A data-driven ship-bridge collision risk early warning method and system

Abstract

The application discloses a kind of data-driven ship bridge collision risk early warning method and system, the method includes the following steps: determining the standardized monitoring area of target bridge, and the historical AIS trajectory data sequence of the area is clustered and is handled, corresponding channel space area is generated;Based on the real-time AIS data of target ship current time and the historical AIS trajectory data sequence in previous preset time window that channel space area is in navigation state, corresponding predicted AIS trajectory data sequence is sequentially calculated;Obtain meteorological sea state data and carry out space-time alignment with real-time AIS data and predicted AIS trajectory data sequence;The risk index including risk of leaving channel, meteorological sea state risk and abnormal risk of steering is scored to each target ship;The scoring results of each risk index are weighted and fused using TOPSIS entropy weight analysis method, and the evaluation result of the comprehensive risk evaluation index of each target ship is obtained, and target ship is risk early warning according to the evaluation result.

Description

Technical Field

[0001] This invention relates to the field of intelligent management and control of water traffic safety and shipping big data analysis technology, and in particular to a data-driven method and system for early warning of ship-bridge collision risks. Background Technology

[0002] As cross-river and cross-sea transportation infrastructure, bridges typically present unique challenges in their navigable waterways, including channel diversion, pier constraints, and high vessel density. Vessels navigating these waterways are susceptible to deviations, boundary crossings, excessively sharp turns, or improper speed control due to factors such as tides, crosswinds, visibility variations, and traffic management. Collisions between vessels and bridge piers, abutments, or other structural elements can lead to serious consequences, including injuries, structural damage, waterway disruption, and secondary environmental pollution. Therefore, developing a collision risk early warning system for bridge areas to accurately identify and provide early warnings of vessel risks is crucial for improving maritime traffic safety.

[0003] The existing bridge collision avoidance and risk warning technology system mainly includes the following three types of methods: The first type is the target detection and close-range warning method based on perception means such as video and radar. This type of method is mainly used to perceive the operating status of ship targets, bridge structures and their adjacent areas in real time, and trigger an alarm when the target approaches the danger distance; The second type is the dangerous trajectory identification method based on rule thresholds. This type of method mainly judges whether a ship is in a dangerous navigation state based on preset distance thresholds, speed thresholds, heading deviation thresholds, boundary crossing rules or behavior rules; The third type is the ship motion prediction and collision probability estimation method based on physical dynamics. This type of method mainly uses ship kinematics or dynamics models to predict and evaluate the ship's future navigation trajectory, attitude changes and collision probability.

[0004] The three methods mentioned above have all played a certain role in the engineering practice of bridge collision avoidance and risk early warning, but they still have significant shortcomings in complex navigation scenarios in bridge areas: First, some methods rely heavily on single sensors, fixed thresholds, or static rules. When the traffic flow density in the bridge area is high, the types of vessels vary greatly, meteorological and hydrological conditions change rapidly, or multiple sources of interference coexist, the environmental adaptability of these methods is insufficient, easily leading to false alarms, missed alarms, or delayed alarms. Second, some methods do not adequately express the spatial constraints of the bridge area, failing to integrate elements such as waterway boundaries, bridge opening navigation ranges, prohibited navigation areas, restricted navigation areas, areas adjacent to bridge structures, and navigation safety buffer zones into the risk assessment process in a unified, structured, and computable manner, thus leading to... First, there are discrepancies between the results and actual regulatory scenarios, management rules, and navigation semantics. Second, bridge collision risks are usually not triggered by a single factor, but rather by the combined effects of multiple factors such as track deviation, abnormal speed, unstable course, abnormal maneuvering, changes in ship-shore distance, limited navigation space, and deteriorating meteorological and hydrological conditions. Existing methods still have shortcomings in multi-factor correlation modeling, unified expression of risk mechanisms, and consistency of indicators, making it difficult to form a unified early warning indicator that is interpretable, quantifiable, and comparable. Third, some methods still have limitations in real-time performance, robustness, and engineering deployment, making it difficult to meet the application requirements of regulatory systems for continuous operation, online updates, stable output, and traceable risk quantification results. Based on the above problems, existing technologies are still unable to achieve ship-bridge collision risk early warning in the complex traffic environment of bridge areas that balances perception accuracy, spatial constraint consistency, multi-factor coupling representation capabilities, and real-time deployment capabilities.

[0005] In conclusion, ship-bridge collision risk monitoring is a crucial foundational capability within the maritime regulatory system. It undertakes the task of continuously sensing and assessing the high-density traffic in bridge areas, and provides direct data support for maintaining waterway order, identifying and warning of risks, and ensuring the structural safety of bridges. Under complex conditions such as fluctuating visibility, increased tidal currents and crosswinds, and adjustments to navigation organization, stable and reliable bridge area monitoring can significantly improve the efficiency of regulatory authorities in detecting and responding to abnormal navigation behavior, and reduce the systemic impact of collision incidents on personnel safety, navigation efficiency, and bridge operation. Summary of the Invention

[0006] To address the low accuracy of ship-bridge collision risk warnings in complex traffic environments, caused by insufficient environmental adaptability, incomplete spatial constraint representation, and inadequate multi-factor coupling modeling in traditional technologies, this invention proposes a data-driven ship-bridge collision risk warning method. Within the bridge monitoring domain, based on historical, current, and future ship trajectories, and integrating channel boundary constraints, meteorological and sea condition factors, and ship maneuvering characteristics, a multi-dimensional risk score is performed. A multi-risk indicator fusion decision algorithm is then used to achieve dynamic real-time assessment and graded warning of the comprehensive risk of ship-bridge collisions. This invention also relates to a data-driven ship-bridge collision risk warning system.

[0007] The technical solution of the present invention is as follows:

[0008] A data-driven method for early warning of ship-bridge collision risks includes the following steps:

[0009] The steps for generating a waterway spatial region are as follows: First, acquire the structural and spatial data of the target bridge, including bridge centerline information, bridge structural height parameters, and the bridge monitoring area. Second, perform coordinate system normalization on the structural and spatial data to construct a standardized monitoring area for the target bridge under a unified geographic reference system. Third, acquire historical vessel crossing records corresponding to the target bridge and form a bridge crossing event set. Fourth, associate the vessel types corresponding to each bridge crossing event and filter the bridge crossing event set according to preset vessel type conditions to obtain a valid crossing event set. Fifth, extract historical AIS trajectory data sequences corresponding to the valid crossing event set based on the standardized monitoring area, summarize them to form a bridge area trajectory point set, and perform preprocessing operations. Sixth, use a hierarchical density clustering algorithm to spatially cluster the preprocessed bridge area trajectory point set to obtain several waterway clusters. Finally, perform surface processing on each waterway cluster to generate the corresponding waterway spatial region.

[0010] The steps for predicting future ship trajectories are as follows: First, identify all target ships currently in operation within the standardized monitoring area and their corresponding ship types. Then, acquire the real-time AIS data of each target ship at the current moment, as well as a sequence of historical AIS trajectory data within a preset time window, forming an AIS trajectory data sequence to be analyzed. Based on the real-time AIS data corresponding to each target ship, and combined with a preset single prediction time step, use a constant speed and direction trajectory extrapolation method to sequentially calculate the east-west and north-south displacement components of each target ship within each single prediction time step. After repeating the preset prediction steps, the predicted AIS trajectory data sequence for each target ship is obtained.

[0011] Meteorological and sea state data acquisition and spatiotemporal alignment steps: acquire multi-source meteorological and sea state data within the entire time range from the current moment to the predicted AIS trajectory data sequence and within the spatial range corresponding to the standardized monitoring area; through spatiotemporal alignment processing, match the multi-source meteorological and sea state data with the real-time AIS data of each target vessel and each AIS data in the predicted AIS trajectory data sequence;

[0012] Multi-dimensional risk quantification and identification steps: For each target vessel, a multi-dimensional index quantification and scoring method is used to score risk indicators including channel departure risk, weather and sea state risk, and abnormal maneuvering risk, to achieve quantitative identification of ship-bridge collision risk. Specifically, this includes: scoring the risk of the vessel leaving the channel based on the relative position of each AIS data point in the real-time AIS data and the predicted AIS trajectory data sequence with the channel spatial area that matches the vessel type of the target vessel, combined with a preset safety boundary distance threshold; and scoring the risk of the vessel leaving the channel based on the real-time AIS data and the predicted AIS data of the target vessel. The multi-source meteorological and sea state data corresponding to each AIS data point in the trajectory data sequence, combined with the target vessel's vessel type and a preset vessel type-meteorological and sea state element level mapping table, determine the risk level of each meteorological and sea state element under each AIS data point. The risk levels are then fused using the most unfavorable element-dominated principle, and the fusion result is normalized to obtain the meteorological and sea state risk score. Based on the heading change rate and speed change rate between adjacent AIS data points in the AIS trajectory data sequence to be analyzed, and combined with the heading change rate anomaly judgment threshold and speed change rate anomaly judgment threshold, the maneuvering anomaly risk is scored.

[0013] Ship-bridge collision risk assessment and early warning steps: For each ship type, the TOPSIS entropy weight analysis method is used to weight and fuse the exit risk score, meteorological and sea state risk score, and maneuver anomaly risk score of each target ship in the aforementioned ship type to obtain the assessment result of the comprehensive ship-bridge collision risk assessment index for each target ship; based on the assessment result of the comprehensive ship-bridge collision risk assessment index corresponding to each target ship, a risk warning is issued for the target ship.

[0014] Preferably, in the waterway spatial region generation step, the steps of obtaining historical vessel crossing records corresponding to the target bridge and forming a bridge crossing event set, filtering the bridge crossing event set to obtain a valid crossing event set, and extracting historical AIS trajectory data sequences corresponding to the valid crossing event set, and summarizing them to form a bridge area trajectory point set, specifically include:

[0015] The database is used to retrieve all vessels that have passed through the target bridge within the monitoring area during the historical period, and the vessel identification and passage information of these vessels are obtained. The passage information includes the time and direction of passage of the vessels through the target bridge.

[0016] For each of the aforementioned passage information, a time window for extracting historical AIS trajectory data sequences of a preset fixed duration is constructed with the crossing time as the center; taking the passage of a ship over the target bridge as a crossing event, the bridge crossing event set of the target bridge is constructed by combining the ship identification, passage information and its corresponding historical AIS trajectory data sequence extraction time window;

[0017] For each crossing event in the bridge crossing event set, the static information of the vessel associated with the crossing event is obtained through the vessel identifier corresponding to the crossing event. The static information of the vessel includes the vessel type.

[0018] The crossing events are filtered according to preset ship type conditions, and crossing events that do not meet the ship type requirements are eliminated to obtain a set of valid crossing events.

