Ship anchoring behavior recognition and anchoring circle construction method

Abstract

The application discloses a ship anchoring behavior recognition and anchoring circle construction method, comprising the following steps: S1, acquiring a plurality of AIS trajectory data in a water area to be analyzed, and preprocessing the AIS trajectory data to obtain a plurality of trajectory segments; S2, determining an adaptive speed threshold based on the trajectory segments, and then obtaining a stay point of each trajectory segment; S3, performing geographical prior filtering on the stay point based on coastline data and anchorage boundary data, and removing a mooring stay point, performing S5 on an anchoring stay point, and performing S4 on a stay point of an undetermined category; S4, classifying the stay point based on a pre-trained K nearest neighbor classification model; removing the mooring stay point, and performing S5 on the anchoring stay point; S5, backtracking each anchoring stay point to obtain a corresponding trajectory point subset, and then performing two-stage filtering to obtain an effective trajectory point set, and constructing an envelope thereof as an anchoring circle. The method has strong applicability and high processing efficiency, and the constructed anchoring circle can completely express anchoring behavior.

Description

Technical Field

[0001] This invention relates to the fields of ship traffic management, intelligent maritime supervision, and geospatial information processing, specifically to a method for identifying ship anchoring behavior and constructing anchoring circles. Background Technology

[0002] Automatic Identification Systems (AIS) can collect information such as a vessel's position, time, speed over land, heading over land, vessel identification, vessel type, and length. Based on AIS trajectory identification of vessel anchoring behavior and the construction of anchorage circles, data support can be provided for anchorage supervision, port traffic organization, anchorage resource allocation, and maritime risk analysis. However, existing technologies have the following drawbacks:

[0003] 1. AIS trajectory-based stop point identification typically uses fixed ground speed thresholds and fixed spatial radii. These fixed parameters make it difficult for the identified stop points to adapt to different port environments and ship types, which can easily lead to missed or false detections of stop events.

[0004] 2. Typically, all stopping points are input into a classifier to separate anchoring and mooring behaviors, without making full use of prior geographical knowledge such as coastlines and anchorages for deterministic discrimination, resulting in a large computational load and low processing efficiency;

[0005] 3. The classification features used by the classifier have high requirements for the integrity of anchoring, making it difficult to distinguish between anchoring behavior and mooring behavior in the early stages of anchoring;

[0006] 4. Directly constructing a convex hull to form an anchoring circle based on the anchoring behavior's dwell point can easily lead to problems such as incomplete expression of anchoring behavior and an undersized anchoring circle area, failing to accurately reflect the actual spatial range of a single anchoring activity. Summary of the Invention

[0007] To address the problems existing in the prior art, this invention provides a method for identifying ship anchoring behavior and constructing anchoring rings. This method is highly applicable and can improve the processing efficiency of anchoring and mooring behavior classification and the accuracy of anchoring ring construction.

[0008] Therefore, the present invention adopts the following technical solution:

[0009] A method for identifying ship anchoring behavior and constructing anchoring circles includes the following steps:

[0010] S1, acquire several AIS trajectory data in the water area to be analyzed, each AIS trajectory data corresponds to a trajectory point, and preprocess the AIS trajectory data to obtain multiple trajectory segments;

[0011] S2: Based on the multiple trajectory segments obtained in S1, determine the adaptive speed threshold; perform speed filtering on each trajectory segment to construct a density interval, select the density space that meets the conditions as candidate stopping points, and then obtain the stopping point of the trajectory segment by density-first deduplication; thus, obtain the stopping points of all trajectory segments.

[0012] S3, based on coastline data and anchorage spatial data, performs geographical prior filtering on stopping points, eliminates moored stopping points, executes S5 on moored stopping points, and executes S4 on stopping points for which the corresponding vessel behavior category is not determined;

[0013] S4: Classify the stopping points based on a pre-trained K-nearest neighbor classification model to determine the ship behavior category corresponding to the stopping points; then, remove moored stopping points and perform S5 on anchored stopping points;

[0014] S5. Backtrack each anchorage point to obtain the corresponding subset of trajectory points, then perform two-level filtering to obtain the set of valid trajectory points, and construct the envelope of the set of valid trajectory points as the anchorage circle.

