Vessel dynamic territorial identification methods and electronic equipment
By using multi-source data fusion and spatiotemporal clustering technology, the problem of distinguishing between short-term transit and long-term jurisdiction in determining the jurisdiction of fishing vessels has been solved, achieving accuracy and traceability of dynamic jurisdiction identification and supporting the needs of fisheries supervision.
Patent Information
- Application Number
- CN202611349470.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-09-02
- Publication Date
- 2026-11-17
AI Technical Summary
Existing technologies rely heavily on static vessel registration or single spatial trajectory clustering for determining the territorial status of fishing vessels. This makes it difficult to distinguish between short-term transit and long-term territorial behavior, and it cannot output dynamic territorial results with effective time periods, thus failing to meet the needs of cross-regional fisheries supervision.
By acquiring multi-source data, including ship static archive data and spatiotemporal behavior data, abnormal data processing and time window segmentation are performed to identify berthing behavior and operational behavior. Spatiotemporal joint distance is constructed for clustering, the comprehensive territorial score is calculated, and combined with evidence sufficiency verification, dynamic territorial identification results with effective time interval and confidence level are generated.
It has improved the accuracy of territorial determination, made the results dynamically traceable, reduced the risk of misjudgment, and can distinguish between long-term primary territorial jurisdiction, seasonal temporary territorial jurisdiction and short-term transit areas, supporting cross-regional fishery administration operations and the tracing of illegal vessels.
Smart Images

Figure CN123116086A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of ship data processing technology, and more specifically, to a method and electronic device for identifying the dynamic homeland of a ship. Background Technology
[0002] In related technologies, current determination of fishing vessel territoriality mostly relies on static vessel registration or single spatial trajectory clustering, which has limited dimensions and makes it difficult to distinguish between short-term transit and long-term territorial behavior. It is also unable to output dynamic territorial results with effective time periods, making it difficult to meet the needs of cross-regional fisheries supervision. Summary of the Invention
[0003] The present invention aims to solve at least one of the technical problems existing in the prior art or related art.
[0004] Therefore, the first aspect of the present invention proposes a method for identifying the dynamic homeland of a ship.
[0005] A second aspect of the present invention provides an electronic device.
[0006] In view of this, a first aspect of the present invention provides a method for identifying the dynamic territorial affiliation of a vessel, comprising: acquiring multi-source data of a target vessel, the multi-source data including at least static vessel data and spatiotemporal behavior data, the spatiotemporal behavior data including multiple behavioral points arranged in chronological order; performing anomaly processing on the spatiotemporal behavior data and dividing the processed spatiotemporal behavior data into multiple time windows; within each time window, identifying the berthing and operational behaviors of the target vessel based on the behavioral points, segmenting the spatiotemporal behavior data into voyage trajectories, and extracting berthing behavior features corresponding to berthing areas and operational distribution features corresponding to operational sea areas from the segmented voyage trajectory data; and based on the relationships between behavioral points... The system constructs a spatiotemporal joint distance based on geographic spatial distance and temporal continuous distance. Based on this joint distance, behavioral points are clustered to obtain stable behavioral clusters, and candidate geographic regions are identified based on these clusters. The system calculates the comprehensive geographic score for each candidate geographic region based on the characteristics of affiliation behavior and work distribution. Candidate primary geographic regions are determined based on their comprehensive geographic scores, and the sufficiency of primary geographic evidence is verified by combining the amount of effective behavioral data and data timeliness within the current time window. Based on the candidate primary geographic regions, comprehensive geographic scores, and the sufficiency of primary geographic evidence verification results for multiple consecutive time windows, the time windows are merged or segmented to generate dynamic geographic identification results with effective time intervals and confidence levels.
[0007] The vessel dynamic territorial identification method provided in this application constructs a full-link judgment system that integrates multi-source data fusion input, abnormal data preprocessing, time window granularity division, voyage behavior feature extraction, spatiotemporal dual-dimensional clustering candidates, multi-dimensional quantitative comprehensive scoring, primary territorial evidence sufficiency verification, and dynamic time segmentation output. It also integrates vessel static registration attributes and dynamic spatiotemporal behavior data, incorporating multi-dimensional information such as time continuity, port berthing patterns, operational sea area distribution, and data evidence strength into the judgment logic. Ultimately, this achieves improved accuracy in territorial determination, dynamic traceability of territorial results, reduced risk of misjudgment, and enhanced fault tolerance in abnormal scenarios. It can distinguish between long-term primary territorial ...
[0008] By acquiring multi-source data of the target vessel, including at least static vessel profile data and spatiotemporal behavior data, the spatiotemporal behavior data comprises multiple behavioral points arranged chronologically. Simultaneously, both static registration data and dynamic behavior data are integrated, breaking the limitation of relying on only a single data dimension. The static vessel profile data provides basic reference information such as registration location and fishing permit scope, while the spatiotemporal behavior data reflects the vessel's actual navigation, berthing, and operational activity patterns. These two data sources complement and cross-validate each other, building a comprehensive data foundation for accurate subsequent location determination and reducing the judgment bias caused by purely matching registration information at the data source level.
[0009] This solution employs anomaly processing on vessel spatiotemporal behavior data and divides the processed data into multiple time windows. On one hand, anomaly processing filters out invalid interference data such as coordinate drift, message duplication, and time reversal, improving the quality and reliability of the basic data and reducing the interference of abnormal messages on subsequent trajectory analysis and feature statistics. On the other hand, by dividing time windows according to preset rules, long-cycle continuous trajectories are decomposed into statistical units of uniform granularity. This not only adapts to the seasonal patterns of fishery production but also provides granular support for subsequent time-segmented statistical analysis of behavioral characteristics and output of dynamic localization results, enabling the localization determination results to accurately correspond to specific time periods.
[0010] This solution identifies the berthing and operational behaviors of target vessels within each time window based on behavioral points. It segments the vessel's spatiotemporal behavior data into voyage trajectories and extracts technical features from these segmented voyage trajectory data. First, it divides the single voyage trajectory using departure and arrival events as boundaries, ensuring each trajectory segment corresponds to a complete departure-voyage-operation-arrival business cycle, resulting in feature statistics with clear business semantics. Simultaneously, it extracts behavioral features from two independent dimensions: port berthing and operational sea area. This characterizes the correlation strength between the vessel and specific areas from two perspectives: the resident port where the vessel berths for replenishment and the core sea area where the vessel operates. Compared to single-dimensional feature determination, this approach more comprehensively and accurately reflects the vessel's territorial affiliation, avoiding the bias caused by relying solely on berthing or operational behaviors.
[0011] This solution constructs a spatiotemporal joint distance based on the geographic and temporal distances between behavioral points. It then clusters these behavioral points to obtain stable behavioral clusters. Furthermore, it identifies candidate geographic regions based on these stable behavioral clusters, incorporating both temporal continuity and spatial proximity into the clustering metric. Unlike clustering algorithms based solely on spatial distance, this approach effectively distinguishes between short-term transit behaviors that overlap spatially but are temporally discrete, and long-term stable activities that are spatially concentrated and temporally continuous. The clustering process only identifies stable behavioral clusters that meet the conditions of spatial neighborhood, temporal continuity, and minimum number of behavioral points. Only the regions corresponding to these stable behavioral clusters are included in the candidate geographic regions, directly filtering out scattered stops and short-term transit areas. This reduces the possibility of misjudgment from the candidate selection stage, fundamentally solving the core problem of spatial clustering's inability to distinguish between short-term transit and long-term geographic regions.
[0012] This solution calculates the comprehensive territorial score for each candidate territorial region based on the characteristics of vessel affiliation and operational distribution. It integrates multi-dimensional behavioral features through normalization and weighting into a quantifiable and ranking comprehensive territorial score, making the territorial determination process transparent and interpretable, replacing the fuzzy determination method that relies on human experience. Simultaneously, the quantitative scoring model supports automated territorial identification for batches of vessels, significantly improving the processing efficiency of fisheries supervision and adapting to batch supervision scenarios with massive amounts of vessel trajectory data.
[0013] In some technical solutions of this application, abnormal data processing is performed on ship spatiotemporal behavior data, specifically including: removing abnormal points from the ship spatiotemporal behavior data based on the geographical location, speed, and timestamp of the behavior points, wherein the abnormal points include at least coordinate boundary crossing points, speed abnormal points, time reversal points, duplicate message points, and position abrupt drift points after positioning interruption; after removing the abnormal points, interactive multi-model filtering is used to smooth the trajectory of the remaining behavior points; wherein the motion model used in interactive multi-model filtering includes at least two of the following: uniform motion model, turning motion model, and low-speed mooring model.
[0014] In some technical solutions of this application, the ship spatiotemporal behavior data also includes the AIS ship name field, the ship static file data includes the registration location, and the ship dynamic location identification method also includes: performing character cleaning and format standardization processing on the AIS ship name field, extracting the pinyin characters at the beginning of the standardized AIS ship name as the ship name pinyin prefix; according to the preset mapping relationship between the pinyin prefix and the administrative division, resolving the ship name pinyin prefix to at least one identifier resolution location and recording the resolution confidence; when the ship name pinyin prefix can be mapped to multiple different administrative divisions, retaining multiple candidate identifier resolution locations, or marking the identifier resolution location as pending confirmation.
[0015] In some technical solutions of this application, the multi-source data also includes port business data, which contains port names from multiple sources; the method also includes standardizing the port names in the port business data, specifically by: mapping port names from different sources to a unified port code through port code matching, port name prefix matching, and character sequence similarity matching in sequence according to a preset priority; stopping subsequent matching steps when the matching confidence of the current matching method reaches the confidence threshold corresponding to the matching method.
[0016] In some technical solutions of this application, identifying the berthing and operational behaviors of target vessels and segmenting the vessel's spatiotemporal behavior data into voyage trajectories specifically includes: identifying berthing events based on the geofence of a preset port, combined with the speed and duration of the behavioral points, and correspondingly determining arrival and departure events; segmenting continuous vessel spatiotemporal behavior data into multiple single-voyage trajectory segments using adjacent departure and arrival events as boundaries; berthing behavior characteristics include at least two of the following: total cumulative berthing time, berthing frequency, average berthing time per berth, continuous berthing time, number of berthings at fixed berths, and the proportion of replenishment and unloading behaviors; operational distribution characteristics include at least two of the following: cumulative grid operational time, operational point density, number of consecutive operational days, seasonal operational distribution, and the proportion of cross-sea area navigation; and dividing and statistically analyzing the operational sea area using equal-area grid units under a projected coordinate system.