[0019] Based on the ship identifiers and historical AIS trajectory data sequences corresponding to each crossing event in the set of valid crossing events, time windows are extracted to obtain the historical AIS trajectory data sequences corresponding to each crossing event; the data volume of each obtained historical AIS trajectory data sequence is checked, and historical AIS trajectory data sequences that do not meet the preset data volume requirements are removed; the checked historical AIS trajectory data sequences are summarized to form a bridge area trajectory point set.

[0020] Preferably, in the ship future trajectory prediction step, the real-time AIS data includes longitude, latitude, timestamp, ground speed, and heading;

[0021] Using the real-time AIS data corresponding to each of the target vessels as the current point, the latitude and longitude of the current point are converted into projected coordinates, and the heading of the current point is converted into radians;

[0022] For each of the target vessels, based on the ground speed, heading and preset single prediction time step at the current point, the east-west displacement component and north-south displacement component of the current point are calculated using a constant speed and heading trajectory extrapolation method.

[0023] Add the projected coordinates of the current point to the corresponding north-south displacement component and east-west displacement component respectively to obtain the projected coordinates of the next predicted AIS data;

[0024] Using the projected coordinates of the next predicted AIS data as the new current point, repeat the above steps until the preset number of prediction steps is reached. Finally, convert the projected coordinates of all predicted AIS data into geographic coordinates to generate a sequence of predicted AIS trajectory data for each target vessel.

[0025] Preferably, in the step of acquiring and aligning meteorological and marine condition data in time and space, the multi-source meteorological and marine condition data includes wind speed, wave height, sea current speed, sea current direction, precipitation, tides, and visibility.

[0026] For each target vessel, the geographical location of each AIS data point in the real-time AIS data and predicted AIS trajectory data sequence of the target vessel is taken as the target location. Four spatial grid points closest to the target location are selected in the multi-source meteorological and sea state data. Based on the selected four spatial grid points, the target location is obtained by calculating the corresponding spatial interpolated meteorological and sea state data through inverse distance weighted interpolation. Then, the multi-source meteorological and sea state data is matched to each AIS data point in the real-time AIS data and predicted AIS trajectory data sequence of each target vessel to achieve spatial dimension alignment.

[0027] For each target vessel, the timestamp of each AIS data point in the real-time AIS data and predicted AIS trajectory data sequence of the target vessel is taken as the target time. Two time grid points adjacent to the target time in the time dimension are selected from the spatially aligned meteorological and sea state data. Linear interpolation is performed on the multi-source meteorological and sea state data corresponding to the selected two time grid points to obtain the multi-source meteorological and sea state data corresponding to the target time. Then, the multi-source meteorological and sea state data is matched to each AIS data point in the real-time AIS data and predicted AIS trajectory data sequence of each target vessel to achieve time dimension alignment.

[0028] Preferably, the step of scoring the risk of a ship leaving the waterway in the multi-dimensional risk quantification identification step specifically includes:

[0029] Based on the relative position of each AIS data point in the real-time AIS data and the predicted AIS trajectory data sequence with the channel space area that matches the ship type of the target ship, determine the relative positional relationship between each AIS data point and the channel space area that matches the ship type of the target ship.

[0030] If at least one AIS data point exists outside the waterway spatial area, the risk score for the vessel leaving the waterway is 1.

[0031] If all AIS data are within the waterway space area, then determine the minimum distance between the trajectory points of all AIS data in the real-time AIS data and the predicted AIS trajectory data sequence and the boundary of the waterway space area that matches the ship type of the target ship.

[0032] If the minimum distance is greater than or equal to the preset safety boundary distance threshold, then the risk score for the ship leaving the waterway will be 0.

[0033] If the minimum distance is less than the preset safety boundary distance threshold, then the risk score of the ship leaving the waterway is calculated based on the ratio of the minimum distance to the preset safety boundary distance threshold.

[0034] Preferably, in the multidimensional risk quantification and identification step, when scoring the risk of abnormal manipulation, the heading change rate of the adjacent AIS data is calculated based on the absolute value of the ratio of the heading difference to the timestamp difference in the adjacent AIS trajectory data sequence to be analyzed; the speed change rate of the adjacent AIS data is calculated based on the absolute value of the ratio of the speed difference to the timestamp difference in the adjacent AIS data.

[0035] Preferably, in the multidimensional risk quantification and identification step, when scoring the risk of abnormal manipulation, the average rate of change of heading of the AIS trajectory data sequence to be analyzed is calculated by using the rate of change of heading between adjacent AIS data in the AIS trajectory data sequence to be analyzed; based on the rate of change of heading between adjacent AIS data in the AIS trajectory data sequence to be analyzed and the average rate of change of heading, the degree of fluctuation of the heading adjustment behavior of the AIS trajectory data sequence to be analyzed is calculated.

[0036] Based on the rate of change of airspeed between adjacent AIS data in the AIS trajectory data sequence to be analyzed, the maximum rate of change of airspeed is determined.

[0037] Based on the degree of fluctuation and the maximum rate of change of speed, combined with the abnormal judgment thresholds for the rate of change of heading and the rate of change of speed, the abnormal control risk score is calculated.

[0038] Preferably, in the ship-bridge collision risk assessment and early warning step, before weighting and fusing the risk index scores of each target ship in the ship type using the TOPSIS entropy weight analysis method, the range standardization method is used to perform dimensionless processing on the risk index scores of each target ship in the ship type.

[0039] For each of the aforementioned ship types, based on the dimensionless processing of the risk index scores corresponding to each target ship in the aforementioned ship type, the TOPSIS entropy weight analysis method is used to determine the index weight coefficients of each risk index corresponding to each target ship in the aforementioned ship type.

[0040] For each of the aforementioned ship types, based on the dimensionless processing of the risk index scores and their corresponding index weight coefficients for each target ship in the aforementioned ship type, the standardized risk values ​​of each risk index for each target ship in the aforementioned ship type are calculated, and a weighted normalized decision matrix for the aforementioned ship type is constructed based on all the standardized risk values ​​under the aforementioned ship type.

[0041] For each of the aforementioned ship types, a positive ideal solution is constructed based on the maximum value of the standardized risk value of each risk indicator in the weighted normalized decision matrix corresponding to the ship type; and a negative ideal solution is constructed based on the minimum value of the standardized risk value of each risk indicator in the weighted normalized decision matrix corresponding to the ship type.

[0042] For each target vessel in each of the aforementioned vessel types, the evaluation result of the ship-bridge collision comprehensive risk assessment index of the target vessel is calculated based on the standardized risk value, positive ideal solution, and negative ideal solution of each risk indicator corresponding to the target vessel.

[0043] Preferably, in the ship-bridge collision risk assessment and early warning step, based on the assessment results of the ship-bridge collision comprehensive risk assessment index corresponding to each target ship, and combined with the preset level classification threshold, the ship-bridge collision risk level of each target ship is classified into normal, caution, and danger.

[0044] Risk warnings are issued for target vessels whose ship-bridge collision risk level is deemed dangerous.

[0045] A data-driven ship-bridge collision risk early warning system includes, in sequence, a waterway spatial area generation module, a ship future trajectory prediction module, a meteorological and sea state data acquisition and spatiotemporal alignment module, a multi-dimensional risk quantification and identification module, and a ship-bridge collision risk assessment and early warning module.

[0046] The waterway spatial region generation module is used to: acquire the structural and spatial data of the target bridge, including bridge centerline information, bridge structural height parameters, and bridge monitoring area; perform coordinate system normalization on the structural and spatial data to construct a standardized monitoring area of ​​the target bridge under a unified geographic reference system; acquire historical vessel crossing records corresponding to the target bridge and form a bridge crossing event set; associate the vessel types corresponding to each bridge crossing event, and filter the bridge crossing event set according to preset vessel type conditions to obtain a valid crossing event set; extract historical AIS trajectory data sequences corresponding to the valid crossing event set based on the standardized monitoring area, summarize them to form a bridge area trajectory point set and perform preprocessing operations; use a hierarchical density clustering algorithm to spatially cluster the preprocessed bridge area trajectory point set to obtain several waterway clusters, and perform surface processing on each waterway cluster to generate the corresponding waterway spatial region.

[0047] The ship future trajectory prediction module is used to: determine all target ships in operation within the standardized monitoring area and their corresponding ship types; acquire the real-time AIS data of each target ship at the current moment and extract the historical AIS trajectory data sequence within a preset time window from the current moment backward to form the AIS trajectory data sequence to be analyzed; based on the real-time AIS data corresponding to each target ship, combined with the preset single prediction time step, and using the constant speed and constant direction trajectory extrapolation method, calculate the east-west and north-south displacement components of each target ship within each single prediction time step in turn, and repeat the calculation for the preset number of prediction steps to obtain the predicted AIS trajectory data sequence of each target ship;

[0048] The meteorological and sea state data acquisition and spatiotemporal alignment module is used to acquire multi-source meteorological and sea state data within the entire time range covered by the predicted AIS trajectory data sequence from the current moment to the spatial range corresponding to the standardized monitoring area; and through spatiotemporal alignment processing, match the multi-source meteorological and sea state data with the real-time AIS data of each target vessel and each AIS data in the predicted AIS trajectory data sequence.

[0049] The multi-dimensional risk quantification and identification module: For each target vessel, a multi-dimensional index quantification and scoring method is used to score risk indicators including channel departure risk, weather and sea state risk, and abnormal maneuvering risk, to achieve quantitative identification of ship-bridge collision risk. Specifically, it includes: scoring the risk of the vessel leaving the channel based on the relative position of each AIS data point in the real-time AIS data and the predicted AIS trajectory data sequence with the channel spatial area that matches the vessel type of the target vessel, combined with a preset safety boundary distance threshold; and scoring the risk of the vessel leaving the channel based on the real-time AIS data and the predicted AIS trajectory data of the target vessel. The multi-source meteorological and sea state data corresponding to each AIS data point in the S-track data sequence are combined with the target vessel's vessel type and a preset vessel type-meteorological and sea state element level mapping table to determine the risk level of each meteorological and sea state element under each AIS data point. The risk levels are then fused using the most unfavorable element-dominated principle, and the fusion result is normalized to obtain the meteorological and sea state risk score. Based on the heading change rate and speed change rate between adjacent AIS data points in the AIS track data sequence to be analyzed, and combined with the heading change rate anomaly judgment threshold and the speed change rate anomaly judgment threshold, the maneuvering anomaly risk is scored.

[0050] The ship-bridge collision risk assessment and early warning module: For each ship type, the TOPSIS entropy weight analysis method is used to weight and fuse the exit risk score, meteorological and sea state risk score, and maneuver anomaly risk score of each target ship in the ship type to obtain the assessment result of the comprehensive ship-bridge collision risk assessment index of each target ship; based on the assessment result of the comprehensive ship-bridge collision risk assessment index corresponding to each target ship, a risk warning is issued for the target ship.