[0015] In step S1 above:

[0016] The AIS trajectory data includes the ship identifier, recording time, longitude, latitude, ground speed, ship type, and ship length of the corresponding trajectory point;

[0017] The preprocessing process includes:

[0018] First, the AIS trajectory data is sorted according to the ship identification and recording time, and AIS trajectory data with duplicate records, missing records, and invalid location records is deleted; AIS trajectory data with a ground speed exceeding 30 knots is deleted.

[0019] For AIS trajectory data of the same vessel, when the recording time interval between adjacent trajectory points exceeds a preset time interval threshold, segmentation is performed to obtain multiple trajectory segments;

[0020] Delete trajectory segments with fewer than the preset minimum number of points to complete the preprocessing of AIS trajectory data.

[0021] The method for determining the adaptive speed threshold in step S2 above is as follows:

[0022] Experts labeled stop event samples based on all trajectory segments obtained from S1;

[0023] The ground-to-ground speed of trajectory points in all the aforementioned dwell event samples is extracted and used to determine the adaptive speed threshold required for dwell identification through a two-level quantile statistical method. The specific steps are as follows:

[0024] First, for each dwell event sample, the 95th quantile of the ground-to-air velocity of all AIS trajectory points within that dwell event sample is calculated, and this quantile is used as the upper speed limit threshold for that dwell event sample. Then, the 95th quantile of the upper speed limit thresholds for all dwell event samples is calculated, and this quantile is used as the adaptive speed threshold. .

[0025] The process of velocity filtering for each trajectory segment in step S2 above is as follows:

[0026] For a certain ship's trajectory segment Perform speed filtering, removing those whose ground speed exceeds the adaptive speed threshold. The trajectory points are used to obtain the filtered trajectory segments. ,in, For the trajectory points.

[0027] The process of constructing the density interval in step S2 above is as follows:

[0028] Extract trajectory segments Corresponding ship type With the captain The adaptive radius of the ship is calculated according to the following formula. :

[0029] ,

[0030] Among them, the numerical codes 30, 60, 70, and 80 corresponding to each ship type are the ship type codes in the AIS standard;

[0031] Traverse the filtered trajectory segments For each trajectory point, the following processing is performed to obtain multiple density intervals:

[0032] With the filtered trajectory segment trajectory points Starting from the anchor point, the search is expanded along both the forward and backward time axes to construct density intervals. , Specifically, this involves checking the trajectory points sequentially. , If the spatial constraints are met, then the trajectory point will be... Included in density range Otherwise, stop expanding in that direction, where the formula for the spatial constraint is:

[0033] ,

[0034] in, For trajectory points and The Euclidean distance between them;

[0035] Record the density intervals obtained after expansion density value and duration Wherein, the density value Density range The total number of trajectory points included; the duration , Density range The earliest timestamp of the trajectory point is recorded in the middle. Density range The timestamp of the latest trajectory point is recorded in the middle.

[0036] The method for obtaining the stopping points corresponding to each trajectory segment in step S2 above is as follows:

[0037] For each density interval If the duration constraint and minimum density constraint are satisfied, then it is included as a candidate dwell point in the trajectory segment. The corresponding set of candidate stops In, among which:

[0038] The formula for the duration constraint is:

[0039] ,

[0040] in, This is the minimum duration that can be set.

[0041] The minimum density constraint formula is:

[0042] ,

[0043] in, This is the minimum number of trajectory points set.

[0044] For each candidate stop point, its centroid is recorded. Coordinates, duration oscillation radius and density value Wherein, the swing radius is the distance between the trajectory point and the centroid. The maximum distance between them;

[0045] Candidate stopping points set based on density value The candidate stop points are sorted in descending order, and the candidate stop point with the highest density value is retained as the trajectory segment. The point of stay.

[0046] Step S3 above includes:

[0047] Obtain the coastline data of the water area to be analyzed. Based on the prior knowledge that the nearshore waters are too shallow for large ships to anchor, perform prior coastline discrimination, as follows:

[0048] Calculate the Euclidean distance from the centroid of the resting point to the nearest coastline. ;

[0049] like If the vessel behavior category corresponding to the stopping point is determined to be mooring behavior, then the stopping point is a mooring stopping point and is removed.

