-edge topology, constructing a route topology network containing 8 nodes, 7 edges, and a total of [number missing] flight segments. N =7. Calculate the route distance threshold. t ,in, m d =2.6km, which is the average length of the 7 segments; c =0.3, l =0.3, β =0.2; Cenv ( e ij Taking into account factors such as water depth, current velocity, and wind and waves, such as e 23 part C env =0.82, indicating a relatively high risk; global average. =0.65; Δ i 23 =68°; f ij Historical frequency of use f max =120 ships / day, a certain waterway f ij =45. Substituting into the calculation, we get... t ≈380 meters. The output initial predicted trajectory, i.e., one UTM coordinate point every 30 seconds, is overlaid with the sea area image for spatial proximity analysis. If a predicted point is more than 380 meters away from the centerline of the current route segment, it is determined that the vessel has entered a new functional area. Finally, the initial trajectory, which is 18.5 km long, is divided into 4 sub-trajectories: T1 is 0 to 6.2 km, which is for direct navigation into port; T2 is 6.2 to 8.5 km, which is for sharp turns; T3 is 8.5 to 15.0 km, which is for deceleration guidance; and T4 is 15.0 to 18.5 km, which is for berthing and alignment. Each sub-trajectory is assigned a corresponding semantic label.

[0042] S4. Perform trajectory pattern recognition on the ship sub-trajectory to obtain the trajectory behavior pattern, evaluate the ship sub-trajectory based on the trajectory behavior pattern, and correct the ship sub-trajectory based on the evaluation results.

[0043] Specifically, trajectory pattern recognition is performed on the ship's sub-trajectory to obtain trajectory behavior patterns, and the ship's sub-trajectory is evaluated based on the trajectory behavior patterns. Based on the evaluation results, trajectory correction is performed on the ship's sub-trajectory, including: S41. Extract the ship behavior feature vector based on the ship sub-trajectory, and use Euclidean distance to perform trajectory pattern matching on the ship behavior feature vector to obtain the trajectory behavior pattern. S42. Based on the trajectory behavior pattern and the pre-acquired navigation rule base, the ship sub-trajectory is evaluated, and the evaluation results are used to judge the abnormal navigation sub-trajectory to obtain the abnormal ship sub-trajectory. S43. Use trajectory behavior patterns to correct the abnormal ship sub-trajectories, obtain the corrected sub-trajectories, and perform trajectory splicing based on the corrected sub-trajectories to obtain the optimized ship prediction trajectory.

[0044] Specifically, trajectory behavior patterns are used to correct abnormal vessel sub-trajectories, resulting in corrected sub-trajectories. These corrected sub-trajectories are then stitched together to obtain optimized predicted vessel trajectories, including: S431. Extract the spatiotemporal state data of ship trajectory points based on abnormal ship sub-trajectories, and construct boundary constraints for correcting the trajectory by combining the trajectory behavior pattern. S432. Based on the boundary constraints of the corrected trajectory and combined with the trajectory cost function, perform global trajectory reconstruction on the abnormal ship sub-trajectories to generate a candidate corrected trajectory set; S433. Calculate the trajectory deviation through candidate corrected trajectories, compare the trajectory deviation with the preset deviation threshold, use the comparison results to perform optimal screening of candidate corrected trajectories, obtain the corrected sub-trajectories, and perform trajectory splicing on the corrected sub-trajectories to obtain the optimized ship prediction trajectory.

