A collision danger dynamic correlation identification method under a complex encounter situation

Abstract

The application discloses a collision danger dynamic correlation identification method under a complex encounter situation, and comprises the following steps: a collision danger degree calculation model based on a domain invasion parameter is constructed, collision risk calculation of a target ship and other target ships under a complex scene based on space-time coupling characteristics is realized, dangerous combinations are screened through a space-time double dynamic threshold mechanism, and high-risk ships are screened; an optimal clustering level is determined by quantifying a dynamic balance relationship between in-cluster compactness and inter-cluster separation, an optimal level number is determined, and division of a cluster with similar danger degrees is realized; information entropy weight calculation of a comprehensive collision danger index is carried out based on an entropy weight method, the ships after cluster division are prioritized for collision avoidance, the ship is operated according to the collision avoidance priority based on an improved speed obstacle method, and the ship returns to the original track navigation after operation according to the collision avoidance priority based on a re-encounter point re-navigation time.

Description

Technical Field

[0001] This invention belongs to the field of maritime traffic risk analysis technology, and relates to a method for dynamic correlation identification of collision hazards under complex encounter situations. Background Technology

[0002] Autonomous vessels, through intelligent decision-making systems that replace manual operation, can systematically eliminate potential navigational accidents caused by human factors, and improve the accuracy of collision risk prevention and control to the sub-second response level.

[0003] Collision risk analysis is a core functional module of autonomous navigation systems, and its technical characteristics directly determine the safety, reliability, and intelligence level of these systems. With the International Maritime Organization's (IMO) advancement of ship autonomy classification standards (from decision support to fully autonomous navigation), collision risk analysis technology has evolved from single-parameter judgment to a multi-dimensional, multi-modal intelligent decision-making system.

[0004] Collision hazard identification, as a core decision-making module of autonomous ship navigation systems, directly defines the safety benchmarks and cognitive dimensions of intelligent navigation systems through its technological evolution. With the deepening implementation of the International Maritime Organization's (IMO) autonomous ship classification standards, this field has broken through the traditional single-parameter threshold judgment of DCPA / TCPA, forming a three-dimensional technical architecture of multimodal data fusion, dynamic risk field prediction, and hybrid enhanced intelligent decision-making. By integrating hydrological and meteorological factors, ship maneuvering characteristics, and complex navigation environment elements, a closed-loop intelligent system of "environmental cognition - situational analysis - game theory decision-making" has been constructed, achieving a leapfrog development from static geometric analysis to intelligent perception of navigational situations. This breakthrough in cognitive computing architecture not only reshapes the safety boundaries of autonomous ship navigation but also becomes a core engine driving the reconstruction of the global shipping technology ecosystem, laying a key technological foundation for the era of intelligent shipping.

[0005] The ship collision hazard analysis technology system has formed an innovative four-dimensional collaborative architecture: In the classical geometric modeling dimension, a dynamic safety assessment model with DCPA (Distance to Closest Encounter) / TCPA (Time to Closest Encounter) as core parameters has been constructed. By introducing ship domain theory, a graded safety threshold has been established, and an improved nearest encounter distance algorithm considering ship maneuvering characteristics has been developed. In the multi-source information fusion dimension, the traditional sensor data barrier has been broken through, integrating multi-modal perception data such as radar, AIS, cameras, and sonar. An extended Kalman filter algorithm is used to achieve spatiotemporal alignment and accuracy compensation of multi-source data. In port and waterway scenarios, 5G+IoT technology is used to achieve real-time sharing of ship-shore perception data. In the intelligent cognitive computing dimension, deep reinforcement learning is used to mine implicit risk features in historical navigation data. Combined with fuzzy mathematics theory, a collision hazard quantification assessment function is constructed to achieve continuous assessment of risk levels. In the rule-based reasoning dimension, a hybrid reasoning engine based on COLREGs rules has been innovatively developed. International collision avoidance rules are encoded into a production rule base. Combined with expert experience, a knowledge graph is constructed. Through semantic network analysis of rule conflict scenarios, decision suggestions that conform to navigation practice are formed. The organic integration of these four technologies signifies that ship collision hazard analysis is moving from traditional threshold judgment to cognitive intelligent decision-making, building a three-dimensional safety protection system for intelligent navigation systems.

[0006] Current ship collision hazard analysis technology faces three major technical bottlenecks in complex encounter scenarios, which seriously restricts the improvement of navigation safety.

[0007] (1) The traditional analysis framework focuses too much on the numerical calculation of collision probability and has not yet established the mechanism of the role of hazard correlation in the optimization of decision-making models, resulting in an exponential increase in the decision-making complexity of multi-ship cooperative avoidance.

