A method and system for identifying hot water areas of waterway water accidents and handling accidents

Abstract

This invention discloses a method and system for identifying and handling accident hotspots in canals, belonging to the field of accident hotspot identification technology. Addressing the problems of existing technologies failing to capture the dynamic migration of accident hotspots under gate control-tidal coupling, and lacking comprehensive risk indicators and attribution explanations, this invention extracts cross-current risk index, ship encounter thermal values, and lock drive characteristics. Based on lock status and tidal phase, it dynamically allocates weights and performs non-linear fusion to construct a risk-weighted spatiotemporal cube. It uses Poisson distribution spatiotemporal scanning statistics to identify hotspot clusters, employs game theory attribution algorithms to analyze dominant risk types, and generates attribution results labeled with gate control-tidal trigger conditions. Through location-risk coupling determination, it adaptively adjusts the iteration interval, generating differentiated patrol, emergency, and dispatch plans. This invention improves the accuracy of hotspot identification, attribution transparency, and the overall intelligence and resource allocation efficiency of the entire chain.

Description

Technical Field

[0001] This invention relates to the field of accident hotspot water area identification technology, and in particular to a method and system for identifying accident hotspot water areas and handling accidents on canals. Background Technology

[0002] With the continuous enhancement of inland waterway transportation's position in the national comprehensive transportation system, the navigation safety of canals, as key channels connecting important waterways and undertaking the transportation of bulk goods, is receiving increasing attention. In recent years, driven by the development strategy of smart shipping, utilizing multi-source information such as Automatic Identification System (AIS) data, hydrological monitoring data, and lock scheduling data to achieve waterway safety situational awareness and risk early warning has become an important development direction in the field of water traffic management.

[0003] However, existing technologies have the following shortcomings: First, they treat accident locations as fixed black spots, failing to consider the migration effects of dynamic factors such as lock scheduling and tidal phase on the spatiotemporal distribution of risks, making it difficult to capture the dynamic changes in accident hotspots with lock opening and closing and tidal fluctuations. Second, in constructing risk assessment indicators, they typically use single-dimensional traffic flow characteristics or historical accident frequencies, lacking a comprehensive quantification and fusion expression of multi-dimensional risk factors such as cross-current risks and ship encounter conflicts, making it difficult to characterize the coupling mechanism of risk formation under the dual-drive of lock control and tides. Third, after identifying high-risk areas, they lack the ability to automatically analyze the causes of risks, failing to distinguish whether the same hotspot is dominated by cross-currents or by dense ship encounters at different times, resulting in the inability to implement differentiated and precise responses in subsequent patrol scheduling and emergency deployment. Fourth, existing methods mostly use fixed-period risk updates, lacking a mechanism to adaptively adjust the monitoring frequency according to changes in the risk situation, which may lead to early warning lags during periods of rapid risk evolution, while wasting computational resources during periods of stable risk.

[0004] Therefore, this invention is proposed. Summary of the Invention

[0005] To address these issues, this invention provides a method and system for identifying and handling accident hotspots in canals, overcoming the problems of existing static identification methods that cannot capture the spatiotemporal dynamic migration patterns of accident hotspots under gate control-tidal coupling, have single risk assessment indicators that lack multi-factor coupling and fusion expression, cannot automatically analyze the causes of hotspot risks to support differentiated and accurate responses, and have fixed risk update frequencies that cannot adaptively adjust the monitoring rhythm.

[0006] To achieve the above objectives, the present invention provides a method for identifying and handling accident hotspots in canals, comprising: S1: Acquire ship navigation data, lock scheduling data, and tidal hydrological data. After data cleaning and spatiotemporal alignment, construct a unified spatiotemporal gridded waterway model. S2, based on a spatiotemporal gridded waterway model, extracts tidal driving features, ship traffic conflict features, and lock driving features; S3, determine the crossflow risk index based on the tidal direction and channel position and the lateral velocity component in the tidal driving characteristics; determine the ship encounter thermal value based on the ship traffic conflict characteristics. S4. Based on the current lock operation status and tidal phase, the crossflow risk index, the ship encounter thermal value, and the lock drive characteristics are dynamically assigned fusion weights. After nonlinear weighted fusion, the comprehensive accident risk index of each spatiotemporal grid unit is calculated, and a risk-weighted spatiotemporal cube is constructed. S5, on the risk-weighted spatiotemporal cube, with the comprehensive accident risk index as input, adopts a spatiotemporal scanning statistical method based on the Poisson distribution model, moves a scanning window of variable size, and identifies hotspot clusters where the comprehensive accident risk index gathers; S6, reverse analyze the contribution of the crossflow risk index and the ship encounter thermal value of the hot spot cluster location, calculate the corresponding gate control state and tidal phase, and generate the hot spot cluster attribution result based on the contribution, gate control state and tidal phase; S7 iterates through S3 to S6 to update the hotspot clusters and their corresponding attribution results for the target time period, and adjusts the time interval of the iteration based on the location of the hotspot clusters and their corresponding attribution results; and generates differentiated incident pre-handling solutions based on the updated hotspot clusters and their corresponding attribution results.

[0007] Furthermore, the acquisition of ship navigation data, lock scheduling data, and tidal hydrological data, after data cleaning and spatiotemporal alignment, to construct a unified spatiotemporal gridded waterway model includes: S11, acquire Automatic Identification System (AIS) data as ship navigation data, acquire lock scheduling record data as lock scheduling data, and acquire measured tide data and tide forecast data as tidal hydrological data. S12, clean the AIS data, remove position jumps, velocity abrupt changes and drift data, and interpolate the missing segments of the cleaned trajectory to obtain a continuous and smooth ship navigation trajectory. S13, convert the lock number, opening and closing time and release direction in the lock scheduling record data into a first continuous time series, and convert the tide level, flow velocity and flow direction in the actual tide data and tide forecast data into a second continuous time series. S14, align the continuous and smooth ship navigation trajectory, the first continuous time series, and the second continuous time series to the same time axis; S15, the canal waterway is divided into geographic grid units with preset side lengths, and a waterway attribute label is attached to each geographic grid unit. The waterway attribute label includes at least one of the following: entrance area, approach channel, bridge area, bend, and waiting anchorage. The unified spatiotemporal gridded waterway model is then constructed.

[0008] Furthermore, the determination of the crossflow risk index based on the tidal direction and channel position, as well as the lateral velocity component, in the tidal driving characteristics includes: S31, obtain the angle between the tidal flow direction and the channel centerline and the lateral velocity component of the tide in each spatiotemporal gridded channel model; S32, the product of the sine of the included angle and the transverse velocity component is determined as the basic value of the transverse flow risk; S33, obtain the channel attribute label of the spatiotemporal grid unit, determine the terrain sensitivity correction coefficient based on the channel attribute label, and correct the crossflow risk base value based on the terrain sensitivity correction coefficient to generate the crossflow risk index.

[0009] Furthermore, determining the encounter thermal value of a ship based on the characteristics of ship traffic conflicts includes: S34 uses a density-based clustering algorithm to identify overtaking, crossing, and encounter situations where ship trajectories are clustered in time and space in the characteristics of ship traffic conflicts. S35, extract the ship encounter density and minimum encounter distance in the areas where each encounter situation occurs; S36. Using the ship encounter density and minimum encounter distance as inputs, a weighted fusion algorithm is used to calculate the ship encounter thermal value; wherein, the ship encounter thermal value is positively correlated with the ship encounter density and negatively correlated with the minimum encounter distance.