[0045] Specifically, if the ship is a 300,000-ton oil tanker, 333 meters long, 58 meters wide, and with a draft of 20.5 meters, the total trajectory is approximately 42 kilometers. Four sub-trajectories are extracted and output: T1 (0-6.2 km), representing a straight approach to port; T2 (6.2-8.5 km), representing a sharp turn; T3 (8.5-15.0 km), representing deceleration and guidance; and T4 (15.0-18.5 km), representing berthing and alignment. A behavioral feature vector is then constructed. Taking T1 as an example, based on the UTM Zone 50N coordinate sequence, with one point every 30 seconds (124 points in total), the calculated average speed is 12.6 knots with a standard deviation of 0.4 knots. The heading stability shows a standard deviation of 2.1°, and the average acceleration is -0.08 knots / min, indicating a slight deceleration trend. The curvature is... The curvature is close to a straight line, and the initial feature vector is constructed as [12.6, 0.4, 2.1, -0.08, 0.0002, ...], with a total of 10 dimensions. For segment T2, the sharp turn characteristics are significantly different; the average speed drops to 9.3 knots, the heading changes from 87° to 156°, the maximum heading change rate reaches 5.2° / min, and the peak curvature reaches [missing value]. The mean acceleration was -0.35 knots / min, and the eigenvectors showed strong dynamics. Matching the Euclidean distance with 500 port approach trajectory templates in the historical database revealed that the minimum distance between segment T2 and the high-curvature turning template was 0.31, with a threshold of 0.5, successfully identifying it as a forced turning behavior pattern in restricted waters. Further evaluation using the navigation safety management rule base, which explicitly requires [further details needed], showed that [further details needed] the high-curvature turning pattern. The speed limit on curves is 8.0 knots. However, the ship reached 8.9 knots at 7.3km on section T2, exceeding the limit by 0.9 knots for 1 minute and 45 seconds, with a curvature of [missing information]. The trajectory was determined to be abnormal. Sections T1, T3, and T4 were assessed as normal. In section T3, the speed smoothly decreased from 9.0 knots to 3.2 knots, meeting the guidance requirement of a speed reduction of approximately 0.5 knots per kilometer. The lateral deviation in section T4 was controlled within ±8 meters, meeting berthing accuracy requirements. Trajectory correction was performed on the abnormal section T2. ​​The original 67 trajectory points of section T2 were extracted. The east coordinates are 547200 to 549100, and the north coordinates are 3321800 to 3324500. Based on the forced steering behavior pattern, correction boundary constraints were constructed: 1) the turning center is locked at (548300±30 meters, 3323200±30 meters); 2) the minimum turning radius is ≥1.05 kilometers; 3) the maximum speed is 8.0 knots; and 4) the rate of curvature change... Based on the above constraints, a trajectory cost function is defined using a linear weighting method. This involves linearly weighting the fuel consumption (proportional to the square of acceleration), the reciprocal of the shortest distance to the channel boundary, and the integral of the square of the rate of change of heading, with weights of 0.4, 0.4, and 0.2, respectively. Fifty candidate corrected trajectories are generated within the constraint space using dynamic programming. The trajectory deviation of each candidate trajectory from the original abnormal trajectory is calculated, i.e., the distance interpolation between each candidate trajectory and the original abnormal trajectory is calculated, summed, and averaged, while considering a preset deviation threshold of 120 meters. Candidate trajectory 9 has a deviation of 112 meters, its peak speed is reduced to 7.8 knots, and its curvature is optimized to... The lowest cost function was 3.01, and it was selected as the optimal corrected trajectory T2'. Finally, T1, T2', T3, and T4 were continuously spliced ​​together to ensure a smooth transition in position, heading, and curvature, generating an optimized ship prediction trajectory with a total length of 18.62 kilometers, which is 120 meters longer than the original trajectory.

[0046] like Figure 2 As shown, according to another embodiment of the present invention, a ship entry and exit trajectory acquisition system based on BeiDou positioning is provided. The ship entry and exit trajectory acquisition system based on BeiDou positioning includes: a ship navigation feature extraction module 1, a ship trajectory prediction module 2, a ship trajectory division module 3, and a ship trajectory correction module 4. Ship navigation feature extraction module 1 is used to acquire ship navigation data, preprocess the ship navigation data to obtain preprocessed ship navigation data, and extract ship navigation features based on the preprocessed ship navigation data. Ship trajectory prediction module 2 is used to filter ship navigation features to obtain navigation trajectory features, and predict ship trajectory based on navigation trajectory features to obtain the predicted ship trajectory. The ship trajectory segmentation module 3 is used to generate a sea area profile based on the pre-acquired navigation sea area data, and to segment the predicted ship trajectory through the sea area profile to obtain ship sub-trajectories; The ship trajectory correction module 4 is used to perform trajectory pattern recognition on the ship sub-trajectory, obtain the trajectory behavior pattern, evaluate the ship sub-trajectory based on the trajectory behavior pattern, and correct the ship sub-trajectory based on the evaluation result.