[0017] In some technical solutions of this application, a spatiotemporal joint distance is constructed based on the geospatial distance and temporal continuity distance between behavioral points. The behavioral points are then clustered based on this spatiotemporal joint distance. Specifically, this includes: selecting any two behavioral points within the same time window, calculating the geospatial distance and temporal continuity distance between the two behavioral points respectively, and standardizing both the geospatial distance and temporal continuity distance; weighting and summing the square values of the standardized geospatial distance and the standardized temporal continuity distance according to spatial and temporal weight coefficients, and taking the square root to obtain the spatiotemporal joint distance; combining spatial neighborhood thresholds, temporal continuity thresholds, and the minimum number of behavioral points, density clustering is performed on the behavioral points within the same time window, and the set of behavioral points that meets the conditions is determined as a stable behavioral cluster; the geospatial distance is calculated using Haversine great circle distance, or the planar distance is calculated after conversion to a projected coordinate system.
[0018] In some technical solutions of this application, the comprehensive score of each candidate territory is calculated, specifically including: for each candidate territory, normalizing and weighting the characteristics of docking behavior and operation distribution to obtain the docking behavior score and operation sea area association score corresponding to the candidate territory; determining the continuity bonus based on the continuous duration of stable behavior clusters in the candidate territory, and determining the short-term transit penalty item based on the docking duration, docking frequency and operation cluster formation in the candidate territory; after weighted summing of the docking behavior score, operation sea area association score and continuity bonus, subtracting the weighted short-term transit penalty item to obtain the initial territory score corresponding to the candidate territory; obtaining the time difference between the current time window and the most recent valid behavior record in the candidate territory, generating an exponential time decay factor based on the time difference and a preset decay coefficient, multiplying the initial territory score by the time decay factor to obtain the comprehensive territory score after historical decay correction; when the time difference exceeds a preset failure threshold, increasing the decay coefficient, or marking the candidate territory as a failed candidate territory.
[0019] In some technical solutions of this application, the sufficiency of evidence for candidate primary home locations is verified by combining the amount of effective behavioral data and data timeliness within the current time window. Specifically, this includes: sorting each candidate home location region from high to low according to the comprehensive home location score, and determining the candidate home location region corresponding to the highest comprehensive home location score as the candidate primary home location; classifying the candidate primary home location according to the home location score level, score leadership, amount of effective behavioral data, and data timeliness as verification dimensions: when the highest comprehensive home location score reaches the first score threshold, the difference between the highest comprehensive home location score and the second highest comprehensive home location score reaches the leading difference threshold, and the number of effective positioning points, effective flight times, effective docking times, and effective operation points within the current time window are verified. When the quantity of evidence reaches the corresponding minimum evidence threshold and the data time interval is not greater than the time limit threshold, the sufficiency of the primary territory evidence is deemed to have passed, the candidate primary territory is determined as the actual activity territory, and the determination status is set to confirmed. When the highest territory comprehensive score reaches the second score threshold which is lower than the first score threshold, or when some conditions of effective behavioral data quantity or data time limit are not met, the determination status is set to low confidence. When the highest territory comprehensive score is lower than the second score threshold, or when any evidence indicator fails to reach the threshold or the data time limit is exceeded, the determination status is set to pending confirmation or insufficient data. If the business system must display the reference territory, the registered territory in the ship static file data is used as the reference territory output, and the source of the reference territory is marked as registration information.
[0020] In some technical solutions of this application, based on the candidate primary affiliation, comprehensive affiliation score, and primary affiliation evidence sufficiency verification results corresponding to multiple consecutive time windows, the time windows are merged or segmented to generate dynamic affiliation identification results with effective time intervals and confidence levels. Specifically, this includes: comparing the candidate primary affiliation and comprehensive affiliation scores corresponding to two adjacent time windows; merging the two adjacent time windows when the candidate primary affiliation is the same and the change in the corresponding comprehensive affiliation score does not reach the segmentation change threshold; setting a affiliation time boundary point between the two adjacent time windows when the candidate primary affiliation is different and at least one of the difference in the corresponding comprehensive affiliation score, the change in the candidate affiliation regional ranking, or the change in confidence level reaches the segmentation change threshold; and generating an effective time interval and confidence level based on the affiliation time boundary point. The start time and end time of failure are defined. When the same candidate territory repeatedly forms a stable behavior cluster within multiple time windows corresponding to the same fishing season, but does not meet the main territory determination conditions within a continuous year-round time window, the candidate territory is marked as a seasonal temporary territory. After generating dynamic territory identification results, the actual activity territory determined by the model is extracted from the dynamic territory identification results. The actual activity territory, the identifier resolution territory, and the registered territory are compared pairwise. When the geographical distance, administrative division level difference, inconsistency duration, and corresponding confidence level between any two types of territories meet the territory conflict conditions, a territory conflict mark is generated, and the conflict type, conflict level, involved territories, and evidence source are output. Conflict types include conflict between identifier resolution territory and registered territory, conflict between identifier resolution territory and actual activity territory, and conflict between registered territory and actual activity territory.
[0021] A second aspect of the present invention provides an electronic device comprising: a memory for storing a computer program and a processor for executing the computer program to implement the ship dynamic territorial identification method as described in any of the above-described technical solutions. Therefore, the electronic device possesses all the beneficial effects of the ship dynamic territorial identification method as described in any of the above-described technical solutions.
[0022] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description
[0023] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which:
[0024] Figure 1 This is a flowchart illustrating a method for identifying the dynamic homeland of a ship according to an embodiment of the present invention.
[0025] Figure 2This is a schematic diagram of trajectory segmentation and behavior recognition according to an embodiment of the present invention;
[0026] Figure 3 This is a schematic diagram of spatiotemporal joint clustering and candidate location identification according to an embodiment of the present invention;
[0027] Figure 4 This is a schematic diagram of territorial scoring, evidence verification, and dynamic time slicing according to an embodiment of the present invention;
[0028] Figure 5 This is a schematic block diagram of an electronic device according to an embodiment of the present invention;
[0029] Figure 6 This is a technical architecture diagram of a ship dynamic homeland identification process according to an embodiment of the present invention;
[0030] Figure 7 This is one of the technical architecture diagrams for the ship name origination resolution, registration verification, and anomaly / low confidence correction process according to an embodiment of the present invention;
[0031] Figure 8 This is the second technical architecture diagram of the ship name origin resolution, registration verification, and anomaly / low confidence correction process according to an embodiment of the present invention. Detailed Implementation
[0032] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.
[0033] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.
[0034] The following reference Figures 1 to 8 A method and electronic device for identifying the dynamic homeland of a ship according to some embodiments of the present invention are described.
[0035] like Figure 1 As shown, an embodiment of this application provides a method for identifying the dynamic homeland of a ship, including:
[0036] Step 102: Obtain multi-source data of the target vessel. The multi-source data includes at least static archive data of the vessel and spatiotemporal behavior data of the vessel. The spatiotemporal behavior data of the vessel includes multiple behavior points arranged in chronological order.
[0037] Step 104: Perform anomaly processing on the ship spatiotemporal behavior data and divide the processed ship spatiotemporal behavior data into multiple time windows;
[0038] Step 106: Within each time window, identify the berthing and operation behaviors of the target vessel based on the behavior points, segment the vessel's spatiotemporal behavior data into voyage trajectories, and extract the berthing behavior characteristics corresponding to the berthing area and the operation distribution characteristics corresponding to the operation sea area from the segmented voyage trajectory data.
[0039] Step 108: Construct a spatiotemporal joint distance based on the geographic spatial distance and temporal continuous distance between behavior points, cluster the behavior points based on the spatiotemporal joint distance to obtain stable behavior clusters, and identify candidate geographical regions based on the stable behavior clusters.
[0040] Step 110: Calculate the comprehensive score of each candidate jurisdiction based on the characteristics of affiliation behavior and the characteristics of work distribution;
[0041] Step 112: Determine the candidate primary location based on the comprehensive score of the location, and verify the sufficiency of the primary location evidence for the candidate primary location by combining the amount of effective behavioral data and the timeliness of the data within the current time window;
[0042] Step 114: Based on the candidate primary location, comprehensive location score, and primary location evidence sufficiency verification results corresponding to multiple consecutive time windows, merge or divide the time windows to generate dynamic location identification results with effective time intervals and confidence levels.
[0043] The vessel dynamic territorial identification method provided in this application constructs a full-link judgment system that integrates multi-source data fusion input, abnormal data preprocessing, time window granularity division, voyage behavior feature extraction, spatiotemporal dual-dimensional clustering candidates, multi-dimensional quantitative comprehensive scoring, primary territorial evidence sufficiency verification, and dynamic time segmentation output. It also integrates vessel static registration attributes and dynamic spatiotemporal behavior data, incorporating multi-dimensional information such as time continuity, port berthing patterns, operational sea area distribution, and data evidence strength into the judgment logic. Ultimately, this achieves improved accuracy in territorial determination, dynamic traceability of territorial results, reduced risk of misjudgment, and enhanced fault tolerance in abnormal scenarios. It can distinguish between long-term primary territorial ...
[0044] By acquiring multi-source data of the target vessel, including at least static vessel profile data and spatiotemporal behavior data, the spatiotemporal behavior data comprises multiple behavioral points arranged chronologically. Simultaneously, both static registration data and dynamic behavior data are integrated, breaking the limitation of relying on only a single data dimension. The static vessel profile data provides basic reference information such as registration location and fishing permit scope, while the spatiotemporal behavior data reflects the vessel's actual navigation, berthing, and operational activity patterns. These two data sources complement and cross-validate each other, building a comprehensive data foundation for accurate subsequent location determination and reducing the judgment bias caused by purely matching registration information at the data source level.
[0045] This solution employs anomaly processing on vessel spatiotemporal behavior data and divides the processed data into multiple time windows. On one hand, anomaly processing filters out invalid interference data such as coordinate drift, message duplication, and time reversal, improving the quality and reliability of the basic data and reducing the interference of abnormal messages on subsequent trajectory analysis and feature statistics. On the other hand, by dividing time windows according to preset rules, long-cycle continuous trajectories are decomposed into statistical units of uniform granularity. This not only adapts to the seasonal patterns of fishery production but also provides granular support for subsequent time-segmented statistical analysis of behavioral characteristics and output of dynamic localization results, enabling the localization determination results to accurately correspond to specific time periods.
[0046] This solution identifies the berthing and operational behaviors of target vessels within each time window based on behavioral points. It segments the vessel's spatiotemporal behavior data into voyage trajectories and extracts technical features from these segmented voyage trajectory data. First, it divides the single voyage trajectory using departure and arrival events as boundaries, ensuring each trajectory segment corresponds to a complete departure-voyage-operation-arrival business cycle, resulting in feature statistics with clear business semantics. Simultaneously, it extracts behavioral features from two independent dimensions: port berthing and operational sea area. This characterizes the correlation strength between the vessel and specific areas from two perspectives: the resident port where the vessel berths for replenishment and the core sea area where the vessel operates. Compared to single-dimensional feature determination, this approach more comprehensively and accurately reflects the vessel's territorial affiliation, avoiding the bias caused by relying solely on berthing or operational behaviors.