[0051] The beneficial effects of this invention are as follows:

[0052] This invention provides a data-driven method for early warning of ship-bridge collision risks. This data-driven risk warning method, tailored to bridge area scenarios, enables real-time monitoring of ship risks around bridges under multi-source data conditions, thereby improving the accuracy, robustness, and engineering usability of bridge collision risk warnings. The method automatically extracts the navigation channels in the bridge area from disordered trajectories into structured navigation space regions by standardizing the monitoring area under a unified geographic reference system and using hierarchical density clustering based on historical AIS trajectories. This provides a precise and calculable navigation space constraint basis for subsequent risk identification. It automatically extracts navigation channel clusters based on historical ship passage records and performs surface processing to generate corresponding navigation space regions. The construction and updating of navigation space regions can be completed without manual intervention, significantly reducing engineering deployment and maintenance costs and improving engineering usability. By standardizing the navigation status screening and ship type matching within the monitoring area, the effectiveness of the analysis object is ensured, and AIS trajectory data analysis is only performed on ships actually navigating, reducing interference from invalid data. Based on short-term AIS trajectory prediction of real-time speed and heading, the analysis of ship navigation status shifts from the "current moment" to... Extending to "future trends," this approach enables early prediction of potential risks such as vessel overtaking or approaching bridges, overcoming the limitations of traditional methods that only warn of the current state. This provides more time for risk warnings and improves their timeliness. Furthermore, the constant speed and direction trajectory calculation method balances computational efficiency and real-time performance, making it suitable for engineering-based real-time monitoring scenarios. Because it predicts short-term AIS trajectories, it eliminates the need for complex model training, ensuring the accuracy of the predictions. By restricting meteorological and sea state data acquisition from both spatial and temporal dimensions, the amount of data acquired is reduced, improving the efficiency of subsequent alignment calculations. The acquired multi-source meteorological and sea state data (such as wind speed, wave height, sea current speed, sea current direction, precipitation, tides, and visibility) comprehensively covers key factors affecting vessel maneuvering. The environmental factors comprehensively capture the complex meteorological and sea conditions in the bridge area, providing a comprehensive data foundation for the quantification of environmental risks. The optimal spatiotemporal alignment process, combining spatial reverse distance weighting and temporal linear interpolation, resolves the heterogeneity between meteorological grid data and ship AIS trajectory data in the spatiotemporal dimensions. This ensures that each AIS data point corresponds to precise environmental parameters, enabling coupled analysis of ship motion status and external environmental conditions. When assessing the risk of ships leaving the channel, the risk assessment is conducted in the channel space area appropriate to the target ship's type. Combined with a tiered scoring system based on preset safety boundary distance thresholds (i.e., distance to the boundary and boundary crossing), the channel risk assessment is more aligned with the ship's navigation characteristics, improving the rationality of channel risk identification.In meteorological and sea state risk assessment, a fusion method combining a pre-defined ship type-meteorological and sea state element level mapping table with the most unfavorable element-dominated principle is adopted. This fully considers the differences in environmental resistance capabilities among different ship types, accurately quantifies the impact of complex meteorological and sea state conditions on ship navigation, and avoids the one-sidedness of judging based on a single meteorological and sea state element. In maneuvering anomaly risk assessment, a score is based on the rate of change of course and speed, transforming the ship's maneuvering behavior characteristics into standardized quantitative indicators. This enables accurate identification of high-frequency abnormal maneuvering behaviors in bridge areas, such as excessively sharp turns and improper speed control. All three dimensions of risk scores are normalized to a unified range, forming comparable and fusion-compatible standardized risk indicators. This solves the problem of bias in single-dimensional judgments and improves the completeness of risk identification under complex navigation conditions. Integrity and stability; weighted fusion is performed separately according to ship type, fully considering the risk characteristics of different ship types and the differences in the impact of various risk indicators. The objective weighting method of TOPSIS entropy weight analysis avoids the subjective bias of manual assignment, making the comprehensive risk assessment results of ship-bridge collisions more consistent with the actual navigation risks of different ship types. Furthermore, it unifies and integrates multi-dimensional heterogeneous risk indicators into a single comprehensive risk assessment indicator, realizing the unified expression of multi-source risk information. Based on the comprehensive risk indicator, hierarchical early warning is carried out, realizing the quantitative classification and accurate alarm of risks. It can support the hierarchical handling and system linkage of regulatory departments, allowing the early warning results to be directly connected to the engineering monitoring and handling process, significantly improving the engineering usability and actual regulatory value of bridge collision risk early warning, and effectively reducing the probability of collision events in bridge areas. This invention, based on a multi-source data-driven approach, generates waterway spatial regions, predicts future ship trajectories, acquires and aligns meteorological and sea state data spatiotemporally, performs multi-dimensional risk scoring, and conducts comprehensive collision risk assessment and early warning. It combines hierarchical density clustering algorithms, constant-speed, constant-direction trajectory prediction methods, spatiotemporal dual-dimensional alignment, and TOPSIS entropy weight analysis. By integrating real-time and predicted AIS trajectory data of ships, effective passage areas of bridges, and multi-source meteorological and sea state information, it performs hierarchical identification and quantitative assessment of ship operational risks. This enables early identification and dynamic early warning of bridge-ship collision risks. It features strong real-time performance, good interpretability, sufficient environmental adaptability, and easy engineering deployment, providing systematic technical support for bridge collision avoidance monitoring and navigation safety management.

[0053] This invention constructs a fixed-length trajectory extraction time window centered on the ship's crossing time, unifying the truncation scale of historical AIS trajectory data for each ship. This avoids feature bias caused by different sampling rates or trajectory lengths for different ships and effectively filters irrelevant historical AIS trajectory data outside the bridge area. By associating ship static information (ship type) and filtering according to preset ship type conditions, crossing events that do not meet the analysis requirements are eliminated. By checking the data volume and eliminating historical AIS trajectory data sequences that do not meet the preset data volume requirements, this invention ensures that each historical AIS trajectory data sequence participating in clustering has sufficient spatial density and continuity, enabling the hierarchical density clustering algorithm to run stably and avoiding misjudgments of false clusters or noise points due to data sparsity, thus enhancing the robustness of the waterway clustering results.

[0054] This invention obtains east-west and north-south displacement components based on ground speed, heading, and fixed time step decomposition. It uses a constant speed and heading trajectory extrapolation method to achieve trajectory extrapolation, which can not only realistically reflect the motion law of ships in inertial navigation, conforming to the short-term navigation characteristics of ships, but also ensure that the prediction logic is simple and the calculation efficiency is high, meeting the engineering requirements of low latency and high response speed for real-time monitoring of bridge areas. The iterative recursive method generates the predicted trajectory point by point, using the prediction result of the previous step as the calculation basis for the next step, and can continuously output complete and coherent multi-step predicted trajectories. This ensures the smoothness and continuity of the trajectory sequence, and can flexibly adapt to the prediction duration requirements of different lengths, accurately covering the entire path of the ship from the current position to the crossing of the waterway space area.

[0055] This invention employs a spatial alignment method that uses the geographic location of each AIS data point as the target location and selects the four nearest spatial grid points for reverse distance weighted interpolation. This method assigns reasonable weights to grid points based on their distance from the ship's location, with closer meteorological observation points receiving higher weights. This fully reflects the real distribution pattern of meteorological elements in space, which is continuously and gradually changing, significantly improving the accuracy of meteorological and sea state data for the ship's location. Linear interpolation is used to complete the temporal alignment, accurately calculating the meteorological and sea state values ​​corresponding to the timestamp of each AIS data point. This ensures that the ship's navigation status at any given time matches the actual meteorological conditions at that time, achieving seamless temporal matching. This dual alignment method, combining spatial and temporal interpolation, constructs a complete spatiotemporal matching system. It enables a one-to-one correspondence between the originally gridded and discontinuous meteorological and sea state data and the continuous real-time AIS data and predicted AIS trajectory data, completely solving the problem of spatiotemporal inconsistency in multi-source heterogeneous data. This provides a standardized data foundation with complete spatiotemporal alignment for subsequent meteorological and sea state risk scoring.

[0056] This invention employs a hierarchical judgment logic of "directly determining high risk for vessels crossing boundaries + assessing edge distance for vessels not crossing boundaries." First, it quickly identifies high-risk situations where vessels clearly cross boundaries, directly assigning a score of 1 to achieve rapid judgment and early warning of high-risk behavior, thus improving the response efficiency of risk monitoring in bridge areas. Then, it performs a refined edge distance analysis on vessels that have not crossed boundaries, balancing the efficiency and refinement of risk judgment. For vessels that have not crossed boundaries, it combines a preset safety boundary distance threshold with a quantified score based on the ratio of the minimum distance to the preset safety boundary distance threshold. This transforms the spatial distance between the vessel and the waterway boundary into a standardized score value within the [0,1] interval, making edge-crossing risk quantifiable and comparable, turning the originally abstract risk of edge-crossing navigation into a concrete numerical indicator.

[0057] This invention scores the risk of maneuvering anomalies by calculating the rate of change of course of adjacent AIS data based on the absolute value of the ratio of the difference in course to the difference in timestamps in the AIS trajectory data sequence to be analyzed; and by calculating the rate of change of speed of adjacent AIS data based on the absolute value of the ratio of the difference in speed to the difference in timestamps in the same AIS data sequence. This scientifically quantifies the magnitude of changes in ship course and speed per unit time, accurately capturing maneuvering behaviors such as excessively sharp turns and sudden acceleration / deceleration. By focusing solely on the degree of change, it aligns with the need to determine maneuvering anomaly risks by focusing only on the severity of the behavior, making the calculation results more consistent with actual risk identification scenarios and improving the accuracy and rationality of maneuvering anomaly risk scoring.

[0058] This invention combines the abnormal heading rate of change threshold and the abnormal speed rate of change threshold to calculate a standardized score by integrating the fluctuation of heading adjustment behavior with the maximum speed rate of change. This takes into account the dual risks of heading and speed control, provides a unified judgment standard for the score results, objectively quantifies the degree of control anomaly, and the output standardized score can be directly integrated with other risk indicators to provide reliable and suitable control anomaly risk data support for subsequent comprehensive risk assessment.

[0059] This invention first uses range standardization to dimensionlessly process the scores of each risk indicator, eliminating the dimensional differences between different risk indicators. Then, it uses the TOPSIS entropy weight analysis method to determine the indicator weight coefficients separately for each ship type, objectively assigning weights based on the actual data differences of risk indicators for each ship type, making the allocation of indicator weight coefficients more closely match the navigation risk characteristics of different ship types. Based on the extreme values ​​of the standardized risk values ​​of each risk indicator in the weighted normalized decision matrix, positive and negative ideal solutions are constructed respectively, accurately anchoring the ideal high and low risk states of each ship type, providing a clear reference for comprehensive risk assessment. Finally, the comprehensive risk assessment index is calculated by combining the ship's standardized risk value with the positive and negative ideal solutions, unifying the multidimensional and heterogeneous risk scores into a single quantifiable assessment result. This achieves accurate ranking and classification of ship risks, significantly improving the objectivity, scientific rigor, and engineering practicality of the comprehensive assessment of bridge collision risks.