[0010] Furthermore, the dynamic allocation of fusion weights to the crossflow risk index, ship encounter thermal value, and lock drive characteristics based on the current lock operating status and tidal phase includes: An adaptive weight regulator is constructed and trained to learn the nonlinear mapping relationship from the combined working conditions of the lock operation state and tidal phase to the optimal weight allocation. During high tide, the output weights of the crossflow risk index and the thermal value encountered by the ship are increased to highlight the collision risk. During the lock discharge period, the output weights of the lock drive characteristics are increased to highlight the bottoming risk.

[0011] During each feature fusion, the current lock operation status and tidal phase are input into the trained adaptive weight regulator to output the crossflow risk index, the ship encounter thermal value, and the fusion weights corresponding to the lock drive features.

[0012] Furthermore, the contribution of the cross-current risk index and the ship encounter thermal value at the location of the hotspot cluster is analyzed in reverse, the corresponding gate control state and tidal phase are calculated, and the hotspot cluster attribution result is generated based on the contribution, gate control state and tidal phase, including: S61, extract the crossflow risk index, ship encounter thermal value, lock drive characteristics and comprehensive accident risk index of each spatiotemporal grid unit within the spatiotemporal window where the hot spot cluster is located; S62 uses a game theory-based attribution decomposition algorithm, taking the cross-current risk index, the ship encounter thermal value, and the lock drive characteristics as participants, calculating the marginal contribution of each participant to the comprehensive accident risk index, and determining the marginal contribution ratio of each participant as their respective contribution degree. S63, the risk type corresponding to the participant with the highest contribution is identified as the dominant risk type of the hotspot cluster; S64, extract the lock operation status and tidal phase of the corresponding spatiotemporal window of the hot spot cluster, and combine them to form the lock control-tidal trigger condition label; S65 combines the dominant risk type with the gate-tidal trigger condition label to generate hotspot cluster attribution results.

[0013] Furthermore, adjusting the iteration execution time interval based on the location of the hotspot cluster and the corresponding hotspot cluster attribution result includes: S71, obtain the location information of each hotspot cluster generated in this iteration and the corresponding hotspot cluster attribution results; S72, based on location information, extracts the channel attribute labels of the locations of each hotspot cluster; S73, the channel attribute labels are coupled with the dominant risk type in the hotspot cluster attribution results to determine the priority of hazard handling for each hotspot cluster; the coupling determination is performed according to a preset location-risk coupling rule table, which defines the hazard level corresponding to different combinations of channel attribute labels and different dominant risk types; S74, adjust the time interval for the next iteration based on the danger handling priority of each hotspot cluster; wherein, when there is a hotspot cluster whose danger handling priority reaches a preset threshold, the time interval is reduced; when the danger handling priorities of all hotspot clusters are lower than the preset threshold, the time interval is increased.

[0014] Furthermore, the accident pre-incident handling plan includes at least a preventive patrol plan, an emergency force pre-deployment plan, and a lock scheduling and traffic flow control plan.

[0015] This invention provides a system for identifying and handling hotspot areas of canal water accidents. The system is used to implement any of the methods for identifying and handling hotspot areas of canal water accidents, and includes: The multi-source data fusion module is used to acquire ship navigation data, lock scheduling data and tidal hydrological data. After data cleaning and spatiotemporal alignment, a unified spatiotemporal gridded waterway model is constructed. The feature extraction module is used to extract tidal driving features, ship traffic conflict features, and lock driving features based on the spatiotemporal gridded waterway model. The risk index calculation module is used to determine the crossflow risk index based on the tidal direction and channel location and the lateral velocity component in the tidal driving characteristics; and to determine the ship encounter thermal value based on the ship traffic conflict characteristics. The dynamic weight fusion module is used to dynamically allocate fusion weights to the crossflow risk index, the ship encounter thermal value, and the lock drive characteristics according to the current lock operation status and tidal phase. After nonlinear weighted fusion, the comprehensive accident risk index of each spatiotemporal grid unit is calculated, and a risk-weighted spatiotemporal cube is constructed. The hotspot identification module is used to identify hotspot clusters where the comprehensive accident risk index is concentrated on a risk-weighted spatiotemporal cube, using a spatiotemporal scanning statistical method based on the Poisson distribution model and moving a scanning window of variable size. The attribution analysis module is used to reverse analyze the contribution of the crossflow risk index and the ship encounter thermal value at the location of the hot spot cluster, calculate the corresponding gate control state and tidal phase, and generate the hot spot cluster attribution result based on the contribution, gate control state and tidal phase. The iteration and solution generation module is used to iteratively execute S3 to S6 to update the hotspot clusters and corresponding hotspot cluster attribution results for the target time period, and adjust the time interval of iteration execution based on the location of the hotspot clusters and the corresponding hotspot cluster attribution results; and generate differentiated accident pre-handling solutions based on the updated hotspot clusters and corresponding hotspot cluster attribution results.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: Firstly, this invention uses lock scheduling data and tidal hydrological data as driving variables for accident hotspot identification, extracts lock-tidal dual-drive risk features including crossflow risk index, ship encounter thermal value, and lock driving characteristics, and dynamically allocates fusion weights according to the current lock operating status and tidal phase. This solves the problem that existing technologies rely solely on historical accident frequency or single traffic flow characteristics and cannot capture the dynamic migration of accident hotspots with lock opening and closing and tidal rise and fall. It improves the accuracy of the comprehensive accident risk index in depicting the spatiotemporal evolution of risk under the coupling effect of lock scheduling and tidal, and enables accident hotspot identification to be dynamically updated with real-time changes in lock control conditions and tidal phase. Secondly, this invention employs a game theory-based attribution decomposition algorithm, using the cross-current risk index, ship encounter thermal value, and lock drive characteristics as game participants. It calculates the marginal contribution of each participant to the comprehensive accident risk index and determines the degree of contribution. It reverse-analyzes the dominant risk type of each hot spot cluster and combines it with the calculated gate control status and tidal phase to generate hot spot cluster attribution results with gate control-tidal trigger condition labels. This solves the problem that existing technologies can only output the risk location but cannot automatically analyze the risk cause. It achieves a technological leap from accident risk attribution, improves the attribution transparency and interpretability of hot spot analysis results, and enables the quantification and differentiation of the risk composition differences of different hot spot clusters at the same time or the same hot spot cluster at different time periods. Third, this invention integrates multi-source data fusion modeling, dual-drive risk feature extraction, dynamic weight fusion, spatiotemporal scanning hotspot identification, game-theoretic attribution analysis, and location-risk coupling-based adaptive iterative control into a complete closed-loop technical solution. This solves the problems of disconnect between risk identification and response actions and the inability to adaptively adjust fixed monitoring frequencies in existing technologies. It improves the efficiency of the entire chain of linkage from risk perception to response actions. While realizing dynamic monitoring and attribution analysis of accident hotspots, it generates differentiated accident pre-processing schemes, including preventive patrol schemes, emergency force pre-deployment schemes, and lock scheduling and control schemes. It also automatically shortens the monitoring interval during high-risk periods and automatically extends the monitoring interval during low-risk periods, thereby improving the intelligence level and resource allocation efficiency of canal navigation safety supervision. Attached Figure Description

[0017] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating a method for identifying hotspot areas of waterway accidents in canals and handling such accidents, provided in an embodiment of the present invention; Figure 2 This is a structural block diagram of a canal water accident hotspot identification and accident handling system provided in an embodiment of the present invention. Detailed Implementation

[0019] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0020] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0021] It should be noted that in the description of this invention, the terms "upper", "lower", "left", "right", "inner", "outer", etc., which indicate directions or positional relationships, are based on the directions or positional relationships shown in the accompanying drawings. This is only for the convenience of description and is not intended to indicate or imply that the device or element must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, it should not be construed as a limitation of this invention.