[0047] In summary, by utilizing the above-mentioned technical solutions of this invention, the present invention achieves ship trajectory prediction and generation through ship trajectory prediction. Simultaneously, by introducing spatial semantic segmentation driven by marine imagery and a closed-loop correction mechanism based on trajectory behavior patterns, it achieves a leap from geometric trajectory fitting to understanding navigation intentions. This significantly improves the usability and decision support capabilities of trajectory data in scenarios such as intelligent pilotage, maritime supervision, navigation safety early warning, and port scheduling optimization. Trajectory segmentation is no longer limited to geometric distance or time segments, but is logically segmented based on the actual navigation context of the ship, enhancing the functional interpretability of sub-trajectories. Through marine imagery, this invention can accurately identify key nodes of the trajectory, achieving refined segmentation of the trajectory and ensuring that each sub-trajectory corresponds to a relatively independent and behaviorally consistent navigation phase. This effectively solves the problem of strong subjectivity in trajectory segmentation in traditional methods. This invention addresses the issue of high threshold dependence, enhancing the comparability of trajectory structures between different vessels and voyages. This lays a solid foundation for subsequent behavior pattern mining and anomaly detection, significantly improving the rationality of vessel navigation planning in complex nearshore waters. Furthermore, by performing spatial semantic segmentation on predicted trajectories based on marine profiles, this invention achieves refined and semantic modeling of vessel entry and exit processes, significantly improving the depth of understanding of trajectory data. Compared to traditional methods that often use fixed time intervals or distance thresholds for mechanical trajectory segmentation, spatial semantic segmentation better reflects the actual behavioral stages of vessels in complex waters. Each sub-trajectory corresponds to an independent behavior with a clear navigation intention and environmental constraints, ensuring a high degree of consistency between the segmentation results and the actual navigation context. This not only achieves refined semantic deconstruction of the vessel's navigation process but also significantly improves the efficiency of subsequent trajectory correction.

[0048] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for obtaining ship entry and exit trajectories based on BeiDou positioning, characterized in that, The method includes: S1. Acquire ship navigation data, preprocess the ship navigation data to obtain preprocessed ship navigation data, and extract ship navigation features based on the preprocessed ship navigation data. S2. Perform feature filtering on the ship's navigation characteristics to obtain navigation trajectory features, and predict the ship's trajectory based on the navigation trajectory features to obtain the predicted ship trajectory. S3. Generate a sea area profile based on the pre-acquired navigation sea area data, and divide the predicted ship trajectory using the sea area profile to obtain the ship sub-trajectory; S4. Perform trajectory pattern recognition on the ship sub-trajectory to obtain the trajectory behavior pattern, evaluate the ship sub-trajectory based on the trajectory behavior pattern, and correct the ship sub-trajectory based on the evaluation results.

2. The method for obtaining ship entry and exit trajectories based on BeiDou positioning according to claim 1, characterized in that, The process of filtering ship navigation characteristics to obtain navigation trajectory characteristics, and then predicting the ship trajectory based on these characteristics, results in the following predicted ship trajectory: S21. Use the Pearson coefficient to perform correlation analysis on the ship's navigation characteristics, and based on the correlation analysis results, remove redundant features from the ship's navigation characteristics to obtain the initial navigation trajectory characteristics. S22. Perform a statistical significance test on the initial navigation trajectory features, and filter the initial navigation trajectory features based on the significance test results and a preset threshold to obtain the navigation trajectory features; S23. Based on the characteristics of the navigation trajectory and the preset correction gain coefficient, the trajectory is predicted and corrected to obtain the predicted trajectory of the ship.