[0050] When not satisfied When this happens, the following judgments are made:

[0051] If anchorage space data cannot be obtained, execute S4 for that stop point;

[0052] If anchorage space data can be obtained, prior knowledge of the operational constraints prohibiting mooring within the anchorage can be used to perform prior anchorage discrimination, as follows:

[0053] If the centroid of the stopping point is located within the anchorage area, the vessel behavior category corresponding to the stopping point is determined to be anchoring behavior, the stopping point is an anchoring stopping point, and S5 is executed for the anchoring stopping point;

[0054] When the candidate stopping point is located outside the anchorage area, S4 is executed for the stopping point;

[0055] When coastline and anchorage boundary data for the water area to be analyzed are unavailable, or the data is missing, inaccurate, outdated, or the spatial relationship between the stop point and the geographic object is unclear, S4 is executed directly for all stop points.

[0056] In step S4 above:

[0057] Points whose corresponding ship behavior category remains undetermined after S3 processing are used as fuzzy samples, and the average neighbor step size feature of the fuzzy samples is extracted. Mean centroid distance characteristics Space compactness index characteristics Characteristics of arc length to radius ratio The model inputs the scale-normalized K-nearest neighbor classification model to calculate the feature distance between the fuzzy sample and the training samples. Based on the feature distance, a preset number of training samples are selected as nearest neighbor samples. The category of the ship behavior corresponding to the fuzzy sample is determined as anchoring behavior or mooring behavior based on the category of the nearest neighbor samples. Then, the stopping point is determined as either an anchor stopping point or a mooring stopping point. If it is an anchor stopping point, S5 is executed; if it is a mooring stopping point, the mooring stopping point is removed.

[0058] In step S4 above:

[0059] The average neighbor step size feature The formula is:

[0060] ,

[0061] The average centroid distance feature The formula is:

[0062] ,

[0063] The space compactness index features The formula is:

[0064] ,

[0065] The arc length to radius ratio feature The formula is:

[0066] .

[0067] Step S5 above includes:

[0068] Backtracking is performed on the trajectory segment corresponding to the anchorage point. ~ All trajectory points within the time range form a subset of trajectory points. ;

[0069] For the subset of trajectory points Each trajectory point in the data undergoes an initial speed threshold screening. If the ground speed of a certain trajectory point exceeds the adaptive speed threshold... If so, mark it as a point to be determined;

[0070] Calculate the Euclidean distance between the point to be determined and the centroid of the corresponding resting point. ;like If the condition is met, the point to be determined is retained; otherwise, the point to be determined is discarded.

[0071] The above operations yield a set of valid trajectory points. ;

[0072] An anchoring stop corresponds to an anchoring event, based on the corresponding set of valid trajectory points. The convex hull of the mooring event is constructed using the Andrew monotonic chain algorithm.

[0073] Compared with the prior art, the present invention has the following beneficial effects:

[0074] 1. This invention constructs an adaptive speed threshold based on AIS trajectory data within the waters to be analyzed, and constructs an adaptive radius based on ship type and length. This effectively solves the problem that fixed parameters are difficult to adapt to differences in different ship types and port environments, and improves the ability to identify stopping points where stopping events occur under different ship types and port environments.

[0075] 2. This invention uses geographical prior filtering based on the coastline data and anchorage boundary data of the water area to be analyzed to quickly determine whether the vessel behavior corresponding to the stopping point is mooring or anchoring behavior; only stopping points whose corresponding vessel behavior category cannot be determined are treated as fuzzy samples and classified based on the K-nearest neighbor classification model, reducing the number of stopping points entering the K-nearest neighbor classification model and improving the classification efficiency of anchoring and mooring behavior.

[0076] 3. This invention introduces spatial compactness index and arc length-to-radius ratio features into the K-nearest neighbor classification model, which effectively improves the distinguishability between the initial stage of anchoring and mooring, and effectively solves the problem that the average neighbor step size feature and the average centroid distance feature are difficult to distinguish between the initial stage of anchoring and mooring.