[0022] Furthermore, it should be noted that, in the description of this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0023] Example 1 like Figure 1 As shown, this invention provides a method for identifying hotspots of waterway accidents in canals and for handling such accidents, including: S1 acquires ship navigation data, lock scheduling data, and tidal hydrological data. After data cleaning and spatiotemporal alignment, a unified spatiotemporal gridded waterway model is constructed, including: S11, acquire Automatic Identification System (AIS) data as ship navigation data, acquire lock scheduling record data as lock scheduling data, and acquire measured tide data and tide forecast data as tidal hydrological data. S12, clean the AIS data, remove position jumps, velocity abrupt changes and drift data, and interpolate the missing segments of the cleaned trajectory to obtain a continuous and smooth ship navigation trajectory. S13, convert the lock number, opening and closing time and release direction in the lock scheduling record data into a first continuous time series, and convert the tide level, flow velocity and flow direction in the actual tide data and tide forecast data into a second continuous time series. S14, align the continuous and smooth ship navigation trajectory, the first continuous time series, and the second continuous time series to the same time axis; S15, the canal waterway is divided into geographic grid units with preset side lengths, and a waterway attribute label is attached to each geographic grid unit. The waterway attribute label includes at least one of the following: entrance area, approach channel, bridge area, bend, and waiting anchorage. The unified spatiotemporal gridded waterway model is then constructed.

[0024] In one possible implementation, AIS base stations deployed along the coast receive and parse navigation data automatically broadcast by ships, including the Maritime Mobile Service Identifier (MMSI), longitude, latitude, speed over land, heading over land, and timestamp, which serves as the ship's navigation data. Lock scheduling data is obtained from the lock operation management system, including lock number, gate opening time, gate closing time, release direction, and a list of ships released within each lock session. Tidal hydrological data is simultaneously accessed from two sources: real-time monitoring data from tide gauges and current meters deployed at key sections of the waterway; and hourly tide level, current velocity, and direction forecasts published by the tide forecasting department. These two sources complement each other, together constituting the tidal hydrological data.

[0025] Each acquired AIS data point is examined. If a record's longitude or latitude changes by more than a preset offset threshold compared to its preceding or following records, it is identified as a position jump and removed. If the rate of change of ground speed between adjacent records exceeds a preset acceleration threshold, it is identified as a speed abrupt change and removed. If a ship's position remains unchanged for an extended period while not anchored, it is identified as drift data and removed. For missing segments in the cleaned trajectory caused by data removal or signal loss, cubic spline interpolation is used to fill in the ship's position coordinates at intermediate moments based on the position and time information of valid trajectory points before and after the missing segment, ultimately obtaining a continuous and smooth navigation trajectory for each ship.

[0026] The opening and closing times of each lock in the lock scheduling record are extracted as time tags, and the release direction is converted into a binary status identifier, for example, upstream is recorded as 1 and downstream as 0. The first continuous time series is constructed on the time axis with locks as the unit. The tide level, flow velocity and flow direction angle in the actual and forecast tide data are arranged in sequence according to their corresponding time tags to form a second continuous time series with a fixed sampling interval, such as every 5 minutes or every 15 minutes.

[0027] Using Coordinated Universal Time (UTC) as the baseline time axis, the timestamps of each trajectory point in the obtained ship navigation trajectory are matched and aligned with the timestamps in the first and second continuous time series. For data points that are not fully synchronized in time, nearest neighbor matching or time-weighted interpolation is used to ensure that the three types of data are consistent in the time dimension.

[0028] The target canal section is divided into continuous geographic grid units along the centerline of the channel according to a preset side length. In this embodiment, this is set to 100 meters. Each grid unit is uniquely identified by the latitude and longitude coordinates of its center point. Simultaneously, channel attribute information is extracted from the electronic nautical chart, and channel attribute labels are attached to each grid unit. Label types include: entrance area (the section connecting the upstream and downstream approach channels and the main channel of the lock), approach channel (the transition section upstream and downstream of the lock), bridge area (the bridge and the water area within a certain range upstream and downstream), bend (the section with a channel curvature radius less than a preset value), and waiting anchorage (the designated water area upstream and downstream of the lock for vessels to wait for passage). A one-to-many mapping relationship is established between the channel attribute labels and the geographic grid units, thereby completing the construction of a unified spatiotemporal gridded channel model.

[0029] This invention solves the problems of trajectory discontinuity and noise interference caused by signal loss or transmission errors in the original AIS data by removing position jumps, velocity abrupt changes, and drift data in AIS data and interpolating missing trajectory segments. This improves the continuity and smoothness of ship navigation trajectories and enhances data accuracy. By dividing the canal channel into geographic grid units with preset side lengths and attaching channel attribute labels including entrance areas, pilotways, bridge areas, bends, and waiting anchorages to each grid unit, this invention addresses the problems of traditional methods lacking unified spatial analysis units and difficulty in distinguishing inherent risk differences in different navigation sections. It provides a standardized spatial reference system for determining accident hotspot locations.

[0030] S2, based on a spatiotemporal gridded waterway model, extracts tidal driving features, ship traffic conflict features, and lock driving features; S3, determine the crossflow risk index based on the tidal direction and channel location, as well as the lateral velocity component, within the tidal driving characteristics; determine the ship encounter thermal value based on ship traffic conflict characteristics, including: S31, obtain the angle between the tidal flow direction and the channel centerline and the lateral velocity component of the tide in each spatiotemporal gridded channel model; S32, the product of the sine of the included angle and the transverse velocity component is determined as the basic value of the transverse flow risk; S33, obtain the channel attribute label of the spatiotemporal grid unit, determine the terrain sensitivity correction coefficient based on the channel attribute label, and correct the crossflow risk base value based on the terrain sensitivity correction coefficient to generate the crossflow risk index.

[0031] S34 uses a density-based clustering algorithm to identify overtaking, crossing, and encounter situations where ship trajectories are clustered in time and space in the characteristics of ship traffic conflicts. S35, extract the ship encounter density and minimum encounter distance in the areas where each encounter situation occurs; S36. Using the ship encounter density and minimum encounter distance as inputs, a weighted fusion algorithm is used to calculate the ship encounter thermal value; wherein, the ship encounter thermal value is positively correlated with the ship encounter density and negatively correlated with the minimum encounter distance.

[0032] In one possible implementation, risk-related features are extracted from three dimensions based on a spatiotemporal gridded waterway model. Regarding tidal-driven features, the tidal level, velocity, direction angle, and tidal phase identifier for each spatiotemporal grid cell are extracted from the second continuous time series of tidal hydrological data, forming a tidal-driven feature set. Regarding vessel traffic conflict features, based on continuous and smooth vessel navigation trajectories, the instantaneous number of vessels, vessel density distribution, relative distance between vessels, relative heading difference, and encounter situation type within each spatiotemporal grid cell are extracted, forming a vessel traffic conflict feature set. Regarding lock-driven features, the lock opening / closing status, release direction, upstream / downstream water level difference, and the increase in discharge velocity caused by gate opening are extracted from the first continuous time series of lock scheduling data, forming a lock-driven feature set.