[0047] This solution constructs a spatiotemporal joint distance based on the geographic and temporal distances between behavioral points. It then clusters these behavioral points to obtain stable behavioral clusters. Furthermore, it identifies candidate geographic regions based on these stable behavioral clusters, incorporating both temporal continuity and spatial proximity into the clustering metric. Unlike clustering algorithms based solely on spatial distance, this approach effectively distinguishes between short-term transit behaviors that overlap spatially but are temporally discrete, and long-term stable activities that are spatially concentrated and temporally continuous. The clustering process only identifies stable behavioral clusters that meet the conditions of spatial neighborhood, temporal continuity, and minimum number of behavioral points. Only the regions corresponding to these stable behavioral clusters are included in the candidate geographic regions, directly filtering out scattered stops and short-term transit areas. This reduces the possibility of misjudgment from the candidate selection stage, fundamentally solving the core problem of spatial clustering's inability to distinguish between short-term transit and long-term geographic regions.
[0048] This solution calculates the comprehensive territorial score for each candidate territorial region based on the characteristics of vessel affiliation and operational distribution. It integrates multi-dimensional behavioral features through normalization and weighting into a quantifiable and ranking comprehensive territorial score, making the territorial determination process transparent and interpretable, replacing the fuzzy determination method that relies on human experience. Simultaneously, the quantitative scoring model supports automated territorial identification for batches of vessels, significantly improving the processing efficiency of fisheries supervision and adapting to batch supervision scenarios with massive amounts of vessel trajectory data.
[0049] This solution determines candidate primary locations based on a comprehensive score and verifies the sufficiency of evidence by combining the amount of valid behavioral data and data timeliness within the current time window. In addition to the score ranking, it adds a data evidence strength verification step, controlling the reliability of the determination from two dimensions: the scale of valid data and data freshness. Only when both the data volume and timeliness meet the threshold requirements is a definitive primary location conclusion output. This avoids forcibly outputting a definitive conclusion in scenarios with insufficient trajectory points, inadequate stop records, or long-term data stagnation. From a rule-based perspective, this ensures the rigor of the location determination results and reduces the risk of misjudgment in sparse sample scenarios.
[0050] This solution utilizes the technical features of merging or segmenting time windows based on candidate primary jurisdictions, comprehensive jurisdictional scores, and the sufficiency of primary jurisdictional evidence across multiple consecutive time windows. This generates dynamic jurisdictional identification results with effective time intervals and confidence levels. By comparing and merging jurisdictional results from adjacent time windows, discrete window-level determination results are stitched together into continuous time interval-level jurisdictional records, outputting dynamic jurisdictional labels with clearly defined start and end times and confidence levels. This design overcomes the limitation of only outputting fixed, static jurisdictional labels, accurately reflecting jurisdictional changes caused by fishing vessels operating across seasons and sea areas, and supporting the determination of jurisdictional responsibility and law enforcement tracing at different time periods.
[0051] In some embodiments of this application, abnormal data processing is performed on ship spatiotemporal behavior data, specifically including: removing abnormal points from the ship spatiotemporal behavior data based on the geographical location, speed, and timestamp of the behavior points, wherein the abnormal points include at least coordinate boundary crossing points, speed abnormal points, time reversal points, duplicate message points, and position abrupt drift points after positioning interruption; after removing the abnormal points, interactive multi-model filtering is used to smooth the trajectory of the remaining behavior points; wherein the motion model used in interactive multi-model filtering includes at least two of the following: uniform motion model, turning motion model, and low-speed mooring model.
[0052] In the above embodiments, the specific implementation method of abnormal data processing is defined. This solution utilizes the technical characteristics of identifying and eliminating five typical abnormal points based on the geographical location, speed, and timestamp of the behavior points. For common abnormal scenarios in AIS data, such as equipment failure, signal obstruction, message duplication, and positioning drift, multi-dimensional anomaly identification rules are designed: coordinate boundary points correspond to erroneous coordinates that exceed the latitude and longitude range of the target sea area; speed anomaly points correspond to messages that exceed the reasonable speed range of fishing vessels; time reversal points correspond to messages with disordered timestamp order; duplicate message points correspond to redundant data repeatedly reported by the same vessel at the same time; and position change drift points after positioning interruption correspond to position data that has changed significantly after a long period of signal loss. By accurately identifying and eliminating abnormal points through the above rules, the interference of obvious erroneous data on subsequent trajectory analysis, feature statistics, and clustering calculations can be directly eliminated, thereby improving the reliability of basic data.
[0053] Building upon this foundation, this solution utilizes interactive multi-model filtering to smooth the trajectory of remaining behavioral points. The filtering incorporates at least two motion models, running multiple motion models in parallel and switching weights in real time to adapt to the movement patterns of fishing vessels in different states, such as straight-line navigation, turning, and anchoring. Compared to a single filtering model, interactive multi-model filtering can more accurately fit the actual trajectory of the fishing vessel, correcting trajectory jitter caused by positioning errors. This results in subsequent extracted statistical results for features such as anchoring duration, operation duration, and voyage boundaries that are closer to reality. The two-stage preprocessing mechanism works together to effectively improve the solution's fault tolerance in complex data scenarios such as missing AIS signals, near-shore obstruction, and equipment failures, ensuring the quality of the basic data for subsequent end-to-end location determination.
[0054] In one embodiment, interactive multi-model filtering sets up three models in parallel: uniform motion model, turning motion model, and low-speed parking model. The process noise and observation noise parameters are set according to the historical AIS error statistics of the same sea area. For position change points that occur more than 2 hours after the positioning interruption, they are directly identified as drift points and removed to avoid large trajectory deviations caused by a single signal anomaly.
[0055] In some embodiments of this application, the ship spatiotemporal behavior data also includes an AIS ship name field, the ship static file data includes the registration location, and the ship dynamic location identification method further includes: performing character cleaning and format standardization processing on the AIS ship name field, extracting the pinyin characters at the beginning of the standardized AIS ship name as the ship name pinyin prefix; according to the preset mapping relationship between the pinyin prefix and the administrative division, resolving the ship name pinyin prefix to at least one identifier resolution location and recording the resolution confidence; when the ship name pinyin prefix can be mapped to multiple different administrative divisions, retaining multiple candidate identifier resolution locations, or marking the identifier resolution location as pending confirmation.
[0056] In the above embodiments, the ship's spatiotemporal behavior data also includes the AIS ship name field, the ship's static file data includes the registration location, and the ship's dynamic location identification method further includes: performing character cleaning and format standardization processing on the AIS ship name field, extracting the pinyin characters at the beginning of the standardized AIS ship name as the ship name pinyin prefix; according to the preset mapping relationship between the pinyin prefix and administrative divisions, resolving the ship name pinyin prefix to at least one identifier resolution location and recording the resolution confidence; when the ship name pinyin prefix can be mapped to multiple different administrative divisions, retaining multiple candidate identifier resolution locations, or marking the identifier resolution location as pending confirmation. This solution utilizes the technical feature of extracting the ship name pinyin prefix from the AIS ship name field and resolving it to an identifier resolution location to mine location clues from dynamically reported message data, constructing a third-party location reference dimension independent of static files and trajectory data. First, character cleaning and format standardization processes are used to remove special symbols, spaces, garbled characters, and case differences from the ship name field, converting it into a unified standard format. This eliminates format differences in the ship name field across different devices and reporting channels, ensuring the consistency and accuracy of prefix extraction. Then, a pre-maintained mapping dictionary of "pinyin prefix - province / region - administrative division code" is used to convert the extracted ship name pinyin prefix into the corresponding identifier resolution location. At the same time, a corresponding confidence level is matched for each resolution result. Preliminary location reference information can be quickly obtained through dynamic AIS messages without relying on a static archive.
[0057] For scenarios where prefixes are ambiguous, this solution utilizes the technical feature of retaining multiple candidate identifiers to resolve the place of origin or marking them as pending confirmation, without forcing a single-value judgment. This avoids resolution errors caused by overlapping naming rules and ensures the rigor of the ship name resolution results. The aforementioned ship name pinyin prefix resolution mechanism provides an independent source of clues for subsequent three-party cross-verification of place of origin, helping to detect issues such as ship identifier misuse, AIS ship name tampering, and abnormal registration information, thereby improving the ability to identify abnormal ships.
[0058] In one embodiment, the mapping dictionary maintains the mapping relationship between the abbreviations of provincial-level administrative regions and key cities across the country, for example, "su" corresponds to region A and "yue" corresponds to region B; for prefixes that correspond to only a single administrative division, the resolution confidence level is set to above 0.9; for ambiguous prefixes that correspond to multiple administrative divisions, a corresponding confidence level is set for each candidate location, and they are uniformly marked as pending manual review.
[0059] In some embodiments of this application, the multi-source data also includes port business data, which contains port names from multiple sources; the method further includes standardizing the port names in the port business data, specifically by: mapping port names from different sources to a unified port code through port code matching, port name prefix matching, and character sequence similarity matching in sequence according to a preset priority; stopping subsequent matching steps when the matching confidence of the current matching method reaches the confidence threshold corresponding to the matching method.
[0060] In the above embodiments, the multi-source data also includes port business data, which contains port names from multiple sources. The method further includes standardizing the port names in the port business data. Specifically, the steps involve mapping port names from different sources to a unified port code according to a preset priority, using port code matching, port name prefix matching, and character sequence similarity matching. When the matching confidence of the current matching method reaches the confidence threshold corresponding to that method, subsequent matching steps are stopped. This solution utilizes the technical features of a four-layer progressive port name matching framework to solve the problem of inconsistent port name representations from multiple sources. Port business data can come from multiple channels such as the fishing port scheduling system, the maritime port entry and exit reporting system, and the fishery law enforcement system. The port names in different channels have problems such as differences between the full name and the abbreviation, differences between different names of the same port, and differences in typos. Direct use will lead to confusion in the statistics of port berthing characteristics. This solution sets up three automatic matching levels in order of accuracy from high to low and priority from high to low: port code matching, port name prefix matching, and character sequence similarity matching. It is supplemented by manual correction as a backup, which not only ensures the accuracy of high priority matching, but also covers the scenarios of non-standard name expression.
[0061] Meanwhile, this solution utilizes the technical feature of stopping subsequent matching once the confidence level of the current matching method reaches a corresponding threshold, avoiding unnecessary similarity calculations and improving the processing efficiency of port name standardization while ensuring matching accuracy. Through a unified coding mapping of port names, port data from different sources and with different descriptions can be aligned to the same standard system, ensuring consistent statistical standards for subsequent port berthing duration, berthing frequency, and replenishment / unloading ratios. This avoids statistical biases caused by different names for the same port, improving the accuracy and comparability of berthing behavior characteristics.