3. The method for obtaining ship entry and exit trajectories based on BeiDou positioning according to claim 2, characterized in that, The process of predicting and correcting the trajectory based on the characteristics of the navigation trajectory and a preset correction gain coefficient to obtain the predicted ship trajectory includes: S231. Based on the characteristics of the navigation trajectory, perform preliminary trajectory prediction to obtain preliminary trajectory prediction results; S232. Calculate the trajectory error based on the preliminary trajectory prediction results, and correct the preliminary trajectory prediction results by combining the preset correction gain coefficient to obtain the ship prediction trajectory.

4. The method for obtaining ship entry and exit trajectories based on BeiDou positioning according to claim 1, characterized in that, The process involves generating a sea area profile based on pre-acquired navigation sea area data, and then dividing the predicted ship trajectory using the sea area profile to obtain ship sub-trajectories, including: S31. Based on the pre-acquired navigation area data, the navigation environment constraints and navigation environment parameters are extracted using BeiDou positioning technology, and navigation area environment data is constructed based on the extracted navigation environment constraints and navigation environment parameters. S32. Convert the navigation area environmental data into a vector layer, and combine it with the pre-acquired ship morphology parameters to perform navigation adaptability filtering on the vector layer to obtain the navigation area layer. S33. Extract the starting point of the route from the navigation area layer, construct the route network skeleton based on the starting point of the route, perform semantic parsing on the route network skeleton, and generate a sea area image based on the semantic parsing results. S34. Utilize sea area images to perform route spatial topology analysis, and based on the results of the route spatial topology analysis, perform spatial semantic segmentation on the predicted ship trajectory to obtain the ship sub-trajectory.

5. The method for obtaining ship entry and exit trajectories based on BeiDou positioning according to claim 4, characterized in that, The process of extracting route start points from the navigation area layer, constructing a route network skeleton based on the route start points, performing semantic parsing on the route network skeleton, and generating a sea area profile based on the semantic parsing results includes: S331. Extract the starting point of the route based on the navigation area layer, and use the starting point of the route as the initial source of route growth. S332. Set the direction of the channel light source, and use the direction of the channel light source to guide the initial route growth source to extend along the optimal navigation path, so as to obtain the preliminary optimized trunk channel layout. S333. Utilize the apex dominance principle to prioritize the expansion of the main routes in the initially optimized main channel layout to obtain the main route network. Then, extend the branch route network according to the main route network to obtain the route network skeleton. S334. Perform functional area analysis on the route network skeleton, and perform semantic annotation based on the functional area analysis results to generate a sea area profile.

6. The method for obtaining ship entry and exit trajectories based on BeiDou positioning according to claim 4, characterized in that, The method involves using marine imagery to perform route spatial topology analysis, and then performing spatial semantic segmentation on the predicted ship trajectories based on the results of the route spatial topology analysis. The resulting ship sub-trajectories include: S341. Extract the spatial geometry and functional zones of air routes based on the sea area image, and perform air route topology modeling based on the spatial geometry and functional zones of air routes to obtain the air route topology network. S342. Perform route spatial topology analysis using route topology network, set route distance thresholds based on route spatial topology analysis results, and perform spatial proximity analysis on ship predicted trajectories based on route distance thresholds. S343. Based on the spatial proximity analysis results, the initial predicted ship trajectory is divided into preliminary ship sub-trajectories, and the preliminary ship sub-trajectories are spatially semantically labeled using the route functional area to obtain the ship sub-trajectories.