[0033] For each spatiotemporal grid cell, the tidal flow direction angle of the cell at the current moment is obtained from the tidal drive feature set, and the direction angle of the channel centerline at the cell's location is obtained from the channel attribute data. The angle θ between the two is calculated. At the same time, the velocity value V of the cell is obtained from the tidal drive feature set, and the velocity V is decomposed into a longitudinal component along the channel direction and a transverse velocity component perpendicular to the channel direction based on the angle θ.

[0034] Multiplying the sine value of the included angle θ, sin(θ), with the lateral velocity component V_lateral yields the basic value of the crossflow risk for that spatiotemporal grid cell. The closer the angle between the flow direction and the channel centerline is to 90 degrees and the larger the lateral velocity component is, the greater the lateral thrust of the crossflow on the ship, and the higher the basic value of the crossflow risk.

[0035] Obtain the channel attribute labels for the spatiotemporal grid cell and query the preset topographic sensitivity correction coefficient table based on the label type. This coefficient table is set according to the channel's hydraulic characteristics and historical accident experience. For example, in the entrance area and bridge area, the flow velocity locally increases due to the water constriction effect, and the correction coefficient can be set to a value greater than 1; in the bend area, due to the superposition effect of centrifugal force and crossflow, the correction coefficient can be set to a value greater than 1; in the pilot channel and waiting anchorage area, the water flow is relatively gentle, and the correction coefficient can be set to 1. Multiply the basic crossflow risk value by the corresponding topographic sensitivity correction coefficient to obtain the final crossflow risk index. This index comprehensively reflects the coupled influence of crossflow intensity and channel topography on the lateral stability of ships.

[0036] Density-based clustering analysis of ship trajectories in space and time, such as a spatiotemporal extension of the DBSCAN algorithm, is performed. By setting neighborhood distance thresholds and minimum trajectory point count thresholds, clusters of spatially adjacent and temporally synchronized trajectory points are identified as ship encounter situations. Based on the relative positions and headings of the ships in the clustering results, the types of encounter situations are further distinguished: those traveling in the same direction with a speed difference leading to a reduced distance are identified as overtaking situations; those with intersecting headings and a risk of collision are identified as crossing encounter situations; and those with approximately opposite headings and approaching each other are identified as oncoming encounter situations.

[0037] For each identified encounter situation area, the frequency of encounter situations within a unit of time and space is statistically analyzed and used as the ship encounter density D for that area. Simultaneously, the minimum encounter distance between all ship pairs within that area is calculated, and the minimum value among all ship pairs is taken as the minimum encounter distance for that area. Using the ship encounter density and minimum encounter distance as inputs, a weighted fusion algorithm is used to calculate the ship encounter heat value H. A higher ship encounter density indicates more frequent traffic conflicts in the area, resulting in a higher heat value; a smaller minimum encounter distance indicates a more urgent encounter situation, also resulting in a higher heat value. The resulting ship encounter heat value comprehensively reflects the degree of aggregation and urgency of collision risks between ships.

[0038] This invention addresses the problem of distorted crossflow risk assessment caused by traditional methods that only use absolute velocity values ​​while ignoring the relative relationship between flow direction and channel within each spatiotemporal grid cell, by obtaining the angle between the tidal flow direction and the channel centerline, as well as the lateral velocity component. It determines the basic value of crossflow risk by multiplying the sine of the angle by the lateral velocity component, thus improving the physical accuracy of crossflow risk quantification. Furthermore, by obtaining the channel attribute labels of the spatiotemporal grid cells and determining a terrain sensitivity correction coefficient based on these labels to correct the basic value of crossflow risk, this invention solves the problem that a unified calculation formula cannot distinguish the differences in the amplification or attenuation effects of different terrain features such as entrance areas, bridge areas, and bends on crossflow effects. This achieves a differential crossflow risk index for channel terrain characteristics. The alienated response improves the spatial precision of cross-current risk assessment. By using a density-based clustering algorithm, it identifies spatiotemporally clustered overtaking, crossing, and head-on encounter situations from the characteristics of ship traffic conflicts. This solves the problem that traditional methods only measure traffic conflicts based on ship number density while ignoring the actual encounter types and spatial relationships between ships, thus improving the accuracy of ship encounter situation identification and the ability to distinguish between types. By extracting the ship encounter density and minimum encounter distance in the areas where each encounter situation occurs, and using both as inputs, a weighted fusion algorithm is used to calculate the ship encounter heat value. This solves the problem that a single indicator cannot simultaneously reflect the frequency and urgency of encounter situations, thus improving the completeness and discriminativeness of the comprehensive characterization of ship collision risks.

[0039] S4. Based on the current lock operation status and tidal phase, the crossflow risk index, the ship encounter thermal value, and the lock drive characteristics are dynamically assigned fusion weights. After nonlinear weighted fusion, the comprehensive accident risk index of each spatiotemporal grid unit is calculated, and a risk-weighted spatiotemporal cube is constructed. Based on the current lock operation status and tidal phase, dynamic fusion weights are assigned to the crossflow risk index, ship encounter thermal value, and lock drive characteristics, including: An adaptive weight regulator is constructed and trained to learn the nonlinear mapping relationship from the combined working conditions of the lock operation state and tidal phase to the optimal weight allocation. During high tide, the output weights of the crossflow risk index and the thermal value encountered by the ship are increased to highlight the collision risk. During the lock discharge period, the output weights of the lock drive characteristics are increased to highlight the bottoming risk.

[0040] During each feature fusion, the current lock operation status and tidal phase are input into the trained adaptive weight regulator to output the crossflow risk index, the ship encounter thermal value, and the fusion weights corresponding to the lock drive features.

[0041] In one possible implementation, an adaptive weighted regulator is constructed. This regulator is implemented using a neural network model, with its input layer receiving lock operation status identifiers and tidal phase identifiers. The lock operation status identifiers include four values: gate closed, gate open - upstream passage, gate open - downstream passage, and water release operation; the tidal phase identifiers include five values: rapid rise, moderate rise, rapid fall, moderate fall, and turning point. These two values ​​combine to form twenty joint operating conditions. The regulator's output layer outputs three weight values, corresponding to the crosscurrent risk index, the ship encounter thermal value, and the lock drive characteristics, respectively, with the sum of the three weight values ​​being 1.

[0042] The adaptive weight regulator is trained using historical accident data. The lock's operational status and tidal phase at the time of each accident are extracted from historical accident records to form the input of the training samples. Based on the accident type labeling in the accident investigation report, the dominant risk factors for each accident are determined, and the expected output of the training samples is constructed accordingly. For example, if a collision occurs during high tide, the weights of the cross-current risk index and the ship encounter thermal value in the expected output should be higher; if a grounding accident occurs during lock discharge, the weights of the lock drive characteristics in the expected output should be higher. Through iterative training with a large number of samples, the regulator learns a nonlinear mapping relationship from combined operating conditions to optimal weight allocation. Specifically, during high tide, the regulator automatically increases the output weights of the cross-current risk index and the ship encounter thermal value to highlight the collision risk caused by cross-current pushing and dense ship encounters; during lock discharge, the regulator automatically increases the output weights of the lock drive characteristics to highlight the risks of grounding and bottoming caused by a sudden drop in water level.

[0043] During each feature fusion operation, the current lock operating status and tidal phase are acquired, converted into corresponding identifier values, and input into a trained adaptive weight regulator. The regulator outputs the fusion weights corresponding to the cross-current risk index, the ship encounter thermal value, and the lock actuation features under the current operating conditions. For each spatiotemporal grid cell, the cross-current risk index, the ship encounter thermal value, and the lock actuation features are normalized separately, and a nonlinear weighted fusion method is used to calculate the comprehensive accident risk index.