[0077] 4. This invention obtains a subset of trajectory points by backtracking the anchoring points, then performs two-level filtering based on velocity and radius to obtain a set of effective trajectory points, and finally constructs an anchoring circle. This solves the problems of incomplete expression of anchoring behavior and small range in anchoring circles constructed by the traditional convex hull method. Attached Figure Description

[0078] Figure 1 This is a flowchart of the ship anchoring behavior recognition and anchoring circle construction method of the present invention;

[0079] Figure 2 In an embodiment of the present invention, a comparison diagram of the mooring circle constructed by the method of the present invention and the conventional convex hull method for trajectory segment 1 is shown, wherein, Figure (a) is the trajectory segment, Figure (b) is the mooring circle constructed by the conventional convex hull method, and Figure (c) is the mooring circle constructed by the method of the present invention.

[0080] Figure 3 In an embodiment of the present invention, a comparison diagram of the mooring circle constructed by the method of the present invention and the conventional convex hull method for trajectory segment 2 is shown, wherein, Figure (a) is the trajectory segment, Figure (b) is the mooring circle constructed by the conventional convex hull method, and Figure (c) is the mooring circle constructed by the method of the present invention;

[0081] Figure 4 In an embodiment of the present invention, a comparison diagram is shown of the anchoring circle constructed by the method of the present invention and the conventional convex hull method for trajectory segment 3. In the figure, (a) is the trajectory segment, (b) is the anchoring circle constructed by the conventional convex hull method, and (c) is the anchoring circle constructed by the method of the present invention. Detailed Implementation

[0082] The technical solution of the invention will be clearly and completely described below with reference to the accompanying drawings and embodiments. Obviously, the following embodiments are only some embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0083] like Figure 1 As shown, the ship anchoring behavior recognition and anchoring circle construction method of the present invention includes the following steps:

[0084] S1, Data Preprocessing:

[0085] Acquire several AIS trajectory data points within the waters to be analyzed. Each AIS trajectory data point corresponds to a trajectory point, including information such as the vessel identifier, recording time, longitude, latitude, ground speed, vessel type, and length.

[0086] The AIS trajectory data is preprocessed, including:

[0087] First, the AIS trajectory data is sorted according to the ship identification and recording time, and duplicate, missing, and invalid location records are deleted; then, AIS trajectory data with a ground speed exceeding 30 knots are deleted.

[0088] Then, based on the recording time of the AIS trajectory data, the AIS trajectory data of the same vessel is segmented according to a preset time interval threshold. Specifically, when the recording time interval between adjacent trajectory points exceeds the time interval threshold, segmentation is performed to obtain multiple trajectory segments. In one embodiment of the present invention, the time interval threshold is set to 60 minutes.

[0089] Track segments with fewer than a preset minimum number of points are deleted to complete the preprocessing of the AIS track data. In one embodiment of the present invention, the minimum number of points is set to 10.

[0090] S2 involves constrained stop point identification of the preprocessed AIS trajectory data obtained in S1. The specific steps are as follows:

[0091] S21, Determine the adaptive speed threshold:

[0092] Based on the preprocessed AIS trajectory data obtained from S1, i.e., based on the labeled stop event samples across all trajectory segments, experts extracted the ground speed of trajectory points from all the stop event samples. This ground speed speed was then used to determine the adaptive speed threshold required for stop identification through a two-layer quantile statistical method. The specific steps are as follows:

[0093] First, for each dwell event sample, the 95th quantile of the ground-to-ground speed of all trajectory points within that dwell event sample is calculated, and this quantile is used as the upper speed threshold for that dwell event sample. Then, the 95th quantile of the upper speed thresholds for all dwell event samples is calculated, and this quantile is used as the adaptive speed threshold. .

[0094] S22, velocity filtering is performed on each trajectory segment to construct density intervals, and density spaces that meet the conditions are selected as candidate stopping points. Then, density-first deduplication is used to obtain the stopping points corresponding to each trajectory segment, as follows:

[0095] For a certain ship's trajectory segment Perform speed filtering, removing those with ground speeds exceeding the adaptive speed threshold. The trajectory points are used to obtain the filtered trajectory segments. .in, For the trajectory points.