7. The method for obtaining ship entry and exit trajectories based on BeiDou positioning according to claim 6, characterized in that, The formula for calculating the route distance threshold is: ； In the formula, τ Indicates the route distance threshold; ε Represents the set of edges in the route topology network; N Indicates the total number of route segments; e ij Indicates the connection node i and j The route segment; d ij Indicates the length of the route segment; μ d This represents the average length of all flight segments; γ Indicates the first weighting coefficient; σ (·) represents an sigmoid function; C env ( e ij This represents the comprehensive marine environmental cost of the route segment; This represents the global average marine environmental cost; λ This represents the second weighting coefficient; Δ θ ij Indicates the connection node i Outbound heading angle and connecting nodes j The change in heading between the incoming heading angles; β Indicates the third weighting coefficient; f ij Indicates the frequency of vessel traffic on a particular section of the route; f max This indicates the maximum frequency of traffic across all air routes.

8. The method for obtaining ship entry and exit trajectories based on BeiDou positioning according to claim 1, characterized in that, The process of performing trajectory pattern recognition on the ship's sub-trajectory to obtain trajectory behavior patterns, evaluating the ship's sub-trajectory based on the trajectory behavior patterns, and correcting the ship's sub-trajectory based on the evaluation results includes: S41. Extract the ship behavior feature vector based on the ship sub-trajectory, and use Euclidean distance to perform trajectory pattern matching on the ship behavior feature vector to obtain the trajectory behavior pattern. S42. Based on the trajectory behavior pattern and the pre-acquired navigation rule base, the ship sub-trajectory is evaluated, and the evaluation results are used to judge the abnormal navigation sub-trajectory to obtain the abnormal ship sub-trajectory. S43. Use trajectory behavior patterns to correct the abnormal ship sub-trajectories, obtain the corrected sub-trajectories, and perform trajectory splicing based on the corrected sub-trajectories to obtain the optimized ship prediction trajectory.

9. A method for obtaining ship entry and exit trajectories based on BeiDou positioning according to claim 8, characterized in that, The process of correcting abnormal vessel sub-trajectories using trajectory behavior patterns to obtain corrected sub-trajectories, and then stitching together these corrected sub-trajectories to obtain optimized vessel prediction trajectories, includes: S431. Extract the spatiotemporal state data of ship trajectory points based on abnormal ship sub-trajectories, and construct boundary constraints for correcting the trajectory by combining the trajectory behavior pattern. S432. Based on the boundary constraints of the corrected trajectory and combined with the trajectory cost function, perform global trajectory reconstruction on the abnormal ship sub-trajectories to generate a candidate corrected trajectory set; S433. Calculate the trajectory deviation through candidate corrected trajectories, compare the trajectory deviation with the preset deviation threshold, use the comparison results to perform optimal screening of candidate corrected trajectories, obtain the corrected sub-trajectories, and perform trajectory splicing on the corrected sub-trajectories to obtain the optimized ship prediction trajectory.

10. A system for acquiring ship arrival and departure trajectories based on BeiDou positioning, used to implement the method for acquiring ship arrival and departure trajectories based on BeiDou positioning as described in any one of claims 1-9, characterized in that, The BeiDou-based system for acquiring ship entry and exit trajectories includes: a ship navigation feature extraction module, a ship trajectory prediction module, a ship trajectory segmentation module, and a ship trajectory correction module. The ship navigation feature extraction module is used to acquire ship navigation data, preprocess the ship navigation data to obtain preprocessed ship navigation data, and extract ship navigation features based on the preprocessed ship navigation data. The ship trajectory prediction module is used to filter the ship's navigation features to obtain navigation trajectory features, and to predict the ship's trajectory based on the navigation trajectory features to obtain the predicted ship trajectory. The ship trajectory division module is used to generate a sea area profile based on pre-acquired navigation sea area data, and to divide the predicted ship trajectory using the sea area profile to obtain ship sub-trajectories. The ship trajectory correction module is used to perform trajectory pattern recognition on the ship sub-trajectory, obtain trajectory behavior patterns, evaluate the ship sub-trajectory based on the trajectory behavior patterns, and correct the ship sub-trajectory based on the evaluation results.

Images