[0044] After traversing all spatiotemporal grid cells and calculating their comprehensive accident risk index one by one, a three-dimensional risk-weighted spatiotemporal cube is constructed by filling the comprehensive accident risk index value of each grid cell into the corresponding position, using the X and Y coordinates of the geographic grid as the spatial dimension and the time axis as the third dimension. Each voxel of this cube represents the risk quantification of a spatiotemporal grid cell, providing a standardized input data structure for subsequent spatiotemporal scanning hotspot identification.

[0045] This invention addresses the problem of inaccurate risk assessment caused by the inability of traditional fixed-weight fusion methods to adapt to dynamic changes in the combined lock control-tidal conditions by constructing and training an adaptive weight adjuster. This learns the nonlinear mapping relationship from the combined lock operation state and tidal phase to the optimal weight allocation, thus improving the adaptive matching capability of the comprehensive accident risk index to spatiotemporally changing scenarios. By automatically increasing the output weights of the crossflow risk index and the thermal values ​​encountered by ships during high tide, and automatically increasing the output weights of the lock drive characteristics during lock discharge, this invention solves the problem that traditional methods using uniform weights cannot highlight major risk factors despite significant differences in dominant risk types under different operating conditions. This ensures that the weight allocation aligns with the accident causation patterns of the current operating condition, improving the operating condition specificity of risk fusion and the sensitivity of dominant risk identification. Furthermore, by constructing the comprehensive accident risk index of each spatiotemporal grid unit as a risk-weighted spatiotemporal cube with geographical coordinates as the spatial dimension and time as the third dimension, this invention solves the problem that traditional two-dimensional risk maps cannot express the evolution of risks along the time axis. This provides a three-dimensional integrated analysis data structure for subsequent spatiotemporal scanning statistics, improving the joint analysis capability of hotspot identification in both time and space dimensions.

[0046] S5, on the risk-weighted spatiotemporal cube, with the comprehensive accident risk index as input, adopts a spatiotemporal scanning statistical method based on the Poisson distribution model, moves a scanning window of variable size, and identifies hotspot clusters where the comprehensive accident risk index gathers; In one possible implementation, a scanning window is defined on a risk-weighted spatiotemporal cube. The scanning window is a cylindrical shape with a circular base, its spatial dimension controlled by the base radius R, and its temporal dimension controlled by the window height H. To capture risk clustering patterns at different scales, the base radius R varies between a preset minimum and maximum radius with a certain step size, for example, from 200 meters to 1000 meters with a step size of 100 meters; the window height H varies between a preset minimum and maximum duration with a certain step size, for example, from 30 minutes to 4 hours with a step size of 30 minutes. The center of the window's base circle traverses the spatial center coordinates of each spatiotemporal grid cell, and the window's temporal midpoint traverses every moment on the time axis, thereby generating a large number of candidate scanning windows covering the entire spatiotemporal cube.

[0047] For each candidate scanning window, its log-likelihood ratio is calculated to measure the significance of risk clustering. The sum of the comprehensive accident risk indices of all spatiotemporal grid cells within the window is used as the observed risk value O within the window. Based on the null hypothesis of the Poisson distribution model, the expected risk value E of the window under the condition of uniform risk distribution is calculated, i.e., E = (number of grid cells in the window / total number of grid cells in the spatiotemporal cube) × total risk value. The log-likelihood ratio LLR of the window is calculated using the following formula: When O > E, LLR = O × ln(O / E) + (T - O) × ln((T - O) / (T - E)); when O ≤ E, LLR = 0. Here, T is the sum of the comprehensive accident risk indices for the entire spatiotemporal cube. The larger the log-likelihood ratio (LLR), the less likely the risk clustering within the window is to be caused by random fluctuations, i.e., the greater the probability of a significant hotspot cluster.

[0048] The Monte Carlo simulation method can be used to determine the significance threshold. The comprehensive accident risk index in the risk-weighted spatiotemporal cube is randomly rearranged multiple times to generate a large number of simulated datasets. The same scanning process is performed on each simulated dataset, and its maximum LLR value is recorded. The preset quantile of the maximum LLR value in the simulated dataset is taken as the significance threshold. All candidate scan windows with LLR values ​​exceeding this significance threshold are determined as significant windows.

[0049] Significant windows are merged by grouping those with a spatial center distance less than a preset spatial proximity threshold (e.g., 500 meters) and a temporal overlap exceeding a preset temporal overlap ratio (e.g., 50%) into a single hotspot cluster. The spatial coverage of each hotspot cluster is the union of the spatial bases of all its internal significant windows, and its temporal coverage is the union of the time periods of all its internal significant windows. This completes the identification of hotspot clusters aggregated by the comprehensive accident risk index; each hotspot cluster represents an accident hotspot water area and its duration.

[0050] This invention, S5, solves the problem that traditional two-dimensional hotspot analysis can only discover clusters in the spatial dimension but cannot simultaneously lock high-risk periods by scanning a three-dimensional spatiotemporal cube with a comprehensive accident risk index as input. It achieves joint localization of accident hotspots in both spatial location and time window dimensions, improving the spatiotemporal integrity of hotspot identification results. By moving a variable-sized scanning window to traverse the entire spatiotemporal cube, it solves the problem that fixed-scale windows cannot adapt to the detection needs of hotspot clusters with different spatial ranges and time spans. This allows for the effective capture of hotspots ranging from short-term clusters in local flight segments to continuous high-risk periods over long distances, improving the multi-scale adaptability and detection sensitivity of hotspot identification.

[0051] S6, reverse analyze the contribution of the cross-current risk index and the ship encounter thermal value at the location of the hotspot cluster, calculate the corresponding gate control state and tidal phase, and generate the hotspot cluster attribution result based on the contribution, gate control state and tidal phase, including: S61, extract the crossflow risk index, ship encounter thermal value, lock drive characteristics and comprehensive accident risk index of each spatiotemporal grid unit within the spatiotemporal window where the hot spot cluster is located; S62 uses a game theory-based attribution decomposition algorithm, taking the cross-current risk index, the ship encounter thermal value, and the lock drive characteristics as participants, calculating the marginal contribution of each participant to the comprehensive accident risk index, and determining the marginal contribution ratio of each participant as their respective contribution degree. S63, the risk type corresponding to the participant with the highest contribution is identified as the dominant risk type of the hotspot cluster; S64, extract the lock operation status and tidal phase of the corresponding spatiotemporal window of the hot spot cluster, and combine them to form the lock control-tidal trigger condition label; S65 combines the dominant risk type with the gate-tidal trigger condition label to generate hotspot cluster attribution results.

[0052] In one possible implementation, based on the spatial and temporal coverage of each identified hotspot cluster, all spatiotemporal grid cells contained within the spatiotemporal window are extracted from the risk-weighted spatiotemporal cube, and the crossflow risk index, ship encounter thermal value, lock drive characteristics, and comprehensive accident risk index of each grid cell are obtained to form the feature dataset of the hotspot cluster.