[0062] In one embodiment, port code matching uses UN / LOCODE international port code regular expression matching with a confidence threshold of 100%; port name prefix matching uses a pre-built standard port name dictionary with a confidence threshold of 90%; character sequence similarity matching uses the longest common subsequence algorithm to calculate similarity with a confidence threshold of 75%; port names that fail to reach the corresponding thresholds in the first three automatic matching layers are automatically entered into a manual correction queue, where business personnel complete the matching and update the standard port dictionary.
[0063] In some embodiments of this application, identifying the berthing and operational behaviors of target vessels and segmenting the vessel's spatiotemporal behavior data into voyage trajectories specifically includes: identifying berthing events based on the geofence of a preset port, combined with the speed and duration of the behavioral points, and determining corresponding arrival and departure events; segmenting continuous vessel spatiotemporal behavior data into multiple single-voyage trajectory segments using adjacent departure and arrival events as boundaries; berthing behavior characteristics include at least two of the following: total cumulative berthing time, berthing frequency, average berthing time per session, continuous berthing time, number of berthings at fixed berths, and the proportion of replenishment and unloading behaviors; operational distribution characteristics include at least two of the following: cumulative grid operational time, operational point density, number of consecutive operational days, seasonal operational distribution, and the proportion of cross-sea area navigation; and dividing and statistically analyzing the operational sea area using equal-area grid cells in a projected coordinate system.
[0064] In the above embodiments, this application refines the specific implementation methods of berthing behavior recognition, voyage segmentation, and feature extraction. The corresponding core technical features are as follows: based on the geofence of a preset port, berthing events are identified by combining the speed and duration of the behavior points, and corresponding arrival and departure events are determined; the continuous spatiotemporal behavior data of ships is segmented into multiple single voyage trajectory segments using adjacent departure and arrival events as boundaries; the berthing behavior features include at least two of the following: total cumulative berthing time, berthing frequency, average berthing time per berth, continuous berthing time, number of berthings at fixed berths, and the proportion of replenishment and unloading behavior; the operation distribution features include at least two of the following: cumulative operation time of the grid, operation point density, number of consecutive operation days, seasonal operation distribution, and the proportion of cross-sea area navigation; the operation sea area is divided and statistically analyzed using equal-area grid units under the projected coordinate system. This solution utilizes the technical features of port geofencing, combined with speed and duration, to identify berthing events. By determining berthing events based on three conditions—spatial range, motion state, and duration—it can accurately distinguish between actual vessel entry into port and berthing, as well as passing through waters near the port. It can also cover non-terminal berthing behaviors such as anchoring and low-speed stops, significantly improving the accuracy of berthing event identification and reducing missed and false diagnoses.
[0065] This solution utilizes multi-dimensional characteristics of port call behavior and multi-dimensional operational distribution to quantify the strength of the connection between vessels and regions from multiple perspectives, including duration, frequency, continuity, location stability, and business attributes. Among these, port call behavior comprehensively depicts the degree of bond between vessels and ports from the dimensions of port call time, frequency, space, and business. Operational distribution characteristics comprehensively depict the degree of connection between vessels and sea areas from the dimensions of operational time density, spatial density, continuity, seasonal patterns, and activity range. These multi-dimensional characteristics together constitute the quantitative basis for determining territorial jurisdiction. Compared with single-characteristic judgment, this method can more comprehensively and accurately reflect the territorial jurisdiction of vessels and significantly reduce judgment bias.
[0066] This scheme utilizes the technical feature of dividing and statistically analyzing the operational sea area using equal-area grid cells under a projected coordinate system. This eliminates the problem of inconsistent grid areas at different latitudes, enabling the statistical results of operational duration and point density in sea areas at different latitudes to have a unified measurement scale. It avoids the problem of inflated density statistics caused by smaller grid areas in high-latitude regions, and improves the statistical accuracy and cross-regional comparability of operational distribution characteristics.
[0067] In one embodiment, the port geofence is delineated using polygon coordinates based on the actual water area of the fishing port, the berthing speed threshold is set to 2 knots, and the berthing time threshold is set to 30 minutes; the operating sea area is divided into equal-area grids with an approximate scale of 500 meters, each grid corresponding to a unique grid code, and the operating points are mapped to the corresponding grids according to latitude and longitude for duration and density statistics.
[0068] In some embodiments of this application, a spatiotemporal joint distance is constructed based on the geographic spatial distance and the temporal continuous distance between behavioral points. The behavioral points are then clustered based on this spatiotemporal joint distance. Specifically, this includes: selecting any two behavioral points within the same time window, calculating the geographic spatial distance and the temporal continuous distance between the two behavioral points respectively, and standardizing both the geographic spatial distance and the temporal continuous distance; weighting and summing the squares of the standardized geographic spatial distance and the standardized temporal continuous distance according to spatial and temporal weight coefficients, and taking the square root to obtain the spatiotemporal joint distance; combining spatial neighborhood thresholds, temporal continuous thresholds, and the minimum number of behavioral points, density clustering is performed on the behavioral points within the same time window, and the set of behavioral points that meets the conditions is determined as a stable behavioral cluster; the geographic spatial distance is calculated using Haversine great circle distance, or the planar distance is calculated after conversion to a projected coordinate system.
[0069] In the above embodiments, the construction method of spatiotemporal joint distance and the execution logic of density clustering are refined. The corresponding core technical features are as follows: select any two behavior points within the same time window, calculate the geospatial distance and temporal continuous distance between the two behavior points respectively, and standardize the geospatial distance and temporal continuous distance respectively; according to the spatial weight coefficient and the temporal weight coefficient, the square value of the standardized geospatial distance and the square value of the standardized temporal continuous distance are weighted and summed, and the square root is taken to obtain the spatiotemporal joint distance; combine the spatial neighborhood threshold, the temporal continuous threshold and the minimum number of behavior points, perform density clustering on the behavior points within the same time window, and determine the set of behavior points that meet the conditions as a stable behavior cluster; the geospatial distance is calculated using Haversine great circle distance, or the planar distance is calculated after conversion to the projected coordinate system. This scheme utilizes the technical feature of standardizing both geospatial distance and temporal continuous distance before weighting them to construct a spatiotemporal joint distance. First, standardization eliminates the dimensional differences between the two types of distances, preventing one dimension from being over- or under-proportioned in the joint distance due to different units. Then, spatial and temporal weight coefficients are used to flexibly adjust the importance of the two dimensions to adapt to the clustering needs of different sea areas and ship types. Finally, the spatiotemporal joint distance is formed by calculating the square and square root, achieving a fusion measurement of spatial and temporal dimensions, enabling the clustering process to simultaneously consider the spatial proximity and temporal continuity of points.
[0070] This scheme utilizes three parameters—spatial neighborhood threshold, temporal continuity threshold, and minimum number of behavior points—to jointly constrain the technical characteristics of density clustering. It sets clustering thresholds from three dimensions: spatial range, temporal span, and point size. Only point sets that simultaneously meet all three conditions will be judged as stable behavior clusters. Discrete points, short-term stopping points, and trajectory points that do not meet the minimum number of points will be marked as noise or temporary activity areas and will not be directly used as primary location candidates, further ensuring the stability and reliability of candidate location areas.
[0071] This solution utilizes the technical characteristics of geospatial distance calculation using Haversine great circle distance or projected coordinate system planar distance calculation, balancing the calculation accuracy for large-scale sea areas with the calculation efficiency for small-scale sea areas. Among them, Haversine great circle distance is based on the Earth's spherical model to calculate the shortest spherical distance between two points. Compared with directly calculating the Euclidean distance of latitude and longitude, it can eliminate the distance error caused by the curvature of the Earth. It is especially suitable for trajectory clustering scenarios in large-scale nearshore sea areas, ensuring the accuracy of distance calculation and thus improving the reliability of clustering results.
[0072] In one embodiment, the spatial neighborhood threshold is set to 5 nautical miles, the time continuity threshold is set to 72 hours, and the minimum number of behavior points is set to 20. The spatial weight coefficient and the time weight coefficient are initially set to 0.6 and 0.4, respectively, and can be adjusted according to the ship activity patterns in different sea areas. For large sea areas such as Sea A, Sea B, and Sea C, the Haversine formula is uniformly used to calculate the great circle distance.
[0073] In some embodiments of this application, the comprehensive score of each candidate territory region is calculated, specifically including: for each candidate territory region, normalizing and weighting the docking behavior characteristics and operation distribution characteristics to obtain the docking behavior score and operation sea area association score corresponding to the candidate territory region; determining the continuity bonus based on the continuous duration of stable behavior clusters within the candidate territory region, and determining the short-term transit penalty item based on the docking duration, docking frequency, and operation cluster formation within the candidate territory region; after weighted summing of the docking behavior score, operation sea area association score, and continuity bonus, subtracting the weighted short-term transit penalty item to obtain the initial territory score corresponding to the candidate territory region; obtaining the time difference between the current time window and the most recent valid behavior record within the candidate territory region, generating an exponential time decay factor based on the time difference and a preset decay coefficient, multiplying the initial territory score by the time decay factor to obtain the comprehensive territory score after historical decay correction; when the time difference exceeds a preset failure threshold, increasing the decay coefficient, or marking the candidate territory region as a failed candidate region.
[0074] In the above embodiments, the calculation logic and historical decay mechanism of the comprehensive score of the territory are refined. The core technical features are as follows: For each candidate territory, the docking behavior characteristics and operation distribution characteristics are normalized and weighted to obtain the docking behavior score and operation sea area association score corresponding to the candidate territory; the continuity score is determined by combining the continuous duration of stable behavior clusters in the candidate territory, and the short-term transit penalty item is determined by combining the docking duration, docking frequency and operation cluster formation in the candidate territory; after weighted summation of docking behavior score, operation sea area association score and continuity score, the weighted short-term transit penalty item is subtracted to obtain the initial territory score corresponding to the candidate territory; the time difference between the current time window and the most recent valid behavior record in the candidate territory is obtained, and an exponential time decay factor is generated according to the time difference and the preset decay coefficient. The initial territory score is multiplied by the time decay factor to obtain the comprehensive territory score after historical decay correction; when the time difference exceeds the preset failure threshold, the decay coefficient is increased, or the candidate territory is marked as a failed candidate area. This scheme utilizes a four-layer scoring structure—"affiliation behavior score + operational sea area association score + continuity bonus - short-term transit penalty"—to construct a comprehensive and highly discriminative quantitative scoring model. Affiliation behavior score and operational sea area association score serve as base scores, corresponding to the association strength of the two core dimensions of port and sea area, respectively. Continuity bonus serves as a positive incentive; the longer the behavior lasts, the more stable the bond between the vessel and the area, resulting in a higher bonus and reinforcing the scoring advantage of areas with long-term stable activity. The short-term transit penalty serves as a negative constraint, imposing corresponding penalty points on areas with short berthing times, low frequency, and no stable operational clusters, lowering the score ranking of short-term transit areas and preventing occasional transit areas from being misclassified as territorial. This combination of bonus and penalty design further widens the score gap between the primary territorial jurisdiction and the transit area, improving the discriminativeness and accuracy of territorial determination.