[0060] This invention also relates to a data-driven ship-bridge collision risk early warning system. This system corresponds to the aforementioned data-driven ship-bridge collision risk early warning method and can be understood as a system that implements the aforementioned data-driven ship-bridge collision risk early warning method. It includes a channel spatial area generation module, a ship future trajectory prediction module, a meteorological and sea state data acquisition and spatiotemporal alignment module, a multi-dimensional risk quantification and identification module, and a ship-bridge collision risk assessment and early warning module. These modules work collaboratively, integrating real-time and predicted AIS trajectories of ships, bridge channel spatial area information, and multi-source meteorological and sea state environmental information to perform hierarchical identification and quantification assessment of ship operational risks. This enables early identification and dynamic early warning of bridge collision risks, featuring strong real-time performance, good interpretability, and easy engineering deployment, providing systematic technical support for bridge collision avoidance monitoring and navigation safety management. The multi-dimensional risk quantification and identification module decomposes collision risks into three basic risks: deviating from the channel, meteorological and sea state, and abnormal maneuvering. It identifies and outputs level results from three dimensions: channel constraints, environmental impact, and ship behavior, reducing the bias caused by single risk indicators and improving the completeness and stability of risk identification under complex navigation conditions. The ship future trajectory prediction module uses real-time AIS data to predict short-term future trajectories. This allows for a joint assessment of current and future risks, expanding early warnings from "current status alerts" to "early trend warnings." This enables earlier detection of potential boundary crossings, abnormal maneuvers, or high-risk approach behaviors, improving the timeliness of bridge collision avoidance warnings. The ship-bridge collision risk assessment and early warning module employs the TOPSIS entropy weight analysis method to weight and fuse multiple risk indicators, forming a single comprehensive risk assessment that supports ranking and classification, achieving a unified expression of multi-source, heterogeneous risk information. Attached Figure Description

[0061] Figure 1 This is a flowchart of the data-driven ship-bridge collision risk early warning method of the present invention.

[0062] Figure 2 This is an example diagram of the waterway space area corresponding to the target bridge of the present invention.

[0063] Figure 3 This is an example diagram for visualizing the results of the integrated risk assessment of ship-bridge collisions according to the present invention.

[0064] Figure 4 This is a structural block diagram of the data-driven ship-bridge collision risk early warning system of the present invention. Detailed Implementation

[0065] The present invention will now be described with reference to the accompanying drawings.

[0066] This invention discloses a data-driven method for early warning of ship-bridge collision risks. It involves a method that, in bridge-area scenarios, comprehensively utilizes dynamic information from the Automatic Identification System (AIS), static information from the bridge, and meteorological and hydrological data from multiple sources. Through multi-indicator fusion decision-making, it achieves real-time assessment and early warning of ship-bridge collision risks. This method is applicable to scenarios such as bridge area supervision of navigation bridges, waterway operation support, Vessel Traffic Service (VTS) auxiliary decision-making, and port and shipping safety management. It can provide risk quantification output and early warning basis for bridge collision warning systems, smart waterway platforms, and maritime traffic situational awareness systems.

[0067] This invention first acquires and collects basic bridge information, standardized monitoring area information, and historical vessel AIS trajectory data for that area. Then, it uses a hierarchical density clustering algorithm to spatially cluster the historical AIS trajectories, obtaining upstream and downstream channel spatial areas, which serve as the data foundation for the entire algorithm system for identifying, comprehensively assessing, and warning of navigation risks in the bridge area. Next, it acquires all vessel information within a preset time period within the current standardized monitoring area of ​​the bridge, as well as multi-source meteorological and sea state data for that time period, and performs cleaning and preprocessing on the raw data. Finally, based on real-time vessel AIS data, it establishes a constant-speed, constant-direction navigation trajectory... Short-term vessel trajectory prediction is performed using an extrapolation method, and the current trajectory point, predicted trajectory point, and historical trajectory point of AIS are spatiotemporally matched with meteorological and sea state data. Based on this, three risk scoring levels are set according to the degree of channel deviation, abnormal vessel maneuvering, and meteorological and sea state element level rules appropriate to the vessel type, and corresponding multi-dimensional risk monitoring algorithms are constructed. Finally, the TOPSIS entropy weight analysis method is used to automatically calculate the weight of each indicator based on various risk scoring data, and the resulting comprehensive early warning index for ship-bridge collision risk monitoring is obtained, realizing continuous early warning intensity output and hierarchical alarm linkage for business purposes.

[0068] In summary, this invention, based on a data-driven approach, utilizes historical AIS track clustering to characterize the spatial morphology of waterways in bridge areas and performs spatiotemporal matching of meteorological and sea state factors with real-time and predicted AIS trajectories to achieve a comprehensive assessment of risks under complex environmental and traffic fluctuation conditions in bridge areas. Simultaneously, through multi-indicator risk identification and quantification, it outputs a single, continuously traceable early warning intensity indicator, facilitating tiered regulatory response and system linkage, thereby improving the accuracy, interpretability, and engineering usability of early warnings. Specifically, such as... Figure 1 As shown, the data-driven ship-bridge collision risk early warning method of the present invention includes the following steps:

[0069] I. Waterway Spatial Region Generation Steps: 1. Obtain the structural and spatial data of the target bridge, including bridge centerline information, bridge structural height parameters, and the bridge monitoring area. 2. Perform coordinate system normalization on the structural and spatial data to construct a standardized monitoring area for the target bridge under a unified geographic reference system. 3. Obtain historical vessel crossing records corresponding to the target bridge and form a bridge crossing event set. 4. Associate the vessel types corresponding to each bridge crossing event and filter the bridge crossing event set according to preset vessel type conditions to obtain a valid crossing event set. 5. Extract historical AIS trajectory data sequences corresponding to the valid crossing event set based on the standardized monitoring area (the historical AIS trajectory data sequence includes several historical AIS data points, each including fields such as vessel identifier MMSI, navigation status, latitude, longitude, timestamp, ground speed, and heading), summarize them to form a bridge area trajectory point set, and perform preprocessing operations. 6. Use a hierarchical density clustering algorithm to spatially cluster the preprocessed bridge area trajectory point set to obtain several waterway clusters. 7. Perform surface processing on each waterway cluster to generate the corresponding waterway spatial region.

[0070] The embodiments of the present invention can obtain structural data, including bridge centerline information and bridge structural height parameters, as well as spatial data including the bridge monitoring area, from the bridge basic information database. Then, all the obtained bridge data are processed by coordinate system standardization to construct a standardized monitoring area of ​​the target bridge under a unified geographic reference system, which is used to limit the spatial range of subsequent ship AIS data screening and navigation channel analysis.

[0071] Secondly, the system retrieves all vessels that have passed through the target bridge within a historical period (e.g., within one year) from the bridge crossing event database, and obtains their vessel identification and passage information. The passage information includes the time and direction of passage of the vessel through the target bridge. It should be noted that within a historical period, each vessel may correspond to multiple passage information records, which may include different directions of passage.

[0072] For each of the aforementioned passage information, a time window for extracting historical AIS trajectory data sequences of a preset fixed duration is constructed, centered on the crossing time; taking the ship's passage through the target bridge as a crossing event (each crossing event corresponds to a unique index used to distinguish crossing events), the ship's identification, passage information, and its corresponding historical AIS trajectory data sequence extraction time window are combined to construct a bridge crossing event set for the target bridge;

[0073] For each crossing event in the bridge crossing event set, the ship static information (such as ship type, length, width, load capacity, etc.) of the ship associated with the crossing event is obtained from the ship static information database through the ship identifier corresponding to the crossing event.

[0074] Next, the crossing events are filtered according to the preset ship type conditions, and crossing events that do not meet the ship type requirements are eliminated to obtain a set of valid crossing events for waterway analysis.

[0075] Then, based on the ship identifiers and historical AIS trajectory data sequences corresponding to each crossing event in each set of valid crossing events, time windows are extracted. The historical AIS trajectory data sequences corresponding to each crossing event are retrieved and extracted from the database storing historical AIS data. The data volume of each of the obtained historical AIS trajectory data sequences is checked, and historical AIS trajectory data sequences that do not meet the preset data volume requirements (such as 50 data points) are removed. The checked historical AIS trajectory data sequences are summarized to form a bridge area trajectory point set and preprocessed. These preprocessed data are used as input data for the hierarchical density clustering algorithm.

[0076] Finally, the trajectory point set of the bridge area is subjected to quality filtering and spatial constraint screening, and uniformly transformed to a plane projection coordinate system to facilitate clustering calculations. Then, a hierarchical density clustering algorithm is used to spatially cluster the preprocessed trajectory point set of the bridge area to obtain several navigation channel clusters. Each navigation channel cluster is then surface-processed to generate corresponding navigation channel spatial regions. Specifically, quality filtering removes historical AIS data with abnormal location jumps, missing attributes (such as missing latitude fields), or time anomalies (such as abnormal timestamp fields), while spatial constraint screening removes historical AIS data whose latitude and longitude coordinates are not within the navigation channel spatial region.

[0077] The specific steps of clustering calculation in this embodiment of the invention include:

[0078] First, calculate any two historical AIS data points (also known as trajectory points) in the bridge area trajectory point set after the plane projection coordinate transformation. , European distance The calculation formula is as follows:

[0079]

[0080] in, Represents trajectory points The x-coordinate of the plane projection. Represents trajectory points The y-coordinate of the plane projection. Represents trajectory points The x-coordinate of the plane projection. Represents trajectory points The y-coordinate of the plane projection.

[0081] Then for any trajectory point Set the minimum neighborhood parameter k, and define The core distance is the distance to the k-th nearest neighbor, and the calculation formula is as follows:

[0082]

[0083] in, express The k-th nearest neighbor distance, Indicates and The trajectory point of the k-th nearest neighbor.

[0084] Secondly, the distance between the two trajectory points is used to reflect the trajectory points. and The formula for calculating the distance between two points, based on their spatial proximity and local density characteristics, is as follows:

[0085]

[0086] in, express and The reachability distance is given by max(.), which means taking the maximum value among them.

[0087] Then, based on all the calculated reach distances, a hierarchical density structure of trajectory points is constructed, combined with a preset minimum cluster size parameter. (e.g., 200), the constructed hierarchical density structure is aggregated to obtain the final clustering result, that is, each trajectory point corresponds to a cluster label, which can be represented by the following expression:

[0088]

[0089] in, Represents trajectory points Clustering label values, the label values ​​can be Any value within, The local density of the trajectory point is too low and the distance between it and other trajectory points is too large, so it cannot be classified into any traffic channel cluster and is therefore a noise point; K represents the number of traffic channel clusters identified. A bridge generally includes at least two traffic channel clusters.