[0096] Extract the ship type of the vessel. With the captain The adaptive radius of the ship is calculated according to the following formula. :

[0097] ,

[0098] Among them, the numerical codes 30, 60, 70, and 80 corresponding to each ship type are the ship type codes in the AIS standard.

[0099] Traverse the filtered trajectory segments For each trajectory point, the following processing is performed to obtain multiple density intervals:

[0100] With the filtered trajectory segment trajectory points Starting from the anchor point, the search is expanded along both the forward and backward time axes to construct density intervals. , Specifically, this involves checking the trajectory points sequentially. ( If the spatial constraints are satisfied, then the trajectory point... Included in density range Otherwise, stop expanding in that direction, where:

[0101] The formula for the spatial constraint is as follows:

[0102] ,

[0103] in, For trajectory points and The Euclidean distance between them.

[0104] Record the density intervals obtained after expansion The density value (the total number of trajectory points included). and duration .in, , Density range The earliest timestamp of the trajectory point is recorded in the middle. Density range The timestamp of the latest trajectory point is recorded in the middle.

[0105] For each density interval If the duration constraint and minimum density constraint are satisfied, then it is included as a candidate dwell point in the trajectory segment. The corresponding set of candidate stops .in:

[0106] The formula for the duration constraint is as follows:

[0107] ,

[0108] in, This is the minimum duration that can be set.

[0109] The formula for the Minimum Density Constraint is as follows:

[0110] ,

[0111] in, This is the minimum number of trajectory points set. In one embodiment of the invention, .

[0112] The aforementioned duration and minimum density constraints ensure sufficient observational support to effectively filter out GPS drift and transient low-speed fluctuations.

[0113] For each candidate stop point, its centroid is recorded. Coordinates (mean of the sum of coordinates of trajectory points), duration , radius of oscillation (distance between trajectory point and centroid) (maximum distance between) and density value .

[0114] To ensure that each dwell event is identified only once, the candidate dwell point set is sorted by density value. The candidate stop points are sorted in descending order, and the candidate stop point with the highest density value (which best represents the core time period of the stop) is retained as the trajectory segment. The point of stay.

[0115] The above operations are used to obtain the stopping points of all trajectory segments.

[0116] S3, perform geographical prior filtering on stop points:

[0117] Obtain the coastline data of the water area to be analyzed. Based on the prior knowledge that the nearshore waters are too shallow for large ships to anchor, perform prior coastline discrimination, as follows:

[0118] Calculate the Euclidean distance from the centroid of the resting point to the nearest coastline. .

[0119] like If the vessel's behavior at the designated stopping point is classified as mooring behavior, then the stopping point is determined to be a mooring stopping point and is removed from the list. In one embodiment of the invention, a distance threshold is used. .

[0120] When not satisfied When this happens, the following judgments are made:

[0121] If anchorage space data cannot be obtained, execute S4 for that stop point;

[0122] If anchorage space data can be obtained, prior knowledge of the operational constraints prohibiting mooring within the anchorage can be used to perform prior anchorage discrimination, as follows:

[0123] If the centroid of the anchorage point is located within the anchorage area, the vessel behavior category corresponding to the anchorage point is determined to be anchoring behavior, and the anchorage point is an anchorage anchorage point. S5 is executed for the anchorage anchorage point.

[0124] When the stopping point is outside the anchorage area, S4 is executed for the stopping point.

[0125] If the coastline data and anchorage boundary data of the water area to be analyzed cannot be obtained (in the case of large-scale cross-regional studies or severe data shortages), or if the data is missing, inaccurate, outdated, or the spatial relationship between the stop point and the geographic object is unclear, then S4 will be executed directly for all stop points.