[0053] A game theory-based attribution decomposition algorithm, such as the Shapley value method, is used to perform risk attribution decomposition for each hotspot cluster. Specifically, the cross-current risk index, ship encounter thermal value, and lock actuation characteristics are considered as three game players, and the comprehensive accident risk index is considered as the total payoff contributed by all three. The goal of attribution decomposition is to calculate the marginal contribution of each player to the total payoff. All possible joining orders of the three players are enumerated, and for each permutation, the marginal gain of each player before and after joining the predicted value of the comprehensive accident risk index is calculated. The average marginal gain of a player across all permutations is taken as the marginal contribution value of that player. Then, the marginal contribution values ​​of the three players are divided by the sum of their marginal contribution values ​​to obtain the contribution degree of each player, and the sum of the three is 1. For example, for a certain hotspot cluster, the calculated contribution degree is: 0.65 for the cross-current risk index, 0.25 for the ship encounter thermal value, and 0.10 for the lock actuation characteristics. This indicates that the overall accident risk of this hotspot cluster is mainly driven by crossflow factors.

[0054] By comparing the three calculated contribution values, the risk type corresponding to the participant with the highest contribution is determined as the dominant risk type of the hotspot cluster. The cross-current risk index corresponds to cross-current dominant risk, mainly manifested as the risk of yaw, loss of control, and grounding caused by the thrust of lateral currents; the ship encounter thermal value corresponds to encounter dominant risk, mainly manifested as the risk of collision caused by dense encounters; and the lock drive characteristics correspond to lock drive risk, mainly manifested as the risk of grounding and bottoming caused by sudden changes in water level and sudden increases in flow velocity due to lock discharge or release. Continuing the example, the highest contribution is the cross-current risk index (0.65), therefore the dominant risk type of this hotspot cluster is determined to be cross-current dominant.

[0055] Extract the lock operation status and tidal phase information within the spatiotemporal window corresponding to the hotspot cluster. From the constructed first continuous time series, obtain the gate opening and closing status and release direction within this time window, and combine them to form a lock status description, such as lock discharge, gate opening - upstream release, or gate closing. From the constructed second continuous time series, obtain the tidal phase identifier within this time window, such as high tide, low tide, or transition point. Combine the two to form a lock control-tidal trigger condition label, such as low tide + lock discharge, high tide + dense gate release, or slack tide + gate closing and waiting. Combine the determined dominant risk type with the generated lock control-tidal trigger condition label to form the complete attribution result for this hotspot cluster.

[0056] This invention extracts the crossflow risk index, ship encounter thermal value, and lock drive characteristics of each grid cell within the spatiotemporal window of a hotspot cluster. It then employs a game theory-based attribution decomposition algorithm to calculate the marginal contribution of each participant to the comprehensive accident risk index. This addresses the problem of traditional hotspot identification methods that only output the location of risk levels without quantifying the risk composition, thus improving the interpretability and attribution accuracy of hotspot analysis results. Furthermore, by considering the marginal gain of each risk feature across all possible addition orders using game theory attribution decomposition algorithms such as the Shapley value method, it solves the problem that simple proportional allocation or single-factor sensitivity analysis cannot fairly measure the true contribution of each feature in the nonlinear fusion model. This ensures that the contribution calculation of each risk feature has game theory fairness and completeness, improving the accuracy and credibility of the attribution results. Finally, by extracting the lock operation status and tidal phase of the corresponding spatiotemporal window of the hotspot cluster and combining them to form a lock control-tidal triggering condition label, it solves the problem of risk attribution being divorced from the environmental context. This establishes a causal relationship between the risk causes and specific lock control operations and tidal conditions, improving the contextual completeness of the attribution conclusions.

[0057] S7 iterates through S3 to S6 to update the hotspot clusters and their corresponding attribution results for the target time period, and adjusts the time interval of the iteration based on the location of the hotspot clusters and their corresponding attribution results; and generates differentiated incident pre-handling solutions based on the updated hotspot clusters and their corresponding attribution results.

[0058] Based on the location of the hotspot clusters and the corresponding hotspot cluster attribution results, the time interval for iteration execution is adjusted, including: S71, obtain the location information of each hotspot cluster generated in this iteration and the corresponding hotspot cluster attribution results; S72, based on location information, extracts the channel attribute labels of the locations of each hotspot cluster; S73, the channel attribute labels are coupled with the dominant risk type in the hotspot cluster attribution results to determine the priority of hazard handling for each hotspot cluster; the coupling determination is performed according to a preset location-risk coupling rule table, which defines the hazard level corresponding to different combinations of channel attribute labels and different dominant risk types; S74, adjust the time interval for the next iteration based on the danger handling priority of each hotspot cluster; wherein, when there is a hotspot cluster whose danger handling priority reaches a preset threshold, the time interval is reduced; when the danger handling priorities of all hotspot clusters are lower than the preset threshold, the time interval is increased.

[0059] The accident pre-incident handling plan should include at least a preventive patrol plan, an emergency force pre-deployment plan, and a lock scheduling and traffic flow control plan.

[0060] In one possible implementation, after each iteration from S3 to S6, the location information of each hotspot cluster generated in this round is obtained, including the latitude and longitude coordinates of the center point of its spatial coverage and the geographic grid cell number it covers; the hotspot cluster attribution results corresponding to each hotspot cluster are obtained, including the dominant risk type, gate-tidal trigger condition label, and the contribution distribution of each risk feature.

[0061] For each hotspot cluster, the corresponding grid cell's channel attribute label is extracted from the constructed spatiotemporal gridded channel model based on its geographic grid cell number in its location information. Channel attribute labels include entrance areas, pilotways, bridge areas, bends, and waiting anchorages. If a hotspot cluster covers multiple grid cells with different channel attribute labels, the label with the largest area percentage is taken as the channel attribute label for that hotspot cluster. The channel attribute label of each hotspot cluster is then coupled with the dominant risk type into a preset location-risk coupling rule table for determination. This rule table defines the hazard level corresponding to different combinations of channel attribute labels and different dominant risk types.

[0062] An example rule table configuration is as follows: When a dominant crossflow risk is superimposed on a bridge area, the water-binding effect of the bridge piers will further amplify the crossflow thrust, making it extremely easy for ships to lose control and collide with the bridge; therefore, the hazard level is high. When a dominant crossflow risk is superimposed on a gate area, the dense flow of ships released from the locks converges into the main channel at the gate area, resulting in an extremely high collision risk; therefore, the hazard level is high. When a dominant crossflow risk is superimposed on a curve, the centrifugal force and the crossflow thrust are superimposed in the same direction, making ships extremely prone to yaw and grounding; therefore, the hazard level is high. Conversely, in the waiting anchorage area, ships are moored or in a low-speed standby state, and the actual hazard level is relatively low regardless of the superimposed risk type. For each hotspot cluster, the combination of its channel attribute label and dominant risk type is found in the rule table to determine the corresponding hazard level, which is then converted into a hazard handling priority. Hotspot clusters with high, medium, and low hazard levels correspond to hazard handling priorities of Level 1, Level 2, and Level 3, respectively.

[0063] If the risk handling priority of any hotspot cluster reaches the preset threshold (in this embodiment, the preset threshold is set to level one), indicating that there is a high-risk cluster that needs close attention in the current flight segment, the time interval for the next iteration will be automatically reduced, for example, from the default 15 minutes to 5 minutes, in order to increase the monitoring frequency and capture the further evolution of the risk in a timely manner.

[0064] If the risk handling priority of all hotspot clusters in this round is lower than the preset threshold, for example, all are level three, indicating that the risk situation of the current flight segment is relatively stable, then the time interval for the next iteration will be automatically increased, for example, from the default 15 minutes to 30 minutes, in order to reduce the consumption of computing resources while ensuring the monitoring effect. If neither of the above two conditions is met, for example, there are level two clusters but no level one clusters, then the default time interval will remain unchanged.