[0090] Finally, in this embodiment, each channel cluster is encapsulated using polygonal boundaries to perform surface processing, generating corresponding channel spatial regions. ( This refers to the closed navigable surface area obtained after the navigable channel clustering process, which can also be called a multi-clustered channel region. Figure 2 As shown, the pink area represents the waterway space region corresponding to the target bridge. The waterway space region of the target bridge is divided into two separate regions, each of which is an irregular polygon. The upper region is the clustering result of ships sailing north (upward), and the lower region is the clustering result of ships sailing south (downward).

[0091] This invention provides a standardized prior input for the "structural constraints - navigation channel" of the bridge area by acquiring and standardizing basic bridge data (structural and spatial data), extracting historical crossing trajectories, and completing the clustering of trajectory points in the bridge area and the generation of the waterway spatial area. This provides a basic support for subsequent ship navigation status identification and bridge collision risk assessment / early warning calculation.

[0092] II. Ship Future Trajectory Prediction Steps: Identify all target ships in operation within the standardized monitoring area and their corresponding ship types, and acquire the real-time AIS data of each target ship at the current moment, as well as the historical AIS trajectory data sequence within a preset time window traced back from the current moment, forming the AIS trajectory data sequence to be analyzed; Based on the real-time AIS data corresponding to each target ship, combined with the preset single prediction time step, using a constant speed and constant direction trajectory extrapolation method, sequentially calculate the east-west and north-south displacement components of each target ship within each single prediction time step, and repeat the preset prediction steps to obtain the predicted AIS trajectory data sequence of each target ship.

[0093] The AIS trajectory data sequence to be analyzed includes a real-time AIS data point at the current moment and several consecutive historical AIS data points extracted from the current moment back within a preset time window; the predicted AIS trajectory data sequence includes several consecutive predicted AIS data points. Each real-time AIS data point, historical AIS data point, and predicted AIS data point includes fields such as vessel identification MMSI, navigation status, latitude, longitude, timestamp, speed over ground (SOG, in km), and heading (HDG).

[0094] Specifically, in this embodiment of the invention, based on the standardized monitoring area of ​​the target bridge, several real-time AIS data are filtered to extract the ships currently in the standardized monitoring area, and the ships are filtered according to the navigation status field in the real-time AIS data to select the ships in the navigation state as the target ships for subsequent analysis.

[0095] When predicting the future trajectory of target vessels, the real-time AIS data corresponding to each target vessel is used. ( Longitude Latitude Using the timestamp as the current point, convert the latitude and longitude of the current point into projected coordinates, and also convert the heading of the current point... Convert to radians;

[0096] For each of the target vessels, based on the ground speed v and heading at the current point... and the preset single prediction time step (e.g., 30 seconds), using a constant speed and direction trajectory extrapolation method, calculate the east-west displacement component and the north-south displacement component of the current point;

[0097] Add the projected coordinates of the current point to the corresponding north-south displacement component and east-west displacement component respectively to obtain the projected coordinates of the next predicted AIS data;

[0098] The next step is to predict the projection coordinates of the AIS data. Convert to geographic coordinates To obtain the next step of predicting AIS data ;

[0099] Using the projected coordinates of the next predicted AIS data as the new current point, repeat the above steps until a preset number of prediction steps N (e.g., 20 steps) is reached, ultimately obtaining the projected coordinates of all predicted AIS data. Convert to geographic coordinates This generates a continuous sequence of predicted AIS trajectory data for each target vessel.

[0100] The calculation formula is as follows:

[0101]

[0102]

[0103]

[0104]

[0105] in, for Projected coordinates This represents the projected x-coordinate of the current point. This represents the projected y-coordinate of the current point. If the current point is the target vessel, it corresponds to the real-time AIS data. ,but and equal, and equal; This represents the projected x-coordinate of the next predicted AIS data for the current point. This represents the projected y-coordinate of the current point in the next predicted AIS data. This represents the east-west displacement component of the current point within a short timescale, assuming the ship maintains constant speed and heading. This represents the north-south displacement component of the current point within a short timescale, assuming the ship maintains its speed and heading unchanged; k represents the prediction step number. This represents the geographic coordinates of the predicted AIS data.

[0106] Based on a standardized monitoring area, this invention completes real-time identification of ships in navigation within the monitoring area, ship type filtering, and extraction of historical AIS data, and further generates a predicted AIS trajectory data sequence, thereby providing dynamic input for subsequent navigation status identification and bridge collision risk calculation based on integrated "real-time-historical-predictive" AIS data.

[0107] III. Meteorological and Sea State Data Acquisition and Spatiotemporal Alignment Steps: Within the entire time range from the current moment to the predicted AIS trajectory data sequence and within the spatial range corresponding to the standardized monitoring area, acquire multi-source meteorological and sea state data (such as wind speed, wave height, sea current speed, sea current direction, precipitation, tides, visibility, etc.); through spatiotemporal alignment processing, match the multi-source meteorological and sea state data with the real-time AIS data of each target vessel and each AIS data in the predicted AIS trajectory data sequence.

[0108] The multi-source meteorological and marine condition data in this embodiment of the invention are collected within the time range corresponding to the real-time AIS data and the predicted AIS trajectory data sequence (i.e., the AIS ship time window) and within the spatial range of the standardized monitoring area. These multi-source meteorological and marine condition elements include wind speed, wave height, sea current speed, sea current direction, precipitation, tides, and visibility. This can reduce the amount of data in subsequent meteorological and marine condition data alignment processing and improve the alignment processing efficiency.

[0109] The specific steps of the alignment process include: for each target vessel, using the geographical location of each AIS data point in the real-time AIS data and predicted AIS trajectory data sequence of the target vessel as the target location. Select the four spatial grid points closest to the target location from the multi-source meteorological and sea state data. By using reverse distance-weighted interpolation, spatially interpolated meteorological and sea state data corresponding to the target location are obtained. The calculation formula is as follows:

[0110]

[0111] in, This represents the weighting coefficient of the back distance of the i-th spatial grid point. The closer the spatial grid point is to the target location, the greater its influence on the interpolation result. This represents the spatial distance between the target location and the i-th spatial grid point. This represents the multi-source meteorological and sea state data at the i-th spatial grid point;

[0112] After interpolation calculations for all target locations, the multi-source meteorological and sea condition data can be matched to each AIS data in the real-time AIS data and predicted AIS trajectory data sequence of each target vessel, thereby aligning each AIS data with the multi-source meteorological and sea condition data in the spatial dimension. In other words, each AIS data corresponds one-to-one with all meteorological and sea condition element data in the spatial dimension.

[0113] For each target vessel, the timestamp of each AIS data point in the real-time AIS data and predicted AIS trajectory data sequence of the target vessel is taken as the target time t. Two time grid points adjacent to the target time in the time dimension are selected from the spatially aligned meteorological and sea state data. and ; for the multi-source meteorological and sea state data corresponding to the two selected time grid points and Linear interpolation was performed to obtain multi-source meteorological and sea state data corresponding to the target time t. The calculation formula is as follows:

[0114]

[0115] in, These are time interpolation coefficients used to make meteorological data at discrete time levels continuous;

[0116] After the calculation of multi-source meteorological and sea state data for all target times is completed, the multi-source meteorological and sea state data can be matched to each data point in the real-time AIS data and predicted AIS trajectory data sequence of each target vessel to achieve time dimension alignment.

[0117] After the above spatial dimension reverse distance weighted interpolation and time dimension linear interpolation processing, meteorological and sea state elements such as wind speed, wave height, sea current speed, sea current direction, precipitation, tide and visibility can be accurately aligned with the ship's current moment and subsequent predicted trajectory points (AIS data) on a spatiotemporal scale. This ensures that the AIS data at each moment corresponds to a set of meteorological and sea state element parameters, thereby forming a spatiotemporal dataset that couples the ship's motion state with external environmental conditions. This provides a unified and continuous environmental constraint input for subsequent ship motion state analysis and bridge collision risk identification.

[0118] IV. Multi-dimensional Risk Quantification and Identification Steps: For each target vessel, a multi-dimensional index quantification scoring method is adopted to score risk indicators including channel departure risk, weather and sea state risk, and abnormal maneuvering risk, to achieve quantitative identification of ship-bridge collision risk. Specifically, this includes: scoring the risk of the vessel leaving the channel based on the relative position of each AIS data point in the real-time AIS data and the predicted AIS trajectory data sequence with the channel spatial area that matches the vessel type of the target vessel, combined with a preset safety boundary distance threshold; and scoring the risk of the vessel leaving the channel based on the real-time AIS data and the predicted AIS trajectory data sequence of the target vessel. The multi-source meteorological and sea state data corresponding to each AIS data point in the S-track data sequence are combined with the target vessel's vessel type and a preset vessel type-meteorological and sea state element level mapping table to determine the risk level of each meteorological and sea state element under each AIS data point. The risk levels are then fused using the most unfavorable element-dominated principle, and the fusion result is normalized to obtain the meteorological and sea state risk score. Based on the heading change rate and speed change rate between adjacent AIS data points in the AIS track data sequence to be analyzed, and combined with the heading change rate anomaly judgment threshold and speed change rate anomaly judgment threshold, the maneuvering anomaly risk is scored.

[0119] This invention provides risk identification for real-time operational status within a standardized monitoring area of ​​a bridge. It assesses vessels in transit from three dimensions: deviation from the waterway, weather and sea conditions, and abnormal maneuvering / operational stability, and outputs corresponding risk ratings.

[0120] A) Risk of deviating from the shipping lane

[0121] Specifically, the scoring of the risk of deviating from the course is based on the real-time AIS data P and the short-term predicted AIS trajectory data sequence. The relative position of each AIS data point (i.e., each trajectory point) to the channel space area (here, the channel space area is set as a polygon A) that matches the ship type to which the target ship belongs is determined.

[0122] As shown in the following formula:

[0123]

[0124]

[0125] in, This indicates that trajectory point P is located outside the airway space region. The trajectory point P is located within the waterway space area, k represents the kth trajectory point for determining relative positional relationship, N represents the total number of trajectory points for determining relative positional relationship, and n represents the total number of trajectory points of the target ship located outside the waterway space area.

[0126] when When this occurs, it indicates that at least one AIS data point exists outside the spatial area of ​​the airway (which can be represented as...). If a vessel's trajectory has deviated from the effective navigation channel area, it can be directly determined that the risk of deviating from the channel is high, and the risk score for deviating from the channel is assigned as 1 ( );

[0127] when When all AIS data (track points) are within the channel space area, it is necessary to determine the minimum distance between the track points of all AIS data in the real-time AIS data and the predicted AIS track data sequence and the boundary of the channel space area that is compatible with the type of vessel to which the target vessel belongs. In other words, the risk of edge contact is assessed based on the proximity of the track points to the channel boundary.