[0126] S4 classifies the stopping points based on a pre-trained K-nearest neighbor classification model to determine the corresponding ship behavior category; then, moored stopping points are removed, and S5 is performed on anchored stopping points. The details are as follows:

[0127] Points where the corresponding ship behavior category (mooring or anchoring) remains undetermined after S3 processing are taken as fuzzy samples, and the average neighbor step size feature of the fuzzy samples is extracted. Mean centroid distance characteristics Space compactness index characteristics Characteristics of arc length to radius ratio The model inputs the scale-normalized data into a pre-trained K-nearest neighbor classification model. It calculates the feature distance between the fuzzy sample and the training samples, selects a predetermined number of training samples as nearest neighbors based on this feature distance, and determines whether the fuzzy sample corresponds to anchoring or mooring behavior based on the category of the nearest neighbors. This determines whether the stopping point is an anchoring stopping point or a mooring stopping point. If it is an anchoring stopping point, proceed to step S5; if it is a mooring stopping point, remove that mooring stopping point.

[0128] in:

[0129] The average neighbor step size feature The formula used to characterize the degree of local movement between adjacent trajectory points is:

[0130] ,

[0131] The average centroid distance feature The formula used to characterize the average spatial dispersion of the stationary trajectory points relative to the centroid of the stationary point is:

[0132] ,

[0133] The space compactness index features The formula used to characterize the concentration of stationary trajectory points within a spatial range is:

[0134] ,

[0135] The arc length to radius ratio feature The formula used to characterize the relationship between path complexity and spatial activity range of a dwell trajectory is:

[0136] .

[0137] S5. Backtracking each anchorage point yields a subset of corresponding trajectory points. This subset is then filtered in two stages to obtain a set of valid trajectory points. The envelope of this valid trajectory point set is then constructed as the anchorage circle, as detailed below:

[0138] Backtracking is performed on the trajectory segment corresponding to the anchorage point. ~ All trajectory points within the time range form a subset of trajectory points. For the subset of trajectory points Each trajectory point in the data undergoes an initial speed threshold screening. If the ground speed of a certain trajectory point exceeds the adaptive speed threshold... If it is, then mark it as a point to be determined.

[0139] Calculate the Euclidean distance between the point to be determined and the centroid of the corresponding resting point. .like If the speed is high, the point to be determined is retained because although the speed of the point to be determined is high, its spatial position has not yet left the range of the anchor chain constraint, so it is regarded as a normal speed fluctuation in the mooring swing; otherwise, the point to be determined is removed.

[0140] The above operations yield a set of valid trajectory points. .

[0141] An anchoring stop corresponds to an anchoring event, based on the corresponding set of valid trajectory points. The convex hull is constructed using the Andrew monotonic chain algorithm and used as the anchor circle for this anchoring event.

[0142] Example

[0143] To demonstrate the effectiveness of the method of the present invention, the following experiments were conducted:

[0144] (1) Using 1,048,561 AIS records from Zhoushan Port as the AIS trajectory data for the waters to be analyzed, the method of this invention effectively identified 2,306 anchorage points. The identification accuracy of the anchorage points was 97.34%. Compared with the case without geographical prior filtering, the method of this invention can reduce the load of K-nearest neighbor calculation by 43.8% while maintaining the classification accuracy, achieving a speedup of 2.12 times. Compared with the traditional convex hull method (directly constructing the convex hull of the anchorage points), the method of this invention effectively expands the effective trajectory points by 5.9% while maintaining geometric stability.

[0145] (2) Visualization of mooring circles: After performing steps S2 to S4 on trajectory segments 1, 2, and 3 respectively, mooring circles are constructed using the traditional convex hull method (directly constructing the convex hull of the stopping point) and the backtracking + two-level filtering method in step S5 of this invention. Trajectory segments 1, 2, and 3, and their corresponding mooring circles constructed using the traditional convex hull method and the method constructed in this invention are shown below. Figure 2 , Figure 3 and Figure 4 It can be seen that the anchoring circle constructed by the traditional convex hull method has some missing trajectory points in the expression of anchoring behavior, while the backtracking + two-level filtering method of this invention effectively recovers the missing trajectory points and improves the ability to accurately express the range of anchoring behavior.