[0065] Based on the updated hotspot cluster location information and attribution results, the following three types of differentiated incident pre-incident handling solutions are automatically generated: The preventative patrol plan extracts the spatial coverage of each hotspot cluster as the target route, extracts the time window from the gate control-tidal trigger condition labels of the hotspot cluster as the suggested patrol time window, and extracts the key vessel types corresponding to the risk characteristics with the highest contribution. For example, for cross-current dominated vessels, the plan focuses on heavy-load, deep-draft vessels, and for encounter dominated vessels, the plan focuses on high-speed container ships. Combining the above information generates a preventative patrol task list that includes patrol routes, patrol times, and key monitoring targets. The emergency response force deployment plan determines the allocation of emergency response forces based on the dominant risk type of each hotspot cluster. For cross-current-dominated hotspots, high-powered tugboats are deployed as the main emergency response force for emergency towing in case of vessel loss of control; for encounter-dominated hotspots, patrol boats are deployed as the main emergency response force for on-site traffic organization and order maintenance; for lock-driven hotspots, emergency rescue vessels and pollution prevention equipment are deployed for grounding rescue and oil spill prevention. The forward deployment location is selected at the upstream boundary of the hotspot cluster's spatial coverage area to ensure that emergency response forces can arrive at the scene in the shortest possible time after an incident. The lock scheduling and traffic flow control scheme generates lock release rhythm adjustment suggestions based on the lock control-tidal trigger condition labels of hotspot clusters. For example, when the label indicates low tide and current shift + lock discharge with crossflow as the dominant risk, it is recommended to suspend discharge operations or reduce the discharge flow until the current shift is complete. Based on the location of grid cells within the hotspot cluster where the crossflow risk index exceeds a preset threshold, it generates suggestions for vessel speed limit zones. Based on areas where the vessel encounter heat value exceeds a preset threshold, it generates temporary anchorage allocation suggestions to guide some vessels to temporarily anchor and wait, thereby reducing vessel density in hotspot areas.

[0066] This invention solves the problem that traditional methods can only perform single static identification and cannot dynamically track the risk evolution process by iteratively executing S3 to S6 to continuously update hotspot clusters and attribution results for the target time period. This achieves a technical upgrade from static snapshots to dynamic continuous monitoring of accident hotspots, improving the timeliness and continuity of risk situation awareness. By obtaining the channel attribute labels of each hotspot cluster's location and coupling them with the dominant risk type in the attribution results to determine the priority of hazard handling, this invention solves the problem that traditional methods ignore the interaction between location attributes and risk types, thus failing to distinguish the urgency of handling. This allows the same dominant risk type to be assigned differentiated handling priorities in different flight segments, improving the refinement of risk classification management and the rationality of decision-making. Furthermore, by ensuring that the hazard handling priority reaches a preset level... By reducing the iteration interval when hotspot clusters are at thresholds and increasing the iteration interval when the priority of handling hotspot clusters is lower than the preset threshold, the problem of delayed early warnings when the fixed monitoring frequency is used and wasted resources when the risk is stable is solved. This achieves adaptive matching between monitoring frequency and risk status, improving the timeliness of early warnings during high-risk periods while reducing computational resource consumption during low-risk periods. By automatically generating differentiated accident pre-handling plans, including preventive patrol plans, emergency force pre-deployment plans, and lock scheduling and traffic flow control plans, based on the updated hotspot cluster locations and attribution results, the problem of disconnect between identification results and handling actions in traditional methods is solved. This achieves a closed loop from risk identification to precise handling, improving the allocation efficiency of maritime regulatory resources and the accuracy of accident prevention.

[0067] Example 2 like Figure 2 As shown, the present invention also provides a canal waterway accident hotspot identification and accident handling system, the system being used to implement the canal waterway accident hotspot identification and accident handling method described in any of Embodiment 1, the system comprising: The multi-source data fusion module is used to acquire ship navigation data, lock scheduling data and tidal hydrological data. After data cleaning and spatiotemporal alignment, a unified spatiotemporal gridded waterway model is constructed. The feature extraction module is used to extract tidal driving features, ship traffic conflict features, and lock driving features based on the spatiotemporal gridded waterway model. The risk index calculation module is used to determine the crossflow risk index based on the tidal direction and channel location and the lateral velocity component in the tidal driving characteristics; and to determine the ship encounter thermal value based on the ship traffic conflict characteristics. The dynamic weight fusion module is used to dynamically allocate fusion weights to the crossflow risk index, the ship encounter thermal value, and the lock drive characteristics according to the current lock operation status and tidal phase. After nonlinear weighted fusion, the comprehensive accident risk index of each spatiotemporal grid unit is calculated, and a risk-weighted spatiotemporal cube is constructed. The hotspot identification module is used to identify hotspot clusters where the comprehensive accident risk index is concentrated on a risk-weighted spatiotemporal cube, using a spatiotemporal scanning statistical method based on the Poisson distribution model and moving a scanning window of variable size. The attribution analysis module is used to reverse analyze the contribution of the crossflow risk index and the ship encounter thermal value at the location of the hot spot cluster, calculate the corresponding gate control state and tidal phase, and generate the hot spot cluster attribution result based on the contribution, gate control state and tidal phase. The iteration and solution generation module is used to iteratively execute S3 to S6 to update the hotspot clusters and corresponding hotspot cluster attribution results for the target time period, and adjust the time interval of iteration execution based on the location of the hotspot clusters and the corresponding hotspot cluster attribution results; and generate differentiated accident pre-handling solutions based on the updated hotspot clusters and corresponding hotspot cluster attribution results.

[0068] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for identifying hotspot areas for waterborne accidents in canals and for handling such accidents, characterized in that, include: S1: Acquire ship navigation data, lock scheduling data, and tidal hydrological data. After data cleaning and spatiotemporal alignment, construct a unified spatiotemporal gridded waterway model. S2, based on a spatiotemporal gridded waterway model, extracts tidal driving features, ship traffic conflict features, and lock driving features; S3, determine the crossflow risk index based on the tidal direction and channel position and the lateral velocity component in the tidal driving characteristics; determine the ship encounter thermal value based on the ship traffic conflict characteristics. S4. Based on the current lock operation status and tidal phase, the crossflow risk index, the ship encounter thermal value, and the lock drive characteristics are dynamically assigned fusion weights. After nonlinear weighted fusion, the comprehensive accident risk index of each spatiotemporal grid unit is calculated, and a risk-weighted spatiotemporal cube is constructed. S5, on the risk-weighted spatiotemporal cube, with the comprehensive accident risk index as input, adopts a spatiotemporal scanning statistical method based on the Poisson distribution model, moves a scanning window of variable size, and identifies hotspot clusters where the comprehensive accident risk index gathers; S6, reverse analyze the contribution of the crossflow risk index and the ship encounter thermal value of the hot spot cluster location, calculate the corresponding gate control state and tidal phase, and generate the hot spot cluster attribution result based on the contribution, gate control state and tidal phase; S7 iterates through S3 to S6 to update the hotspot clusters and their corresponding attribution results for the target time period, and adjusts the time interval of the iteration based on the location of the hotspot clusters and their corresponding attribution results; and generates differentiated incident pre-handling solutions based on the updated hotspot clusters and their corresponding attribution results.