[0128] set up Let represent the set of boundary points of the channel space region, then the minimum distance from the k-th predicted trajectory point to the channel boundary is... It can be defined as:

[0129]

[0130] Where q represents any point in the set of boundary points of the waterway space region;

[0131] Furthermore, the minimum distance from the predicted AIS trajectory data sequence to the channel boundary can be defined. for:

[0132]

[0133] set up The risk score for a ship leaving the waterway is calculated based on the preset safety boundary distance threshold. It can be defined as:

[0134]

[0135] If the minimum distance ≥Preset safety boundary distance threshold If so, the risk score for the ship leaving the waterway will be 0;

[0136] If the minimum distance <Preset safety boundary distance threshold> Based on Calculate the risk score for ships leaving the waterway.

[0137] Therefore, The larger the value, the more likely the ship's predicted trajectory points (predicted AIS trajectory data sequence) are to deviate from the waterway space area, and the higher the risk of deviating from the waterway.

[0138] B) Weather and sea state risks

[0139] In this embodiment of the invention, for meteorological and sea state risk scoring, due to the significant differences in maneuverability and environmental resistance among different ship types, it is necessary to perform a graded mapping based on multi-source meteorological and sea state data corresponding to each AIS data point in the real-time AIS data and predicted AIS trajectory data sequence of the target ship, combined with the ship type of the target ship, in a preset ship type-meteorological and sea state element level mapping table. This yields the risk level of each meteorological and sea state element for each AIS data point (risk levels include normal, slightly unfavorable, significantly deteriorating, and extremely severe, with corresponding values ​​of 0, 1, 2, and 3, respectively). Then, the risk levels are fused using the most unfavorable element-dominant principle, and the fusion result is normalized to obtain the meteorological and sea state risk scoring result. The risk level of the s-th type of meteorological and sea state element corresponding to the k-th data point (trajectory point) is set as follows: The meteorological and sea state risk score result can then be expressed by the following formula. :

[0140]

[0141] Where N represents the total number of trajectory points involved in the calculation, and S represents the total number of meteorological and sea state elements for parameter calculation;

[0142] Therefore, The larger the value, the worse the current and short-term future environmental conditions, and the higher the risk of navigation for ships affected by weather and sea conditions.

[0143] C) Risk of Manipulation Abnormalities

[0144] For the maneuvering anomaly risk scoring, firstly, the rate of change of heading between adjacent AIS data (two adjacent trajectory points) is calculated based on the absolute value of the ratio of the heading difference to the timestamp difference in the AIS trajectory data sequence to be analyzed. The calculation formula is as follows:

[0145]

[0146] in, and This indicates the headings corresponding to two adjacent trajectory points. and The timestamps represent the timestamps corresponding to two adjacent trajectory points, i represents the i-th trajectory point in the AIS trajectory data sequence to be analyzed, and M represents the total number of trajectory points in the AIS trajectory data sequence to be analyzed.

[0147] Secondly, the rate of change of airspeed between adjacent AIS data points is calculated based on the absolute value of the ratio of the airspeed difference to the timestamp difference between adjacent AIS data points. The calculation formula is as follows:

[0148]

[0149] in, and This represents the speed corresponding to two adjacent trajectory points;

[0150] Next, based on the rate of change of heading between adjacent data points in the AIS trajectory data sequence to be analyzed, the average rate of change of heading for the AIS trajectory data sequence to be analyzed is calculated. ;

[0151] Then, based on the rate of change of heading between adjacent AIS data in the AIS trajectory data sequence to be analyzed... With the average rate of change of heading The degree of fluctuation in the heading adjustment behavior of the AIS trajectory data sequence to be analyzed was calculated. The calculation formula is as follows:

[0152]

[0153] Next, based on the rate of change of airspeed between adjacent AIS data points in the AIS trajectory data sequence to be analyzed... Determine the maximum rate of change of speed This is used to reflect the degree of drastic change in ship speed, and its expression is as follows:

[0154]

[0155] Finally, based on the aforementioned degree of fluctuation and the maximum rate of change of speed Thresholds are determined by combining the abnormal rate of change of heading. Threshold for judging abnormal speed change rate The manipulation anomaly risk score was calculated. The calculation formula is as follows:

[0156]

[0157] in, and These are weighting coefficients, and they satisfy the formula... .

[0158] Therefore, Furthermore, the larger the value, the more unstable the ship's maneuvering behavior is within a short time window, and the higher the risk of maneuvering anomalies.

[0159] The multi-dimensional risk quantification and identification step in this invention ultimately outputs three positive risk indicators that can be directly used in the comprehensive assessment: risk of deviating from the navigation channel, risk of meteorological and sea state conditions, and risk of abnormal maneuvering. (All three indicators are positive risk indicators; that is, the larger the indicator value, the higher the risk of ship-bridge collision.) These three indicators constitute a risk indicator vector.

[0160]

[0161] in, This indicates the result of the risk assessment for leaving the shipping lane. This indicates the meteorological and sea condition risk score result. This indicates the result of the manipulation anomaly risk score.

[0162] V. Ship-Bridge Collision Risk Assessment and Early Warning Steps: For each ship type, the TOPSIS entropy weight analysis method is used to weight and fuse the exit risk score, meteorological and sea state risk score, and maneuver anomaly risk score of each target ship in the aforementioned ship type to obtain the assessment result of the comprehensive ship-bridge collision risk assessment index for each target ship; based on the assessment result of the comprehensive ship-bridge collision risk assessment index corresponding to each target ship, a risk warning is issued for the target ship.

[0163] In this embodiment of the invention, before weighting and fusing the risk index scores of each target vessel in the vessel type using the TOPSIS entropy weight analysis method, the range standardization method is first used to perform dimensionless processing on the risk index scores of each target vessel in the vessel type.

[0164] Assuming that within the same assessment period, there are T samples of a certain type of vessel to be assessed, and these samples may correspond to different target vessels (in another embodiment, the samples may correspond to the same target vessel at different times), then the risk indicators output by the multi-dimensional risk quantification and identification steps can form an indicator matrix. Where i represents the i-th sample and j represents the j-th risk indicator. This represents the off-course risk score result for the i-th sample. This represents the meteorological and sea state risk score result for the i-th sample. This represents the manipulation anomaly risk score result for the i-th sample.

[0165] Since the above three risk indicators have all been standardized as positive risk indicators where "the larger the value, the higher the risk," the following range standardization method can be used directly for dimensionless processing to obtain the normalized result:

[0166]

[0167] in, This represents the normalized result of the j-th risk indicator (positive risk indicator) for the i-th sample;

[0168] Based on this, the risk index scores are calculated using the dimensionless processing of the data for each target vessel within the aforementioned vessel type. The TOPSIS entropy weight analysis method is used to determine the index weight coefficients of each risk indicator for each target ship in the aforementioned ship type. Specifically, the weight of the j-th risk indicator in the i-th sample is first calculated. The formula is shown below:

[0169]

[0170] Then, use the following formula to calculate the information entropy of the j-th risk indicator. and coefficient of difference :

[0171]

[0172]

[0173] Thus, the index weight coefficient (also known as entropy weight) of the j-th risk indicator. It can be represented as:

[0174]

[0175] Then, for each of the aforementioned ship types, based on the dimensionless risk index scores and their corresponding index weight coefficients for each target ship within that ship type, the standardized risk values ​​for each risk index corresponding to each target ship within that ship type are calculated. Finally, a weighted normalized decision matrix for that ship type is constructed based on all the standardized risk values ​​for that ship type. :

[0176]

[0177] Next, for each of the aforementioned ship types, the maximum value of the standardized risk value of each risk indicator in the weighted normalized decision matrix corresponding to that ship type is determined (where the maximum value of the standardized risk value of the j-th risk indicator can be expressed as...). Construct the positive ideal solution for the ship type. Based on the minimum standardized risk value of each risk indicator in the weighted normalized decision matrix corresponding to the ship type (where the minimum standardized risk value of the j-th risk indicator can be expressed as...), Construct the negative ideal solution for the ship type. .

[0178] Therefore, the correct solution is... Corresponding to the ideal high-risk state of "highest risk of deviating from the channel, highest risk of weather and sea conditions, and highest risk of abnormal maneuvering"; negative ideal solution This corresponds to the ideal low-risk state, which includes "the lowest risk of deviating from the course, the lowest risk of weather and sea conditions, and the lowest risk of abnormal maneuvering." In other words, if a sample simultaneously possesses high... higher and higher If the scores of all three risk indicators are low, then the state is closer to the positive ideal solution; conversely, if the scores of all three risk indicators are low, then the state is closer to the negative ideal solution.

[0179] Based on this, the Euclidean distance between the i-th sample and the positive ideal solution and the negative ideal solution, as well as the comprehensive risk assessment index for ship-bridge collisions, are calculated respectively:

[0180]

[0181]

[0182] in, Let represent the Euclidean distance between the i-th sample and the positive ideal solution. Let represent the Euclidean distance between the i-th sample and the negative ideal solution, j represent the j-th risk indicator, and m represent the total number of risk indicators. This represents the comprehensive risk assessment index for ship-bridge collisions in the i-th sample. The larger the value, the closer the i-th sample is to the high-risk ideal state, and the higher the risk of ship-bridge collision; conversely, the smaller the value, the closer the sample is to the low-risk ideal state, and the lower the risk of ship-bridge collision.

[0183] Finally, based on the assessment results of the ship-bridge collision comprehensive risk assessment indicators corresponding to each of the target vessels, Based on a preset level classification threshold, each target vessel is classified into a ship-bridge collision risk level (including normal, caution, danger, etc.), and a risk warning is issued for target vessels with a ship-bridge collision risk level of danger.

[0184] The embodiments of the present invention, through the comprehensive evaluation process of the above-mentioned TOPSIS entropy weight analysis method, can uniformly map multi-dimensional and heterogeneous risk identification results into a single comprehensive early warning indicator, which is used to sort, classify or determine the triggering early warning threshold for ship-bridge collision risks, thereby providing an intuitive and quantifiable comprehensive risk output for bridge collision avoidance monitoring and decision support.

[0185] Embodiments of the present invention can display ship-bridge collision risk warning information on a visualized bridge risk monitoring screen, such as... Figure 3 The image shown is an example of a vessel-bridge collision risk warning information for the Sutong Yangtze River Bridge. The left side displays current vessel information near the bridge, including basic vessel information such as name, type, MMSI, update time, current location, speed / draft, bow / direction, length / beam, and corresponding risk level. The upper right side displays a map of the area near the bridge and vessel navigation distribution information. The lower right side displays the statistical quantity of various vessel types on the map using bar charts. This visualization method can present multi-dimensional and multi-type monitoring data in a structured and intuitive way, enabling bridge collision avoidance monitoring and navigation safety management personnel to quickly grasp the distribution, quantity, and trends of vessel data, effectively improving data readability and emergency decision-making efficiency.