[0075] This solution utilizes an exponential time decay factor to correct the initial territorial score, enabling dynamic updates to the territorial score. Behavioral records closer to the current time contribute more to the territorial score, while older historical data contribute less, ensuring that the territorial score reflects the latest ship activity patterns in real time and reducing interference from long-term data on the current territorial determination. When the time difference exceeds a preset failure threshold, the system enters a rapid decay range by increasing the decay coefficient or directly marking failure candidate areas, further strengthening the impact of data timeliness on the determination results and adapting to scenarios of dynamically changing ship operating areas.
[0076] The above-mentioned multi-dimensional weighted scoring combined with the historical decay mechanism enables the comprehensive score of the territory to fully and dynamically reflect the true degree of correlation between the vessel and the candidate area. The judgment process is interpretable and traceable, while effectively reducing the risk of misjudgment caused by occasional behavior and outdated data.
[0077] In one embodiment, the initial weight of the affiliation behavior score is set to 0.6, and the initial weight of the operation sea area association score is set to 0.4; the attenuation coefficient is initially set to 0.1 / month, that is, the score decreases by about 10% for every month of delay; when the most recent valid record is more than 6 months away from the current time, the attenuation coefficient is increased to 0.3 / month, and when there is no valid record for more than 12 months, the corresponding candidate area is marked as a failed candidate area.
[0078] In some embodiments of this application, the sufficiency of evidence for candidate primary home locations is verified by combining the amount of effective behavioral data and data timeliness within the current time window. Specifically, this includes: sorting each candidate home location region from high to low according to the comprehensive home location score, and determining the candidate home location region corresponding to the highest comprehensive home location score as the candidate primary home location; classifying the candidate primary home location based on the home location score level, score leadership, amount of effective behavioral data, and data timeliness as verification dimensions: when the highest comprehensive home location score reaches the first score threshold, the difference between the highest comprehensive home location score and the second highest comprehensive home location score reaches the leading difference threshold, and the number of effective positioning points, effective flight times, effective docking times, and number of effective operation points within the current time window are verified. When the quantity of evidence reaches the corresponding minimum evidence threshold and the data time interval is not greater than the timeliness threshold, the sufficiency of the primary territory evidence is deemed to have passed the verification, and the candidate primary territory is determined as the actual activity territory, with the judgment status set to confirmed. When the highest territory comprehensive score reaches the second score threshold which is lower than the first score threshold, or when some conditions of effective behavioral data quantity or data timeliness are not met, the judgment status is set to low confidence. When the highest territory comprehensive score is lower than the second score threshold, or when any evidence indicator fails to reach the threshold or the data timeliness is exceeded, the judgment status is set to pending confirmation or insufficient data. If the business system must display the reference territory, the registered territory in the ship static file data is used as the reference territory output, and the source of the reference territory is marked as registration information.
[0079] In the above embodiments, the hierarchical judgment rules and fallback output mechanism for verifying the sufficiency of primary territory evidence are refined. The core technical features are as follows: Candidate territories are sorted from highest to lowest based on their comprehensive territories score, and the candidate territories with the highest comprehensive territories score are determined as candidate primary territories; candidate primary territories are classified and judged based on territories score level, score leadership, effective behavioral data volume, and data timeliness as verification dimensions: when the highest comprehensive territories score reaches the first score threshold, the difference between the highest and second-highest comprehensive territories score reaches the leading difference threshold, and the number of effective positioning points, effective flight times, effective docking times, and effective operation points within the current time window all reach the corresponding maximum... When the threshold for minor evidence and the data timeliness interval are not greater than the timeliness threshold, the sufficiency verification of the primary territory evidence is deemed to have passed, the candidate primary territory is determined as the actual activity territory, and the judgment status is set to confirmed. When the highest territory comprehensive score reaches the second score threshold which is lower than the first score threshold, or when some conditions of effective behavioral data volume or data timeliness are not met, the judgment status is set to low confidence. When the highest territory comprehensive score is lower than the second score threshold, or when any evidence indicator fails to reach the threshold or the data timeliness exceeds the time limit, the judgment status is set to pending confirmation or insufficient data. If the business system must display the reference territory, the registered territory in the ship static file data is used as the reference territory output, and the source of the reference territory is marked as registration information.
[0080] The implementation process of the beneficial technical effects, based on the aforementioned technical features, is as follows: This solution utilizes a hierarchical judgment mechanism with four verification dimensions and three levels of judgment status to achieve a strict correspondence between the judgment result and the strength of evidence. The four verification dimensions comprehensively measure the sufficiency of the judgment basis from four aspects: absolute score, relative score advantage, data sample size, and data freshness. Compared to single-dimensional verification, this provides a more comprehensive assessment of the reliability of the conclusion. The three levels of judgment status correspond to different levels of evidence strength: the "confirmed" status corresponds to scenarios with sufficient evidence and reliable conclusions, which can be directly used for law enforcement jurisdiction determination; the "low confidence" status corresponds to scenarios where the score meets the standard but the evidence is flawed, prompting reviewers to pay close attention; and the "pending confirmation / insufficient data" status corresponds to scenarios with missing evidence, and no definitive conclusion is output. This hierarchical judgment mechanism avoids the limitations of binary judgments of "either / or," making the confidence level of the territorial result quantifiable and distinguishable, adapting to the usage needs of different business scenarios.
[0081] This solution utilizes the technical features of using the registered location as a fallback reference location and clearly marking the source. When there is insufficient evidence to output the model's judgment result, it also takes into account the display needs of the business system. At the same time, by clearly marking the source of the reference location as the registration information, it avoids mistaking the registration information for the trajectory model's judgment result, ensuring the traceability and rigor of the location result and preventing confusion in business use.
[0082] The aforementioned evidence sufficiency verification and grading mechanism ensures the rigor of the territorial determination results from a rule-based perspective, and significantly reduces the risk of misjudgment caused by data sparsity and outdated data.
[0083] In one embodiment, the first scoring threshold is set to 0.8, the second scoring threshold is set to 0.5, and the leading difference threshold is set to 0.2; the minimum evidence thresholds are: no less than 100 valid positioning points, no less than 5 valid flight times, no less than 3 valid docking times, and no less than 20 valid operation points; the data timeliness threshold is set to 30 days, that is, the most recent valid record is no more than 30 days from the current time.
[0084] In some embodiments of this application, based on the candidate primary affiliation, comprehensive affiliation score, and primary affiliation evidence sufficiency verification results corresponding to multiple consecutive time windows, the time windows are merged or segmented to generate dynamic affiliation identification results with effective time intervals and confidence levels. Specifically, this includes: comparing the candidate primary affiliation and comprehensive affiliation scores corresponding to two adjacent time windows; merging the two adjacent time windows when the candidate primary affiliation is the same and the change in the corresponding comprehensive affiliation score does not reach the segmentation change threshold; setting a affiliation time boundary point between the two adjacent time windows when the candidate primary affiliation is different and at least one of the difference in the corresponding comprehensive affiliation score, the change in the candidate affiliation regional ranking, or the change in confidence level reaches the segmentation change threshold; and generating an effective time interval based on the affiliation time boundary point. The start time and end time of failure are defined. When the same candidate territory repeatedly forms a stable behavior cluster within multiple time windows corresponding to the same fishing season, but does not meet the main territory determination conditions within a continuous year-round time window, the candidate territory is marked as a seasonal temporary territory. After generating dynamic territory identification results, the actual activity territory determined by the model is extracted from the dynamic territory identification results. The actual activity territory, the identifier resolution territory, and the registered territory are compared pairwise. When the geographical distance, administrative division level difference, inconsistency duration, and corresponding confidence level between any two types of territories meet the territory conflict conditions, a territory conflict mark is generated, and the conflict type, conflict level, involved territories, and evidence source are output. Conflict types include conflict between identifier resolution territory and registered territory, conflict between identifier resolution territory and actual activity territory, and conflict between registered territory and actual activity territory.
[0085] In the above embodiments, the execution logic of dynamic time sharding and the three-party location conflict verification mechanism are refined. The core technical features are as follows: comparing the candidate primary location and the comprehensive location score corresponding to two adjacent time windows; merging the two adjacent time windows when the candidate primary location corresponding to two adjacent time windows is the same and the change in the corresponding comprehensive location score does not reach the sharding change threshold; setting a location time boundary point between the two adjacent time windows when the candidate primary location corresponding to two adjacent time windows is different and at least one of the following—the difference in the corresponding comprehensive location score, the change in the candidate location region ranking, or the change in confidence—reaches the sharding change threshold; generating an effective start time and an invalid end time based on the location time boundary point; and when the same candidate location region is in multiple corresponding... If a stable cluster of behaviors repeatedly forms within the same fishing season window, but fails to meet the criteria for determining the primary place of residence within a continuous year-round window, the candidate place of residence is marked as a seasonal temporary place of residence. After generating dynamic place of residence identification results, the actual activity place of residence determined by the model is extracted from the dynamic place of residence identification results. The actual activity place of residence, the identifier resolution place of residence, and the registered place of residence are compared pairwise. When the geographical distance, administrative division level difference, inconsistency duration, and corresponding confidence level between any two types of place of residence meet the place of residence conflict conditions, a place of residence conflict marker is generated, and the conflict type, conflict level, involved place of residence, and source of evidence are output. The conflict types include conflict between identifier resolution place of residence and registered place of residence, conflict between identifier resolution place of residence and actual activity place of residence, and conflict between registered place of residence and actual activity place of residence.
[0086] The beneficial technical effects achieved by this solution are as follows: This solution utilizes the technical features of comparing, merging, and segmenting adjacent time windows to automatically generate dynamic location time intervals; continuous windows with consistent location status are merged into complete effective time periods, reducing fragmented time segments and improving the readability and business applicability of the results; demarcation points are set at locations where substantial changes occur in location to ensure consistency of location status within each time segment. Finally, location records with precise start and end times are generated, achieving a refined expression of the time dimension of location results.
[0087] This solution utilizes the technical features of seasonal temporary territorial marking to identify areas that operate during fixed fishing seasons each year but do not meet the main territorial criteria throughout the year, based on the seasonal patterns of fishery production. These areas are then separately marked as seasonal temporary territorial territories, further refining the granularity of territorial types and enabling better adaptation to business scenarios such as fishery administration's closed season management and cross-regional operation reporting.