Claims

1. A method for identifying ship anchoring behavior and constructing anchoring circles, characterized in that, Includes the following steps: S1, acquire several AIS trajectory data in the water area to be analyzed, each AIS trajectory data corresponds to a trajectory point, and preprocess the AIS trajectory data to obtain multiple trajectory segments; S2: Based on the multiple trajectory segments obtained in S1, determine the adaptive speed threshold; perform speed filtering on each trajectory segment to construct a density interval, select the density space that meets the conditions as candidate stopping points, and then obtain the stopping point of the trajectory segment by density-first deduplication; thus, obtain the stopping points of all trajectory segments. S3, based on coastline data and anchorage spatial data, performs geographical prior filtering on stopping points, eliminates moored stopping points, executes S5 on moored stopping points, and executes S4 on stopping points for which the corresponding vessel behavior category is not determined; S4: Classify the stopping points based on a pre-trained K-nearest neighbor classification model to determine the ship behavior category corresponding to the stopping points; then, remove moored stopping points and perform S5 on anchored stopping points; S5. Backtrack each anchorage point to obtain the corresponding subset of trajectory points, then perform two-level filtering to obtain the set of valid trajectory points, and construct the envelope of the set of valid trajectory points as the anchorage circle.

2. The method for identifying ship anchoring behavior and constructing anchoring circles according to claim 1, characterized in that, In S1: The AIS trajectory data includes the ship identifier, recording time, longitude, latitude, ground speed, ship type, and ship length of the corresponding trajectory point; The preprocessing process includes: First, the AIS trajectory data is sorted according to the ship identification and recording time, and AIS trajectory data with duplicate records, missing records, and invalid location records is deleted; AIS trajectory data with a ground speed exceeding 30 knots is deleted. For AIS trajectory data of the same vessel, when the recording time interval between adjacent trajectory points exceeds a preset time interval threshold, segmentation is performed to obtain multiple trajectory segments; Delete trajectory segments with fewer than the preset minimum number of points to complete the preprocessing of AIS trajectory data.

3. The method for identifying ship anchoring behavior and constructing anchorage circles according to claim 2, characterized in that, The method for determining the adaptive speed threshold in S2 is as follows: Experts labeled stop event samples based on all trajectory segments obtained from S1; The ground-to-ground speed of trajectory points in all the aforementioned dwell event samples is extracted and used to determine the adaptive speed threshold required for dwell identification through a two-level quantile statistical method. The specific steps are as follows: First, for each dwell event sample, the 95th percentile of the ground-to-air velocity of all AIS trajectory points within that dwell event sample is calculated, and this 95th percentile is used as the upper speed limit threshold for that dwell event sample. Then, the 95th percentile of the upper speed limit thresholds for all dwell event samples is calculated, and this 95th percentile is used as the adaptive speed threshold. .

4. The method for identifying ship anchoring behavior and constructing anchoring circles according to claim 3, characterized in that, The process of velocity filtering for each trajectory segment in S2 is as follows: For a certain ship's trajectory segment Perform speed filtering, removing those whose ground speed exceeds the adaptive speed threshold. The trajectory points are used to obtain the filtered trajectory segments. ,in, For the trajectory points.

5. The method for identifying ship anchoring behavior and constructing anchoring circles according to claim 4, characterized in that, The process of constructing density intervals in S2 is as follows: Extract trajectory segments Corresponding ship type With the captain The adaptive radius of the ship is calculated according to the following formula. : ， Among them, the numerical codes 30, 60, 70, and 80 corresponding to each ship type are the ship type codes in the AIS standard; Traverse the filtered trajectory segments For each trajectory point, the following processing is performed to obtain multiple density intervals: With the filtered trajectory segment trajectory points Starting from the anchor point, the search is expanded along both the forward and backward time axes to construct density intervals. , Specifically, this involves checking the trajectory points sequentially. , If the spatial constraints are met, then the trajectory point will be... Included in density range Otherwise, stop expanding in that direction, where the formula for the spatial constraint is: ， in, For trajectory points and The Euclidean distance between them; Record the density intervals obtained after expansion density value and duration Wherein, the density value Density range The total number of trajectory points included; the duration , Density range The earliest timestamp of the trajectory point is recorded in the middle. Density range The timestamp of the latest trajectory point is recorded in the middle.