[0186] Based on the same inventive concept, one or more embodiments of this specification also provide a data-driven ship-bridge collision risk warning system. Since the working principle of the data-driven ship-bridge collision risk warning system is the same as that of the aforementioned data-driven ship-bridge collision risk warning method, the implementation of the data-driven ship-bridge collision risk warning system can refer to the aforementioned implementation of the data-driven ship-bridge collision risk warning method, and the repeated parts will not be described again.

[0187] Figure 4 This specification provides a structural block diagram of a data-driven ship-bridge collision risk early warning system for one or more embodiments. Figure 4 As shown, the system includes, in sequence, a waterway spatial region generation module 101, a ship future trajectory prediction module 102, a meteorological and sea state data acquisition and spatiotemporal alignment module 103, a multi-dimensional risk quantification and identification module 104, and a ship-bridge collision risk assessment and early warning module 105. Among them,

[0188] The waterway spatial region generation module 101 is used to acquire the structural and spatial data of the target bridge, including bridge centerline information, bridge structural height parameters, and bridge monitoring area. The module performs coordinate system normalization on the structural and spatial data to construct a standardized monitoring area of ​​the target bridge under a unified geographic reference system. It acquires historical vessel crossing records corresponding to the target bridge and forms a bridge crossing event set. It associates the vessel types corresponding to each bridge crossing event and filters the bridge crossing event set according to preset vessel type conditions to obtain a valid crossing event set. Based on the standardized monitoring area, it extracts historical AIS trajectory data sequences corresponding to the valid crossing event set, summarizes them to form a bridge area trajectory point set, and performs preprocessing operations. It uses a hierarchical density clustering algorithm to spatially cluster the preprocessed bridge area trajectory point set to obtain several waterway clusters, and performs surface processing on each waterway cluster to generate the corresponding waterway spatial region.

[0189] The ship future trajectory prediction module 102 is used to determine all target ships in the navigation state within the standardized monitoring area and their corresponding ship types, and to acquire the real-time AIS data of each target ship at the current moment and the historical AIS trajectory data sequence within a preset time window traced back from the current moment to form the AIS trajectory data sequence to be analyzed; based on the real-time AIS data corresponding to each target ship, combined with the preset single prediction time step, the module uses a constant speed and constant direction trajectory extrapolation method to calculate the east-west and north-south displacement components of each target ship within each single prediction time step, and repeats the calculation for the preset number of prediction steps to obtain the predicted AIS trajectory data sequence of each target ship;

[0190] The meteorological and sea state data acquisition and spatiotemporal alignment module 103 is used to acquire multi-source meteorological and sea state data within the entire time range covered by the predicted AIS trajectory data sequence from the current moment to the spatial range corresponding to the standardized monitoring area; and through spatiotemporal alignment processing, the multi-source meteorological and sea state data is matched with the real-time AIS data of each target vessel and each AIS data in the predicted AIS trajectory data sequence.

[0191] The multi-dimensional risk quantification and identification module 104, for each target vessel, employs a multi-dimensional index quantification and scoring method to score risk indicators including channel departure risk, weather and sea state risk, and abnormal maneuvering risk, thereby achieving quantitative identification of ship-bridge collision risk. Specifically, this includes: scoring the channel departure risk based on the relative position of each AIS data point in the real-time AIS data and the predicted AIS trajectory data sequence with the channel spatial area corresponding to the vessel type of the target vessel, combined with a preset safety boundary distance threshold; and scoring the channel departure risk based on the real-time AIS data and the predicted AIS trajectory data sequence of the target vessel. The multi-source meteorological and sea state data corresponding to each AIS data point in the AIS trajectory data sequence are combined with the target vessel's vessel type and a preset vessel type-meteorological and sea state element level mapping table to determine the risk level of each meteorological and sea state element under each AIS data point. The risk levels are then fused using the most unfavorable element-dominated principle, and the fusion result is normalized to obtain the meteorological and sea state risk score. Based on the rate of change of heading and the rate of change of speed between adjacent AIS data points in the AIS trajectory data sequence to be analyzed, and combined with the anomaly judgment thresholds for the rate of change of heading and the rate of change of speed, the risk of maneuvering anomalies is scored.

[0192] The ship-bridge collision risk assessment and early warning module 105, for each ship type, uses the TOPSIS entropy weight analysis method to weight and fuse the exit risk score, meteorological and sea state risk score, and maneuver anomaly risk score of each target ship in that ship type to obtain the assessment result of the comprehensive ship-bridge collision risk assessment index for each target ship; based on the assessment result of the comprehensive ship-bridge collision risk assessment index corresponding to each target ship, it provides a risk warning for the target ship.

[0193] This invention discloses a data-driven method and system for early warning of ship-bridge collision risks. Within a standardized bridge monitoring area, it integrates bridge structure and waterway boundary constraints, AIS real-time trajectory, meteorological elements, and ship maneuvering characteristics to construct a feature expression and analysis framework suitable for real-time early warning. This framework enables multi-dimensional risk identification and assessment of ship navigation status. Furthermore, it employs a multi-index fusion algorithm to integrate and calculate multiple sub-risk indicators, outputting a unique comprehensive ship-bridge collision risk quantification index for early warning determination. Ultimately, this achieves interpretable, comparable, and traceable early warning of ship-bridge collision risks.

[0194] It should be noted that the specific embodiments described above enable those skilled in the art to more fully understand the present invention, but do not limit the present invention in any way. Therefore, although the present invention has been described in detail with reference to the accompanying drawings and embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the present invention. In short, all technical solutions and improvements that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the present invention patent.

Claims

1. A data-driven method for early warning of ship-bridge collision risks, characterized in that, Includes the following steps: The steps for generating a waterway spatial region are as follows: First, acquire the structural and spatial data of the target bridge, including bridge centerline information, bridge structural height parameters, and the bridge monitoring area. Second, perform coordinate system normalization on the structural and spatial data to construct a standardized monitoring area for the target bridge under a unified geographic reference system. Third, acquire historical vessel crossing records corresponding to the target bridge and form a bridge crossing event set. Fourth, associate the vessel types corresponding to each bridge crossing event and filter the bridge crossing event set according to preset vessel type conditions to obtain a valid crossing event set. Based on the standardized monitoring area, extract the historical AIS trajectory data sequence corresponding to the set of valid crossing events, summarize it to form a bridge area trajectory point set and perform preprocessing operations; Hierarchical density clustering algorithm is used to spatially cluster the preprocessed bridge area trajectory point set to obtain several waterway clusters. Each waterway cluster is then surface-processed to generate the corresponding waterway spatial region. The steps for predicting future ship trajectories are as follows: First, identify all target ships currently in operation within the standardized monitoring area and their corresponding ship types. Then, acquire the real-time AIS data of each target ship at the current moment, as well as a sequence of historical AIS trajectory data within a preset time window, forming an AIS trajectory data sequence to be analyzed. Based on the real-time AIS data corresponding to each target ship, and combined with a preset single prediction time step, use a constant speed and direction trajectory extrapolation method to sequentially calculate the east-west and north-south displacement components of each target ship within each single prediction time step. After repeating the preset prediction steps, the predicted AIS trajectory data sequence for each target ship is obtained. Meteorological and sea state data acquisition and spatiotemporal alignment steps: acquire multi-source meteorological and sea state data within the entire time range from the current moment to the predicted AIS trajectory data sequence and within the spatial range corresponding to the standardized monitoring area; through spatiotemporal alignment processing, match the multi-source meteorological and sea state data with the real-time AIS data of each target vessel and each AIS data in the predicted AIS trajectory data sequence; Multi-dimensional risk quantification and identification steps: For each target vessel, a multi-dimensional index quantification and scoring method is used to score risk indicators including channel departure risk, weather and sea state risk, and abnormal maneuvering risk, to achieve quantitative identification of ship-bridge collision risk. Specifically, this includes: scoring the risk of the vessel leaving the channel based on the relative position of each AIS data point in the real-time AIS data and the predicted AIS trajectory data sequence with the channel spatial area that matches the vessel type of the target vessel, combined with a preset safety boundary distance threshold; and scoring the risk of the vessel leaving the channel based on the real-time AIS data and the predicted AIS data of the target vessel. The multi-source meteorological and sea state data corresponding to each AIS data point in the trajectory data sequence, combined with the target vessel's vessel type and a preset vessel type-meteorological and sea state element level mapping table, determine the risk level of each meteorological and sea state element under each AIS data point. The risk levels are then fused using the most unfavorable element-dominated principle, and the fusion result is normalized to obtain the meteorological and sea state risk score. Based on the heading change rate and speed change rate between adjacent AIS data points in the AIS trajectory data sequence to be analyzed, and combined with the heading change rate anomaly judgment threshold and speed change rate anomaly judgment threshold, the maneuvering anomaly risk is scored. Ship-bridge collision risk assessment and early warning steps: For each ship type, the TOPSIS entropy weight analysis method is used to weight and fuse the exit risk score, meteorological and sea state risk score, and maneuver anomaly risk score of each target ship in the aforementioned ship type to obtain the assessment result of the comprehensive ship-bridge collision risk assessment index for each target ship; based on the assessment result of the comprehensive ship-bridge collision risk assessment index corresponding to each target ship, a risk warning is issued for the target ship.

2. The method according to claim 1, characterized in that, In the waterway spatial region generation step, the steps of obtaining historical vessel crossing records corresponding to the target bridge and forming a bridge crossing event set, filtering the bridge crossing event set to obtain a valid crossing event set, and extracting the historical AIS trajectory data sequence corresponding to the valid crossing event set, and summarizing them to form a bridge area trajectory point set, specifically include: The database is used to retrieve all vessels that have passed through the target bridge within the monitoring area during the historical period, and the vessel identification and passage information of these vessels are obtained. The passage information includes the time and direction of passage of the vessels through the target bridge. For each of the aforementioned passage information, a time window for extracting historical AIS trajectory data sequences of a preset fixed duration is constructed with the crossing time as the center; taking the passage of a ship over the target bridge as a crossing event, the bridge crossing event set of the target bridge is constructed by combining the ship identification, passage information and its corresponding historical AIS trajectory data sequence extraction time window; For each crossing event in the bridge crossing event set, the static information of the vessel associated with the crossing event is obtained through the vessel identifier corresponding to the crossing event. The static information of the vessel includes the vessel type. The crossing events are filtered according to preset ship type conditions, and crossing events that do not meet the ship type requirements are eliminated to obtain a set of valid crossing events. Based on the ship identifiers and historical AIS trajectory data sequences corresponding to each crossing event in the set of valid crossing events, time windows are extracted to obtain the historical AIS trajectory data sequences corresponding to each crossing event; the data volume of each obtained historical AIS trajectory data sequence is checked, and historical AIS trajectory data sequences that do not meet the preset data volume requirements are removed; the checked historical AIS trajectory data sequences are summarized to form a bridge area trajectory point set.