[0088] This solution utilizes the technical characteristics of cross-verification among the actual activity location, the identifier resolution location, and the registration location to construct a multi-dimensional abnormal vessel identification mechanism. Through comprehensive judgment based on four dimensions—geographical distance, differences in administrative division level, duration of inconsistency, and corresponding confidence level—false alarms caused by single temporary dockings or minor naming differences are avoided. Conflict markers are only triggered when the location difference reaches a comprehensive threshold. Three conflict types correspond to different abnormal scenarios: a conflict between the identifier resolution location and the registration location may correspond to incorrect vessel registration information or non-standard vessel naming; a conflict between the identifier resolution location and the actual activity location may correspond to AIS vessel name tampering or vessel counterfeiting; and a conflict between the registration location and the actual activity location may correspond to vessels operating in different locations for extended periods or calling at ports outside their registered area. Different types of conflict markers provide fisheries law enforcement personnel with clear verification directions, significantly improving the efficiency and accuracy of abnormal vessel investigations.
[0089] In one embodiment, the segmentation change threshold is set as follows: a difference of 0.3 in the comprehensive score, a change of more than 2 places in the candidate region ranking, or a decrease of 1 level or more in the confidence level. Segmentation is performed when any one of these conditions is met. The territorial conflict conditions are set as follows: the two types of territories belong to different prefecture-level cities, the inconsistency lasts for more than 3 months, and the corresponding territorial determination confidence level reaches low confidence or above. A conflict marker is generated when all of the above conditions are met. The conflict level is divided into three levels: general, moderate, and severe, based on the administrative division difference level and duration.
[0090] This embodiment provides a method for dynamic identification of fishing vessel location by integrating multi-source spatiotemporal data. It addresses the regulatory needs of fishery administration departments at all levels along the coast for determining fishing vessel location, controlling cross-regional operations, and tracing illegal vessels. The method is implemented using six sequential processing modules, namely: multi-source data access and cleaning module, fishing vessel trajectory spatiotemporal segmentation and behavior extraction module, port berthing and operation area multi-feature quantification calculation module, spatiotemporal clustering candidate region identification module, location scoring and conflict verification and dynamic segmentation module, and result storage and business output module.
[0091] I. Multi-source heterogeneous data access and standardized preprocessing
[0092] 1. Multi-source data source access
[0093] This embodiment connects to four types of data sources:
[0094] (1) Static archive of fishing vessels: stores basic registration information of vessels, including fields such as vessel MMSI (Maritime Mobile Service Identifier), vessel name, port of registry, fishing license operating area, vessel tonnage, registration time, etc.
[0095] (2) AIS dynamic spatiotemporal data: that is, the real-time trajectory data reported by the Automatic Identification System of ships, which consists of multiple behavioral points arranged in chronological order. Each behavioral point includes fields such as latitude and longitude, timestamp, speed, heading, MMSI, equipment status, and AIS ship name.
[0096] (3) Port entry and exit business data: docking records from port scheduling, fishery administration and other systems, including fields such as port name, entry time, departure time, docking duration, supply identifier, unloading identifier, etc.
[0097] (4) Fishery operation location data: Fishing operation location data collected from channels such as manual patrol, radar monitoring, and satellite remote sensing.
[0098] 2. Data cleaning and trajectory correction
[0099] First, the AIS dynamic spatiotemporal data is filtered for anomalies, removing five types of invalid anomaly points: coordinate boundary points (points whose latitude and longitude exceed the reasonable range of the target sea area), speed anomaly points (points whose speed exceeds the normal navigation speed range of fishing vessels), timestamp reversal points (out-of-order messages whose reporting time is later than subsequent points), duplicate message points (redundant points repeatedly reported by the same vessel at the same time), and position change drift points (points whose position changes drastically after a long period of positioning signal interruption).
[0100] After removing outlier points, interactive multi-model filtering (IMM) is used to smooth the remaining discrete trajectory points. The filter model set includes three types: uniform motion model, turning motion model, and low-speed mooring model. The optimal fitting model can be automatically switched according to the real-time motion status of the ship. The process noise and observation noise parameters of the model can be set according to the historical AIS error statistics of the same sea area or the speed change range of the corresponding ship type to ensure the accuracy of trajectory smoothing.
[0101] 3. Standardization of port names
[0102] To address the issue of inconsistent port name representations from different sources, a four-layer progressive matching framework is adopted to uniformly map them to standard port codes. The matching priority from high to low is as follows: the first layer is UN / LOCODE (United Nations Port Code) regular expression matching, which performs accurate matching based on the official port code library; the second layer is port dictionary prefix matching, which performs fuzzy prefix matching based on a pre-built standard port name dictionary; the third layer is longest common subsequence similarity matching, which uses string similarity calculation to adapt to scenarios such as name misspellings and abbreviations; and the fourth layer is manual correction, which performs manual calibration for special port names that cannot be covered by automatic matching.
[0103] During the matching process, when the previous matching is successful and the confidence level reaches the threshold of the corresponding level, the subsequent matching steps are stopped, balancing matching accuracy and processing efficiency.
[0104] 4. Analysis of AIS ship name country prefixes
[0105] The specific process for extracting location clues from the AIS ship name field is as follows: First, the AIS ship name field is cleaned and standardized by lowercase to remove special symbols, spaces and garbled characters; then, the first 2-3 characters of the ship name are identified and extracted as the pinyin prefix of the province / region abbreviation, which is the pinyin prefix of the ship name.
[0106] The system maintains a mapping dictionary of "pinyin prefix - province / region - administrative division code", maps the extracted ship name pinyin prefix to the corresponding identifier resolution location, and records the confidence level of each resolution result; for prefixes that are ambiguous and can correspond to multiple administrative divisions, all candidate identifier resolution locations are retained, or they are marked as pending manual confirmation.
[0107] For example, the ship name field “suqiyu12345” can be parsed to have the prefix “su”, which corresponds to the clue of Jiangsu Province (A); the ship name field “yueganyu12345” can be parsed to have the prefix “yue”, which corresponds to the clue of Dong Province (A).
[0108] 5. Spatiotemporal unified rasterization
[0109] The target sea area is divided into fishing ground grid units of uniform scale for statistical analysis of the characteristics of the operating sea area. In this embodiment, it is preferred to use an equal-area grid under the projected coordinate system with a grid scale of approximately 500 meters to eliminate statistical bias caused by inconsistencies in the grid area at different latitudes. If a latitude and longitude grid is used directly, the applicable latitude range is limited to control statistical errors.
[0110] II. Fishing Vessel Trajectory Segmentation and Key Behavioral Feature Extraction
[0111] The core of this "behavior extraction" step is to identify ship behavior events based on trajectory data, segment single voyage units, and extract two core features: port berthing and operational sea area. For details, please refer to [link / reference needed]. Figure 2 A schematic diagram of fishing vessel voyage trajectory segmentation and behavior recognition.
[0112] like Figure 2 As shown, the single voyage trajectory is divided by the boundary of "departure event - arrival event": the left side is port A, where a departure event is triggered when the vessel leaves port A, serving as the starting point of the voyage; the right side is port B, where an arrival event is triggered when the vessel enters port B, serving as the ending point of the voyage; the complete trajectory between the two points constitutes a single voyage trajectory segment. Within the voyage segment, candidate operating areas corresponding to low-speed / berthing points can be identified, and three core indicators for the voyage—voyage duration, operating duration, and berthing duration—can be statistically analyzed.
[0113] 1. Flight trajectory segmentation and event recognition
[0114] Based on port geofencing combined with speed and duration conditions, berthing, arrival, and departure events are identified: a corresponding port geofencing (i.e., the spatial boundary of the port waters) is pre-defined for each port; when multiple consecutive action points are located within the port geofencing, and the speed of the points is lower than the berthing speed threshold, and the duration reaches the berthing time threshold, it is determined as a berthing event; the time when the vessel enters the port geofencing is recorded as the arrival event time, and the time when the vessel leaves the port geofencing is recorded as the departure event time.
[0115] Using adjacent departure and arrival events as boundaries, continuous ship spatiotemporal behavior data is divided into multiple independent single-voyage trajectory segments. Each voyage segment is marked with origin and destination ports, sailing duration, berthing duration, and operation duration attributes.
[0116] 2. Port berthing feature extraction
[0117] For each candidate port, the characteristics of vessel berthing behavior are statistically analyzed, including: total cumulative berthing time, monthly berthing frequency, average berthing time per trip, number of consecutive months of berthing, number of berthings at fixed berths, and the proportion of replenishment and unloading activities. These characteristics comprehensively depict the bond strength between vessels and ports from the dimensions of time, frequency, space, and business.
[0118] 3. Feature extraction of the operating sea area
[0119] Based on the statistical characteristics of the sea area operations using equal-area grid cells, the points corresponding to the operational activities are mapped to their respective grids, and relevant indicators for each grid are statistically analyzed, specifically including: cumulative operation time per grid, density of fishing points, number of consecutive days of operation, seasonal distribution of operations, and proportion of cross-sea area navigation. These characteristics comprehensively reflect the activity level and patterns of vessels in different sea areas.
[0120] 4. Time window slicing
[0121] Time windows are divided according to natural months, fishing seasons, or regulatory cycles, and a sliding window mechanism is used for statistics. The window parameters include the window length L and the sliding step δ. For example, L can be set to 1 month and δ to 1 month, or L can be set to 1 fishing season and δ to 1 month. The window parameters can be flexibly adjusted according to the regulatory granularity requirements. All feature statistics and subsequent clustering, scoring, and judgment are performed on a per-time-window basis to ensure the time granularity of territorial determination.
[0122] III. Quantitative Calculation of Multiple Features of Port Call-in and Operational Sea Areas
[0123] This step corresponds to the "multi-feature quantization calculation" stage, and also corresponds to... Figure 4The core of the scoring section on the left is to construct a multi-dimensional weighted scoring model to calculate the comprehensive score of each candidate territory and quantify the correlation between ships and regions.
[0124] 1. Basic Formula for Overall Score
[0125] For each fishing vessel v, each candidate port / candidate region r, and each time window t, calculate the comprehensive territorial score using the following formula:
[0126] S(v,r,t)=w1×Sport(v,r,t)+w2×Sfishing(v,r,t)+w3×Scontinuity(v,r,t)-w4×Stransit(v,r,t);
[0127] Among them, Sport is the port berthing behavior score, which is obtained by weighting the cumulative berthing time, monthly berthing frequency, number of berthings at fixed berths, and the proportion of replenishment and unloading after normalization.
[0128] Sfishing is the operational sea area correlation score, which is obtained by weighting the grid operation duration, fishing point density, and number of consecutive operation days after normalization.
[0129] Scontinuity is a continuous score, determined based on the continuous duration of stable behavior clusters within the candidate region; the longer the behavior lasts, the higher the score.
[0130] Stransit is a short-term transit penalty item, determined based on the stop duration, stop frequency, and operation cluster formation within the candidate area. Areas with occasional stops or without stable operation clusters will have corresponding points deducted.