6. The method for identifying ship anchoring behavior and constructing anchoring circles according to claim 5, characterized in that, The method for obtaining the stopping points corresponding to each trajectory segment in S2 is as follows: For each density interval If the duration constraint and minimum density constraint are satisfied, then it is included as a candidate dwell point in the trajectory segment. The corresponding set of candidate stops In, among which: The formula for the duration constraint is: ， in, This is the minimum duration that can be set. The minimum density constraint formula is: ， in, This is the minimum number of trajectory points set. For each candidate stop point, its centroid is recorded. Coordinates, duration oscillation radius and density value Wherein, the swing radius is the distance between the trajectory point and the centroid. The maximum distance between them; Candidate stopping points set based on density value The candidate stop points are sorted in descending order, and the candidate stop point with the highest density value is retained as the trajectory segment. The point of stay.

7. The method for identifying ship anchoring behavior and constructing anchoring circles according to claim 6, characterized in that, S3 include: Obtain the coastline data of the water area to be analyzed. Based on the prior knowledge that the nearshore waters are too shallow for large ships to anchor, perform prior coastline discrimination, as follows: Calculate the Euclidean distance from the centroid of the resting point to the nearest coastline. ; like If the vessel behavior category corresponding to the stopping point is determined to be mooring behavior, then the stopping point is a mooring stopping point and is removed. When not satisfied When this happens, the following judgments are made: If anchorage space data cannot be obtained, execute S4 for that stop point; If anchorage space data can be obtained, prior knowledge of the operational constraints prohibiting mooring within the anchorage can be used to perform prior anchorage discrimination, as follows: If the centroid of the stopping point is located within the anchorage area, the vessel behavior category corresponding to the stopping point is determined to be anchoring behavior, the stopping point is an anchoring stopping point, and S5 is executed for the anchoring stopping point; When the candidate stopping point is located outside the anchorage area, S4 is executed for the stopping point; When coastline and anchorage boundary data for the water area to be analyzed are unavailable, or the data is missing, inaccurate, outdated, or the spatial relationship between the stop point and the geographic object is unclear, S4 is executed directly for all stop points.

8. The method for identifying ship anchoring behavior and constructing anchoring circles according to claim 7, characterized in that, In S4: Points whose corresponding ship behavior category remains undetermined after S3 processing are used as fuzzy samples, and the average neighbor step size feature of the fuzzy samples is extracted. Mean centroid distance characteristics Space compactness index characteristics Characteristics of arc length to radius ratio The model inputs the scale-normalized K-nearest neighbor classification model to calculate the feature distance between the fuzzy sample and the training samples. Based on the feature distance, a preset number of training samples are selected as nearest neighbor samples. The category of the ship behavior corresponding to the fuzzy sample is determined as anchoring behavior or mooring behavior based on the category of the nearest neighbor samples. Then, the stopping point is determined as either an anchor stopping point or a mooring stopping point. If it is an anchor stopping point, S5 is executed; if it is a mooring stopping point, the mooring stopping point is removed.

9. The method for identifying ship anchoring behavior and constructing anchoring circles according to claim 8, characterized in that, In S4: The average neighbor step size feature The formula is: ， The average centroid distance feature The formula is: ， The space compactness index features The formula is: ， The arc length to radius ratio feature The formula is: 。 10. The method for identifying ship anchoring behavior and constructing anchoring circles according to claim 9, characterized in that, S5 include: Backtracking is performed on the trajectory segment corresponding to the anchorage point. ~ All trajectory points within the time range form a subset of trajectory points. ; For the subset of trajectory points Each trajectory point in the data undergoes an initial speed threshold screening. If the ground speed of a certain trajectory point exceeds the adaptive speed threshold... If so, mark it as a point to be determined; Calculate the Euclidean distance between the point to be determined and the centroid of the corresponding resting point. ;like If the condition is met, the point to be determined is retained; otherwise, the point to be determined is discarded. The above operations yield a set of valid trajectory points. ; An anchoring stop corresponds to an anchoring event, based on the corresponding set of valid trajectory points. The convex hull of the mooring event is constructed using the Andrew monotonic chain algorithm.

Images