3. The method according to claim 1, characterized in that, In the ship future trajectory prediction step, the real-time AIS data includes longitude, latitude, timestamp, ground speed, and heading; Using the real-time AIS data corresponding to each target vessel as the current point, the latitude and longitude of the current point are converted into projected coordinates, and the heading of the current point is converted into radians; For each of the target vessels, based on the ground speed, heading and preset single prediction time step at the current point, the east-west displacement component and north-south displacement component of the current point are calculated using a constant speed and heading trajectory extrapolation method. Add the projected coordinates of the current point to the corresponding north-south displacement component and east-west displacement component respectively to obtain the projected coordinates of the next predicted AIS data; Using the projected coordinates of the next predicted AIS data as the new current point, repeat the above steps until the preset number of prediction steps is reached. Finally, convert the projected coordinates of all predicted AIS data into geographic coordinates to generate a sequence of predicted AIS trajectory data for each target vessel.

4. The method according to claim 1, characterized in that, In the step of acquiring and aligning meteorological and marine state data in time and space, the multi-source meteorological and marine state data includes wind speed, wave height, sea current speed, sea current direction, precipitation, tides, and visibility. For each target vessel, the geographical location of each AIS data point in the real-time AIS data and predicted AIS trajectory data sequence of the target vessel is taken as the target location. Four spatial grid points closest to the target location are selected in the multi-source meteorological and sea state data. Based on the selected four spatial grid points, the target location is obtained by calculating the corresponding spatial interpolated meteorological and sea state data through inverse distance weighted interpolation. Then, the multi-source meteorological and sea state data is matched to each AIS data point in the real-time AIS data and predicted AIS trajectory data sequence of each target vessel to achieve spatial dimension alignment. For each of the target vessels, the timestamp of each AIS data in the real-time AIS data and predicted AIS trajectory data sequence of the target vessel is taken as the target time. Two time grid points that are adjacent to the target time in the time dimension are selected in the spatially aligned meteorological and sea state data. Linear interpolation is performed on the multi-source meteorological and sea state data corresponding to the two selected time grid points to obtain the multi-source meteorological and sea state data corresponding to the target time. Then, the multi-source meteorological and sea state data is matched to each AIS data in the real-time AIS data and predicted AIS trajectory data sequence of each target vessel to achieve time dimension alignment.

5. The method according to claim 1, characterized in that, The multidimensional risk quantification and identification step specifically includes the following steps for scoring the risk of a ship leaving the waterway: Based on the relative position of each AIS data point in the real-time AIS data and the predicted AIS trajectory data sequence with the channel space area that matches the ship type of the target ship, determine the relative positional relationship between each AIS data point and the channel space area that matches the ship type of the target ship. If at least one AIS data point exists outside the waterway spatial area, the risk score for the vessel leaving the waterway is 1. If all AIS data are within the waterway space area, then determine the minimum distance between the trajectory points of all AIS data in the real-time AIS data and the predicted AIS trajectory data sequence and the boundary of the waterway space area that matches the ship type of the target ship. If the minimum distance is greater than or equal to the preset safety boundary distance threshold, then the risk score for the ship leaving the waterway will be 0. If the minimum distance is less than the preset safety boundary distance threshold, then the risk score of the ship leaving the waterway is calculated based on the ratio of the minimum distance to the preset safety boundary distance threshold.

6. The method according to claim 1 or 5, characterized in that, In the multidimensional risk quantification and identification step, when scoring the risk of abnormal manipulation, the heading change rate of the adjacent AIS data is calculated based on the absolute value of the ratio of the heading difference to the timestamp difference in the adjacent AIS trajectory data sequence to be analyzed; the speed change rate of the adjacent AIS data is calculated based on the absolute value of the ratio of the speed difference to the timestamp difference in the adjacent AIS data.

7. The method according to claim 6, characterized in that, In the multidimensional risk quantification and identification step, when scoring the risk of abnormal manipulation, the average rate of change of heading of the AIS trajectory data sequence to be analyzed is calculated by using the rate of change of heading between each adjacent AIS data in the AIS trajectory data sequence to be analyzed; based on the rate of change of heading between each adjacent AIS data in the AIS trajectory data sequence to be analyzed and the average rate of change of heading, the degree of fluctuation of the heading adjustment behavior of the AIS trajectory data sequence to be analyzed is calculated. Based on the rate of change of airspeed between adjacent AIS data in the AIS trajectory data sequence to be analyzed, the maximum rate of change of airspeed is determined. Based on the degree of fluctuation and the maximum rate of change of speed, combined with the abnormal judgment thresholds for the rate of change of heading and the rate of change of speed, the abnormal control risk score is calculated.

8. The method according to claim 1, characterized in that, In the ship-bridge collision risk assessment and early warning step, before weighting and integrating the risk index scores of each target ship in the ship type using the TOPSIS entropy weight analysis method, the range standardization method is used to perform dimensionless processing on the risk index scores of each target ship in the ship type. For each of the aforementioned ship types, based on the dimensionless processing of the risk index scores corresponding to each target ship in the aforementioned ship type, the TOPSIS entropy weight analysis method is used to determine the index weight coefficients of each risk index corresponding to each target ship in the aforementioned ship type. For each of the aforementioned ship types, based on the dimensionless processing of the risk index scores and their corresponding index weight coefficients for each target ship in the aforementioned ship type, the standardized risk values ​​of each risk index for each target ship in the aforementioned ship type are calculated, and a weighted normalized decision matrix for the aforementioned ship type is constructed based on all the standardized risk values ​​under the aforementioned ship type. For each of the aforementioned ship types, a positive ideal solution is constructed based on the maximum value of the standardized risk value of each risk indicator in the weighted normalized decision matrix corresponding to the ship type; and a negative ideal solution is constructed based on the minimum value of the standardized risk value of each risk indicator in the weighted normalized decision matrix corresponding to the ship type. For each target vessel in each of the aforementioned vessel types, the evaluation result of the ship-bridge collision comprehensive risk assessment index of the target vessel is calculated based on the standardized risk value, positive ideal solution, and negative ideal solution of each risk indicator corresponding to the target vessel.

9. The method according to claim 1 or 8, characterized in that, In the ship-bridge collision risk assessment and early warning step, based on the assessment results of the ship-bridge collision comprehensive risk assessment index corresponding to each target ship, and combined with the preset level classification threshold, the ship-bridge collision risk level of each target ship is classified, including normal, caution, and danger. Risk warnings are issued for target vessels whose ship-bridge collision risk level is deemed dangerous.

10. A data-driven ship-bridge collision risk early warning system, characterized in that, The module comprises, in sequence, a channel spatial region generation module, a ship future trajectory prediction module, a meteorological and sea state data acquisition and spatiotemporal alignment module, a multi-dimensional risk quantification and identification module, and a ship-bridge collision risk assessment and early warning module. The waterway spatial region generation module is used to: acquire the structural and spatial data of the target bridge, including bridge centerline information, bridge structural height parameters, and bridge monitoring area; perform coordinate system normalization on the structural and spatial data to construct a standardized monitoring area of ​​the target bridge under a unified geographic reference system; acquire historical vessel crossing records corresponding to the target bridge and form a bridge crossing event set; associate the vessel types corresponding to each bridge crossing event, and filter the bridge crossing event set according to preset vessel type conditions to obtain a valid crossing event set; extract historical AIS trajectory data sequences corresponding to the valid crossing event set based on the standardized monitoring area, summarize them to form a bridge area trajectory point set and perform preprocessing operations; use a hierarchical density clustering algorithm to spatially cluster the preprocessed bridge area trajectory point set to obtain several waterway clusters, and perform surface processing on each waterway cluster to generate the corresponding waterway spatial region; The ship future trajectory prediction module is used to: determine all target ships in operation within the standardized monitoring area and their corresponding ship types; acquire the real-time AIS data of each target ship at the current moment and extract the historical AIS trajectory data sequence within a preset time window from the current moment backward to form the AIS trajectory data sequence to be analyzed; based on the real-time AIS data corresponding to each target ship, combined with the preset single prediction time step, and using the constant speed and constant direction trajectory extrapolation method, calculate the east-west and north-south displacement components of each target ship within each single prediction time step in turn, and repeat the calculation for the preset number of prediction steps to obtain the predicted AIS trajectory data sequence of each target ship; The meteorological and sea state data acquisition and spatiotemporal alignment module is used to acquire multi-source meteorological and sea state data within the entire time range covered by the predicted AIS trajectory data sequence from the current moment to the spatial range corresponding to the standardized monitoring area; and through spatiotemporal alignment processing, match the multi-source meteorological and sea state data with the real-time AIS data of each target vessel and each AIS data in the predicted AIS trajectory data sequence. The multi-dimensional risk quantification and identification module: For each target vessel, a multi-dimensional index quantification and scoring method is used to score risk indicators including channel departure risk, weather and sea state risk, and abnormal maneuvering risk, to achieve quantitative identification of ship-bridge collision risk. Specifically, it includes: scoring the risk of the vessel leaving the channel based on the relative position of each AIS data point in the real-time AIS data and the predicted AIS trajectory data sequence with the channel spatial area that matches the vessel type of the target vessel, combined with a preset safety boundary distance threshold; and scoring the risk of the vessel leaving the channel based on the real-time AIS data and the predicted AIS trajectory data of the target vessel. The multi-source meteorological and sea state data corresponding to each AIS data point in the S-track data sequence are combined with the target vessel's vessel type and a preset vessel type-meteorological and sea state element level mapping table to determine the risk level of each meteorological and sea state element under each AIS data point. The risk levels are then fused using the most unfavorable element-dominated principle, and the fusion result is normalized to obtain the meteorological and sea state risk score. Based on the heading change rate and speed change rate between adjacent AIS data points in the AIS track data sequence to be analyzed, and combined with the heading change rate anomaly judgment threshold and the speed change rate anomaly judgment threshold, the maneuvering anomaly risk is scored. The ship-bridge collision risk assessment and early warning module: For each ship type, the TOPSIS entropy weight analysis method is used to weight and fuse the exit risk score, meteorological and sea state risk score, and maneuver anomaly risk score of each target ship in the ship type to obtain the assessment result of the comprehensive ship-bridge collision risk assessment index of each target ship; based on the assessment result of the comprehensive ship-bridge collision risk assessment index corresponding to each target ship, a risk warning is issued for the target ship.

Images