[0131] The weight parameters can be adjusted according to the labeled samples or expert rules. Initially, w1=0.6 and w2=0.4 can be set. After introducing the continuous bonus and transit penalty, all weights are normalized to satisfy w1+w2+w3+w4=1.
[0132] 2. Historical attenuation correction
[0133] To reduce the impact of outdated historical data on current judgments, an exponential time decay mechanism is introduced to decay and correct the initial score. The corrected score formula is as follows:
[0134] Sadj(v,r,t)=S(v,r,t)·e^(-λ×Δt)
[0135] Where Δt is the month difference between the current time window and the most recent valid docking or valid operation record within the candidate area, and λ is the preset attenuation coefficient. Sadj is the scoring formula.
[0136] When the most recent valid record is more than 6 months old, the region enters the rapid decay interval, the decay coefficient λ is increased, and the weight of old data is further reduced; if the maximum valid time limit is exceeded, the candidate region can be directly marked as an invalid candidate region and will no longer be included in the main location determination range.
[0137] IV. Spatiotemporal Clustering Candidate Region Identification
[0138] For specific principles and processes, please refer to... Figure 3 A schematic diagram of spatiotemporal joint clustering and candidate location identification.
[0139] like Figure 3 As shown, the input to the clustering algorithm is the spatial dimension of port locations, trajectory locations, and work grid locations, and the spatiotemporal joint distance is used as the similarity metric between locations.
[0140] The final clustering outputs two types of effective stable behavior clusters: cluster C1 corresponds to candidate territories of ports with long-term berthing and cluster C2 corresponds to candidate territories of sea areas with seasonal operations; discrete points that do not meet the clustering conditions are marked as noise points, corresponding to short-term transit areas, and are not included in the candidate territories area, nor are they used as the basis for determining the primary territories.
[0141] V. Local jurisdiction scoring, conflict verification of registered clues, sufficiency of evidence verification, and dynamic zoning.
[0142] 1. Ranking of candidate primary geographic locations
[0143] For all candidate geographic regions within a single time window, they are sorted from highest to lowest according to their comprehensive geographic region score (Sadj) after historical attenuation correction, and the candidate geographic region with the highest score is determined as the candidate primary geographic region.
[0144] 2. Verification of the sufficiency of evidence regarding the principal jurisdiction
[0145] like Figure 4 As shown in the middle section, before outputting the final judgment result, the sufficiency of evidence for the candidate primary location is first verified. The verification dimensions include four types of data scale indicators and one type of timeliness indicator.
[0146] Based on the verification results and score levels, the three-tiered classification judgment rule is applied:
[0147] (1) Confirmed status: When the highest comprehensive score reaches the first score threshold (high threshold), the difference between the highest score and the second highest score reaches the leading difference threshold, and all the above evidence indicators meet the threshold requirements, the evidence of the main territory is deemed sufficient, the candidate main territory is determined as the actual activity territory, and the status is marked as "confirmed".
[0148] (2) Low confidence status: When the highest comprehensive score reaches the second score threshold (medium threshold) which is lower than the first score threshold, or when some evidence indicators do not meet the requirements, the status is marked as "low confidence", indicating that there is uncertainty in the results of the territory;
[0149] (3) Pending confirmation / Insufficient data status: When the highest comprehensive score is lower than the second score threshold, or any evidence indicator fails to reach the threshold, or the data timeliness expires, the status is marked as "pending confirmation" or "insufficient data".
[0150] 3. Cusp-of-Potential Output Rules
[0151] When there is insufficient evidence or the data has not been updated for a long time, the system will not force the highest-scoring area to be determined as the primary location. If the business system must display the reference location, the registered port of registry in the ship's static file will be output as the reference location. At the same time, the source of the reference location will be clearly marked as "registration information" to distinguish it from the actual activity location determined by the trajectory model, so as to avoid confusion in the use by the business side.
[0152] 4. Verification of territorial conflicts among the three parties
[0153] The system performs pairwise cross-comparisons of three types of location clues: the location of the identifier obtained from AIS vessel name resolution, the registration location in the fishing vessel's static file, and the actual activity location determined by the model.
[0154] The conflict determination comprehensively considers four dimensions: geographical distance between the two jurisdictions, differences in administrative division levels, duration of inconsistency, and confidence level of the corresponding jurisdiction results. When all four dimensions meet the jurisdiction conflict conditions, a jurisdiction conflict marker is generated, and the conflict type, conflict level, jurisdiction involved, source of evidence, and reason for recommendation for review are output to avoid false alarms due to a single temporary stop or error in a single field.
[0155] There are three types of conflicts:
[0156] (1) Conflict between the identification resolution location and the registration location: This indicates that there may be an anomaly in the ship registration information or that the AIS ship name is not filled in correctly;
[0157] (2) Conflict between the AIS identifier resolution location and the actual activity location: This indicates that there may be abnormal situations such as AIS ship name identifier tampering or ship license plate cloning;
[0158] (3) Conflict between the registered location and the actual location of the activity: This indicates that the vessel has been operating in different locations for a long time and calling at other ports.
[0159] Conflict markers can provide fishery law enforcement personnel with directions for identifying abnormal vessels, supporting manual review and cross-regional collaborative law enforcement.
[0160] 5. Dynamic time-segmentation
[0161] Based on the judgment results of multiple consecutive time windows, time intervals are automatically merged or segmented to generate dynamic location records with a time dimension. The specific rules are as follows:
[0162] Compare the candidate primary location and the comprehensive score of two adjacent time windows. If the candidate primary location of adjacent windows is the same and the change in comprehensive score does not reach the segmentation change threshold, then the two time windows are merged into the same effective period of the location. If the candidate primary location of adjacent windows is different and at least one of the following reaches the segmentation change threshold: the difference in comprehensive score, the change in candidate region ranking, or the change in confidence, then a location time boundary point is set between the two windows.
[0163] Each generated dynamic location record includes the following fields: location, effective start time, expiration end time, and confidence level. The segmentation change threshold can be determined using a fixed threshold, historical distribution percentiles, or the optimal threshold from the validation samples.
[0164] In response to the seasonal patterns of fishery production, a special seasonal temporary territorial marker is set up: if the same candidate territorial area repeatedly forms a stable cluster of behaviors within the same fishing season window each year, but does not meet the criteria for determining the primary territorial location within a continuous time window throughout the year, it is marked as a seasonal temporary territorial location, which is suitable for business scenarios such as fishing moratorium management and cross-regional operation reporting.
[0165] Figure 4 The time sharding example below visually demonstrates the dynamic sharding results: January-March corresponds to Port A, and the status is confirmed; April-June has insufficient data, and the status is pending confirmation; July-September corresponds to Port B, and the status is confirmed; October-December uses the port of registration as a reference, and the status is low confidence.
[0166] VI. Result Storage and Business Output
[0167] MMSI, Vessel Name, AIS Vessel Name Resolution Location, Vessel Name Resolution Confidence, Registered Port of Registry, Model-Determined Port of Attribution, Primary Location Determination Status, Reference Location Source, Location Effective Period, Location Overall Score, Confidence Level, Registration Clue Conflict Marker, Conflict Type, Conflict Level, Anomaly Marker, Main Fishing Grounds, Historical Call-in Statistics, Last Valid Track Time, Number of Valid Track Points, Number of Valid Voyages, Number of Valid Call-ins, Reasons for Insufficient Evidence, Data Update Time.
[0168] The primary attribution status is categorized into four types: "confirmed," "pending confirmation," "low confidence," and "insufficient data." The reference attribution sources are categorized into four types: "model determination," "registered ship flag," "AIS ship name resolution," and "manual confirmation."
[0169] The output results can be directly connected to business systems such as the fishery administration supervision dashboard, the cross-regional law enforcement transfer system, and the fishing moratorium control and early warning system, providing data support for regulatory operations.
[0170] The effectiveness of this method can be verified by comparing it with the static ship registration matching method and the single spatial clustering method through indicators such as accuracy, recall, false positive rate, manual review pass rate, and processing time of a single ship's annual trajectory.
[0171] like Figure 5 As shown, an embodiment of this application provides an electronic device 200, including a memory 204 and a processor 202. The memory 204 stores a computer program, and the processor 202 executes the computer program to implement the ship dynamic territorial identification method as described in any of the above embodiments. Therefore, the electronic device 200 possesses all the beneficial effects of the ship dynamic territorial identification method as described in any of the above embodiments.
[0172] like Figure 6 As shown, Figure 6 The system demonstrates six sequential processing steps: multi-source data access, data cleaning and standardization, behavior extraction, candidate area identification, and territory determination output. It accesses AIS vessel name, file, port entry and exit, and operation point data, completes trajectory correction, port coding unification, and sea area rasterization, performs voyage segmentation, berthing and operation sea area feature extraction, and performs scoring and ranking, evidence verification, and time segmentation after spatiotemporal DBSCAN clustering. The output results dataset includes fields such as MMSI, vessel name resolution territory, registered port of registry, model-determined territory, conflict type, and reasons for insufficient evidence.
[0173] This application defines the voyage boundary using the departure event at port A and the arrival event at port B, identifies low-speed / berthing points along the way as candidate operation areas, and counts the sailing time, operation time, and berthing time of each voyage segment to complete the segmentation of a single voyage trajectory, providing basic units for the subsequent extraction of port berthing and operation area related features.
[0174] This application spatially aggregates port, trajectory, and operational grid points, employing a spatiotemporal joint distance formula including Haversine great circle distance for calculation. Through DBSCAN spatial neighborhood threshold εs, temporal continuity threshold εt, and minimum behavior point count MinPts constraints, it separates cluster C1 representing long-term berthing and cluster C2 representing seasonal operations, marking discrete points as noise points representing short-term transit, thus filtering out effective candidate jurisdictional areas. Based on two types of scoring inputs—berthing behavior and operational sea area—this application calculates a comprehensive score Sadj with exponential historical decay correction and ranks candidate ports. Then, it conducts evidence sufficiency verification, distinguishing between insufficient evidence requiring confirmation and sufficient evidence allowing for the output of the primary jurisdiction based on indicators such as trajectory points, voyages, number of stops, and update time. Combining high and medium-level thresholds, it classifies primary jurisdiction, temporary jurisdiction, and transit jurisdiction, automatically generating segmented time axes. Examples demonstrate the segmented jurisdictional effects for port A (confirmed), insufficient data requiring confirmation, port B (confirmed), and registered port with low confidence. Figure 7 and Figure 8 First, AIS anomaly filtering is completed, and the mapping regions of ship name prefixes such as "suqiyu" and "yueganyu" are parsed. After IMM trajectory smoothing and port name standardization, two types of registration evidence, namely the registration file location and the ship name resolution location, are summarized. The model is used to determine the location, the registration file location, and the ship name resolution location for three-way conflict comparison. Cross-regional operations are identified, abnormal situations are marked, and manual review prompts are output. This completes the ship name prefix resolution and three-way conflict verification processing chain that was not covered in the draft.