2. The method for identifying hotspot areas and handling accidents on canals according to claim 1, characterized in that, The acquisition of ship navigation data, lock scheduling data, and tidal hydrological data, followed by data cleaning and spatiotemporal alignment, to construct a unified spatiotemporal gridded waterway model includes: S11, acquire Automatic Identification System (AIS) data as ship navigation data, acquire lock scheduling record data as lock scheduling data, and acquire measured tide data and tide forecast data as tidal hydrological data. S12, clean the AIS data, remove position jumps, velocity abrupt changes and drift data, and interpolate the missing segments of the cleaned trajectory to obtain a continuous and smooth ship navigation trajectory. S13, convert the lock number, opening and closing time and release direction in the lock scheduling record data into a first continuous time series, and convert the tide level, flow velocity and flow direction in the actual tide data and tide forecast data into a second continuous time series. S14, align the continuous and smooth ship navigation trajectory, the first continuous time series, and the second continuous time series to the same time axis; S15, the canal waterway is divided into geographic grid units with preset side lengths, and a waterway attribute label is attached to each geographic grid unit. The waterway attribute label includes at least one of the following: entrance area, approach channel, bridge area, bend, and waiting anchorage. The unified spatiotemporal gridded waterway model is then constructed.

3. The method for identifying hotspot areas and handling accidents on canals according to claim 2, characterized in that, The determination of the crossflow risk index based on the tidal direction and channel position, as well as the lateral velocity component, in the tidal driving characteristics includes: S31, obtain the angle between the tidal flow direction and the channel centerline and the lateral velocity component of the tide in each spatiotemporal gridded channel model; S32, the product of the sine of the included angle and the transverse velocity component is determined as the basic value of the transverse flow risk; S33, obtain the channel attribute label of the spatiotemporal grid unit, determine the terrain sensitivity correction coefficient based on the channel attribute label, and correct the crossflow risk base value based on the terrain sensitivity correction coefficient to generate the crossflow risk index.

4. The method for identifying hotspot areas and handling accidents on canals according to claim 1, characterized in that, The determination of the encounter thermal value of a ship based on the characteristics of ship traffic conflicts includes: S34 uses a density-based clustering algorithm to identify overtaking, crossing, and encounter situations where ship trajectories are clustered in time and space in the characteristics of ship traffic conflicts. S35, extract the ship encounter density and minimum encounter distance in the areas where each encounter situation occurs; S36. Using the ship encounter density and minimum encounter distance as inputs, a weighted fusion algorithm is used to calculate the ship encounter thermal value; wherein, the ship encounter thermal value is positively correlated with the ship encounter density and negatively correlated with the minimum encounter distance.

5. The method for identifying hotspot areas and handling accidents on canals according to claim 1, characterized in that, The dynamic allocation of fusion weights to the crossflow risk index, ship encounter thermal value, and lock drive characteristics based on the current lock operating status and tidal phase includes: An adaptive weight regulator is constructed and trained to learn the nonlinear mapping relationship from the combined working conditions of the lock operation state and tidal phase to the optimal weight allocation. During high tide, the output weights of the crossflow risk index and the thermal value encountered by the ship are increased to highlight the collision risk. During the lock discharge period, the output weights of the lock drive characteristics are increased to highlight the bottoming risk. During each feature fusion, the current lock operation status and tidal phase are input into the trained adaptive weight regulator to output the crossflow risk index, the ship encounter thermal value, and the fusion weights corresponding to the lock drive features.

6. The method for identifying hotspot areas for waterborne accidents in canals and handling accidents according to claim 1, characterized in that, The reverse analysis determines the contribution of the cross-current risk index and the ship encounter thermal value at the location of the hotspot cluster, calculates the corresponding gate control state and tidal phase, and generates hotspot cluster attribution results based on the contribution, gate control state, and tidal phase, including: S61, extract the crossflow risk index, ship encounter thermal value, lock drive characteristics and comprehensive accident risk index of each spatiotemporal grid unit within the spatiotemporal window where the hot spot cluster is located; S62 uses a game theory-based attribution decomposition algorithm, taking the cross-current risk index, the ship encounter thermal value, and the lock drive characteristics as participants, calculating the marginal contribution of each participant to the comprehensive accident risk index, and determining the marginal contribution ratio of each participant as their respective contribution degree. S63, the risk type corresponding to the participant with the highest contribution is identified as the dominant risk type of the hotspot cluster; S64, extract the lock operation status and tidal phase of the corresponding spatiotemporal window of the hot spot cluster, and combine them to form the lock control-tidal trigger condition label; S65 combines the dominant risk type with the gate-tidal trigger condition label to generate hotspot cluster attribution results.

7. The method for identifying hotspot areas and handling accidents on canals according to claim 2, characterized in that, The adjustment of the iteration execution time interval based on the location of the hotspot cluster and the corresponding hotspot cluster attribution result includes: S71, obtain the location information of each hotspot cluster generated in this iteration and the corresponding hotspot cluster attribution results; S72, based on location information, extracts the channel attribute labels of the locations of each hotspot cluster; S73, the channel attribute labels are coupled with the dominant risk type in the hotspot cluster attribution results to determine the priority of hazard handling for each hotspot cluster; the coupling determination is performed according to a preset location-risk coupling rule table, which defines the hazard level corresponding to different combinations of channel attribute labels and different dominant risk types; S74, adjust the time interval for the next iteration based on the danger handling priority of each hotspot cluster; wherein, when there is a hotspot cluster whose danger handling priority reaches a preset threshold, the time interval is reduced; when the danger handling priorities of all hotspot clusters are lower than the preset threshold, the time interval is increased.

8. The method for identifying hotspot areas and handling accidents on canals according to claim 1, characterized in that, The accident pre-incident handling plan includes at least a preventive patrol plan, an emergency force pre-deployment plan, and a lock scheduling and traffic flow control plan.

9. A system for identifying hotspot areas of waterway accidents in canals and for handling such accidents, characterized in that, The system is used to implement the canal waterway accident hotspot identification and accident handling method according to any one of claims 1 to 8, the system comprising: The multi-source data fusion module is used to acquire ship navigation data, lock scheduling data and tidal hydrological data. After data cleaning and spatiotemporal alignment, a unified spatiotemporal gridded waterway model is constructed. The feature extraction module is used to extract tidal driving features, ship traffic conflict features, and lock driving features based on the spatiotemporal gridded waterway model. The risk index calculation module is used to determine the crossflow risk index based on the tidal direction and channel location and the lateral velocity component in the tidal driving characteristics; and to determine the ship encounter thermal value based on the ship traffic conflict characteristics. The dynamic weight fusion module is used to dynamically allocate fusion weights to the crossflow risk index, the ship encounter thermal value, and the lock drive characteristics according to the current lock operation status and tidal phase. After nonlinear weighted fusion, the comprehensive accident risk index of each spatiotemporal grid unit is calculated, and a risk-weighted spatiotemporal cube is constructed. The hotspot identification module is used to identify hotspot clusters where the comprehensive accident risk index is concentrated on a risk-weighted spatiotemporal cube, using a spatiotemporal scanning statistical method based on the Poisson distribution model and moving a scanning window of variable size. The attribution analysis module is used to reverse analyze the contribution of the crossflow risk index and the ship encounter thermal value at the location of the hot spot cluster, calculate the corresponding gate control state and tidal phase, and generate the hot spot cluster attribution result based on the contribution, gate control state and tidal phase. The iteration and solution generation module is used to iteratively execute S3 to S6 to update the hotspot clusters and corresponding hotspot cluster attribution results for the target time period, and adjust the time interval of iteration execution based on the location of the hotspot clusters and the corresponding hotspot cluster attribution results; and generate differentiated accident pre-handling solutions based on the updated hotspot clusters and corresponding hotspot cluster attribution results.

Images