[0175] In the claims, description, and accompanying drawings of this invention, the term "plural" refers to two or more. Unless otherwise explicitly defined, the terms "upper," "lower," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and simplifying the descriptive process, and are not intended to indicate or imply that the device or element referred to must have the described specific orientation, or be constructed and operated in a specific orientation. Therefore, these descriptions should not be construed as limiting the invention. The terms "connected," "installed," "fixed," etc., should be interpreted broadly. For example, "connected" can be a fixed connection between multiple objects, a detachable connection between multiple objects, or an integral connection; it can be a direct connection between multiple objects or an indirect connection between multiple objects through an intermediate medium. For those skilled in the art, the specific meaning of the above terms in this invention can be understood based on the specific circumstances described above.
[0176] In the claims, description, and accompanying drawings of this invention, the terms "one embodiment," "some embodiments," "specific embodiment," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of the invention. In the claims, description, and accompanying drawings of this invention, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.
[0177] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.
Claims
1. A method for identifying the dynamic homeland of a vessel, characterized in that, include: Acquire multi-source data of the target vessel, wherein the multi-source data includes at least static archive data of the vessel and spatiotemporal behavior data of the vessel, wherein the spatiotemporal behavior data of the vessel includes multiple behavior points arranged in chronological order; The ship's spatiotemporal behavior data is subjected to anomaly processing, and the processed ship's spatiotemporal behavior data is divided into multiple time windows; Within each time window, the docking and operational behaviors of the target vessel are identified based on the behavioral points. The vessel's spatiotemporal behavior data is segmented into voyage trajectories, and the docking behavior features corresponding to the docking area and the operational distribution features corresponding to the operational sea area are extracted from the segmented voyage trajectory data. A spatiotemporal joint distance is constructed based on the geographic spatial distance and temporal continuous distance between the behavioral points. The behavioral points are clustered based on the spatiotemporal joint distance to obtain stable behavioral clusters. Candidate geographical regions are identified based on the stable behavioral clusters. Based on the affiliation behavior characteristics and the operation distribution characteristics, calculate the comprehensive score of each candidate territory region; Candidate primary locations are determined based on the comprehensive score of the location, and the sufficiency of primary location evidence is verified by combining the amount of effective behavioral data and the timeliness of the data within the current time window. Based on the candidate primary location, comprehensive location score, and primary location evidence sufficiency verification results corresponding to multiple consecutive time windows, the time windows are merged or segmented to generate dynamic location identification results with effective time intervals and confidence levels.
2. The method for identifying the dynamic homeland of a ship according to claim 1, characterized in that, The abnormal data processing of the ship's spatiotemporal behavior data specifically includes: Based on the geographical location, speed and timestamp of the behavior points, abnormal points in the ship's spatiotemporal behavior data are removed. The abnormal points include at least coordinate boundary crossing points, abnormal speed points, time reversal points, duplicate message points, and position change drift points after positioning interruption. After removing outlier points, interactive multi-model filtering is used to smooth the trajectory of the remaining behavioral points. The interactive multi-model filtering uses at least two of the following motion models: uniform motion model, turning motion model, and low-speed parking model.
3. The method for identifying the dynamic homeland of a ship according to claim 1, characterized in that, The ship spatiotemporal behavior data also includes an AIS ship name field, the ship static file data includes the registration location, and the ship dynamic location identification method further includes: The AIS ship name field is cleaned and formatted, and the pinyin characters at the beginning of the standardized AIS ship name field are extracted as the pinyin prefix of the ship name. Based on the preset mapping relationship between pinyin prefixes and administrative divisions, the pinyin prefixes of the ship name are resolved into at least one identifier resolution location and the resolution confidence level is recorded; When the pinyin prefix of the ship name can be mapped to multiple different administrative divisions, multiple candidate identifier resolution locations are retained, or the identifier resolution location is marked as pending confirmation.
4. The method for identifying the dynamic homeland of a ship according to claim 1, characterized in that, The multi-source data also includes port business data, which contains port names from multiple sources; the method further includes standardizing the port names in the port business data, specifically through the following steps: According to the preset priority, port names from different sources are mapped to a unified port code by matching port code, matching port name prefix, and matching character sequence similarity in sequence. When the matching confidence of the current matching method reaches the confidence threshold corresponding to the matching method, the subsequent matching steps are stopped.
5. The method for identifying the dynamic homeland of a ship according to claim 1, characterized in that, The step of identifying the berthing and operational behaviors of the target vessel based on the behavioral points, and segmenting the vessel's spatiotemporal behavior data into voyage trajectories, specifically includes: Based on the geofence of the preset port, the docking event is identified by combining the speed and duration of the behavior point, and the corresponding arrival and departure events are determined. Using adjacent departure and arrival events as boundaries, the continuous spatiotemporal behavior data of the ships is divided into multiple single-voyage trajectory segments; The docking behavior characteristics include at least two of the following: total cumulative docking time, docking frequency, average docking time per trip, continuous docking time, number of dockings at fixed berths, and the proportion of resupply and unloading behavior. The operation distribution characteristics include at least two of the following: cumulative grid operation time, operation point density, number of consecutive operation days, seasonal operation distribution, and proportion of cross-sea area navigation. The sea area for operation is divided and statistically analyzed using equal-area grid cells in a projected coordinate system.
6. The method for identifying the dynamic homeland of a ship according to claim 1, characterized in that, The step of constructing a spatiotemporal joint distance based on the geographic spatial distance and temporal continuity distance between the behavioral points, and clustering the behavioral points based on the spatiotemporal joint distance, specifically includes: Select any two behavioral points within the same time window, calculate the geographic spatial distance and the temporal continuous distance between the two behavioral points, and standardize the geographic spatial distance and the temporal continuous distance respectively. The spatiotemporal joint distance is obtained by weighting and summing the squared values of the standardized geospatial distance and the standardized temporal continuous distance according to the spatial weighting coefficient and the temporal weighting coefficient, and taking the square root. By combining spatial neighborhood threshold, temporal continuity threshold and minimum number of behavior points, density clustering is performed on the behavior points within the same time window, and the set of behavior points that meet the conditions is determined as the stable behavior cluster. The geospatial distance is calculated using the Haversine great circle distance, or the planar distance is calculated after conversion to a projected coordinate system.
7. The method for identifying the dynamic homeland of a ship according to claim 1, characterized in that, The calculation of the comprehensive score for each of the candidate territories specifically includes: For each candidate territory, the affiliation behavior characteristics and the operation distribution characteristics are normalized and weighted to obtain the affiliation behavior score and the operation sea area association score corresponding to the candidate territory. The continuity bonus is determined by combining the continuous duration of stable behavior clusters within the candidate territory, and the short-term transit penalty is determined by combining the stop duration, stop frequency and operation cluster formation within the candidate territory. After weighting and summing the affiliation behavior score, the operation sea area association score, and the continuity bonus, the weighted short-term transit penalty item is subtracted to obtain the initial territory score corresponding to the candidate territory area. Obtain the time difference between the current time window and the most recent valid behavior record in the candidate territory area; generate an exponential time decay factor based on the time difference and a preset decay coefficient; multiply the initial territory score by the time decay factor to obtain the territory comprehensive score after historical decay correction. When the time difference exceeds the preset failure threshold, the attenuation coefficient is increased, or the candidate region is marked as a failure candidate region.
8. The method for identifying the dynamic homeland of a ship according to claim 1, characterized in that, The step of combining the amount of valid behavioral data and the timeliness of the data within the current time window to perform a sufficiency verification of the primary attribution evidence for the candidate primary attribution location specifically includes: The candidate territories are sorted from highest to lowest according to their comprehensive territories score, and the candidate territories with the highest comprehensive territories score are determined as the candidate primary territories. The candidate primary affiliations are classified and determined based on their location score level, leading score, amount of effective behavioral data, and data timeliness. When the highest comprehensive score of the territory reaches the first score threshold, the difference between the highest comprehensive score of the territory and the second highest comprehensive score of the territory reaches the leading difference threshold, and the number of effective positioning points, effective flight times, effective docking times, and effective operation points within the current time window all reach the corresponding minimum evidence threshold and the data time interval is not greater than the timeliness threshold, the sufficiency verification of the primary territory evidence is deemed to have passed, the candidate primary territory is determined as the actual activity territory, and the determination status is set to confirmed. When the highest comprehensive score of the territory reaches a second score threshold that is lower than the first score threshold, or when some conditions such as the amount of effective behavioral data or the timeliness of data are not met, the judgment status is set to low confidence. When the highest comprehensive score of the territory is lower than the second score threshold, or any evidence indicator fails to reach the threshold or the data timeliness expires, the judgment status is set to pending confirmation or insufficient data. If the business system must display a reference territory, the registered territory in the ship's static file data will be used as the reference territory for output, and the source of the reference territory will be marked as registration information.
9. The method for identifying the dynamic homeland of a ship according to claim 3, characterized in that, The step involves merging or segmenting the time windows based on the candidate primary affiliation, comprehensive affiliation score, and primary affiliation evidence sufficiency verification results corresponding to multiple consecutive time windows, generating dynamic affiliation identification results with effective time intervals and confidence levels. Specifically, this includes: Compare the candidate primary affiliation with the comprehensive score of the location corresponding to two adjacent time windows; When the candidate primary home locations corresponding to two adjacent time windows are the same, and the change in the corresponding comprehensive score of the home location does not reach the segmentation change threshold, the two adjacent time windows are merged. When the candidate primary home locations corresponding to two adjacent time windows are different, and at least one of the differences in the comprehensive scores of the corresponding home locations, the changes in the ranking of the candidate home locations, or the changes in confidence reaches the segmentation change threshold, a home location time boundary point is set between the two adjacent time windows. The effective start time and the invalidation end time are generated based on the local time boundary point; When the same candidate territory area repeatedly forms the stable behavior cluster in multiple time windows corresponding to the same fishing season, but does not meet the main territory determination conditions in a continuous year-round time window, the candidate territory area is marked as a seasonal temporary territory. After generating the dynamic territorial identification result, the actual activity territoriality determined by the model is extracted from the dynamic territorial identification result; The actual activity location, the identifier resolution location, and the registration location are compared pairwise; When the geographical distance, administrative division level difference, duration of inconsistency and corresponding confidence level between any two types of territories meet the territorial conflict conditions, a territorial conflict marker is generated, and the conflict type, conflict level, involved territories and source of evidence are output. The conflict types include conflicts between the identifier resolution location and the registration location, conflicts between the identifier resolution location and the actual activity location, and conflicts between the registration location and the actual activity location.
10. An electronic device, characterized in that, include: A memory and a processor, the memory for storing a computer program and the processor for executing the computer program to implement the ship dynamic territorial identification method as described in any one of claims 1 to 9.