[0008] (2) In terms of risk quantification, although tools such as geometric probability index (DCPA / TCPA), ship domain model and speed barrier method (VO) have achieved single ship risk assessment, they generally lack the ability to dynamically analyze the topology of multi-ship conflict networks and have not constructed a criterion for the correlation between conflict targets.

[0009] (3) Existing methods fail to establish a time-series coordination mechanism for ship avoidance actions in complex encounter scenarios, making it difficult to achieve spatiotemporal decoupling of maneuvering response and conflict resolution of action sequences, resulting in a significant decrease in collision avoidance decision-making efficiency in complex encounter situations. Summary of the Invention

[0010] To address the aforementioned problems, the technical solution adopted by this invention is: a method for dynamic association and identification of collision hazards under complex encounter situations, comprising the following steps:

[0011] Based on the dual parameters DDV and TDV of the spatial intrusion characteristics of the ship domain, a three-dimensional evaluation framework is constructed through spatial intrusion state, time dimension prediction and quantitative dimension characterization. Combined with the asymmetric Gaussian distribution fitting algorithm, a collision risk calculation model based on domain intrusion parameters is constructed to realize the calculation of collision risk between the ship and other target ships in complex scenarios based on spatiotemporal coupling characteristics.

[0012] The collision risk assessment model evaluates the collision risk between the vessel and other target vessels in complex scenarios based on spatiotemporal coupling characteristics. It then uses a dual dynamic threshold mechanism of space and time to screen for dangerous combinations and identify high-risk vessels.

[0013] For high-risk vessels after screening, an improved hierarchical clustering algorithm is used to integrate the time collision risk index into the improved profile coefficient. The optimal clustering level is determined by quantifying the dynamic balance between intra-cluster compactness and inter-cluster separation, and the optimal number of levels is determined to achieve the division of clusters with similar time collision risk.

[0014] The information entropy weights of the comprehensive collision hazard index are calculated using the entropy weight method, and the collision avoidance priorities of ships after being grouped into clusters are ranked.

[0015] Based on the improved speed barrier method, this vessel can operate according to collision avoidance priorities.

[0016] Based on the timing of the resumption of navigation at the nearest encounter point, the vessel will return to its original trajectory after performing operations according to the collision avoidance priority.

[0017] Furthermore, the collision hazard calculation model based on the domain intrusion parameter includes the following:

[0018] Based on the circular ship domain model, the intrusion value (DDV) of a target ship into the security domain of the ship is derived and the DDV calculation formula is as follows:

[0019] DDV = max(1 - DCPA / D) ij ,0) (1)

[0020] Among them, D ij Radius of the ship's domain;

[0021] The membership function r is established based on the Gaussian equation. s for:

[0022]

[0023] Where a represents the peak collision risk level, b represents the peak location, and c represents the variance, which controls the width of the Gaussian curve;

[0024] For any two-ship encounter situation, the Time to Intrude (TDV) remaining in the domain invasion is calculated as follows:

[0025]

[0026] Among them, V R This represents the relative speed of ship j relative to ship i;

[0027] The time-based collision hazard set is R u Membership function r u for:

[0028]

[0029] in, This indicates the time it takes for the two ships to reach the applicable distance in the situation.

[0030] Furthermore, the process of screening high-risk vessels through a dual spatial and temporal dynamic threshold mechanism for selecting hazardous combinations is as follows:

[0031] When the spatial collision risk level between the target ship and the ship meets r s ≥r0, where r0 is a preset safety threshold, and the time-based collision risk level reaches r t When ≥r0+Δr, the ship pair (i,j) will be automatically identified as a high-risk combination;

[0032]

[0033] Where Δr is a dynamically adjusted parameter, satisfying And the set of ship pair sequences where i ≠ j is P, satisfying The set of ship pair sequences where i,j∈P and i≠j is P';

[0034] The degree of dispersion of encountering ships is quantified by time similarity. The time similarity function is:

[0035]

[0036] in: For the time dimension, it represents the membership function of collision risk.

[0037] A quantitative model of the similarity of the motion of encountering ships is constructed based on the collision risk membership function in the time dimension.

[0038] Construct a motion similarity quantification matrix for meeting ship pairs to determine the degree of temporal connection:

[0039]

[0040] M simThe larger the value, the more similar the risk characteristics of the two ships in terms of time urgency. The time collision risk correlation is calculated only when the ship pair (i,j) belongs to the high-risk set P'. Otherwise, it is directly assigned a value of 0, indicating that there is no significant conflict correlation between the two ships.

[0041] Furthermore, the improved hierarchical clustering algorithm includes the following process:

[0042] First, each ship in P' is considered as a cluster (C1, C2, C3, ..., C N Initial cluster state C 0 The description is as follows:

[0043] C 0 =[C1,C2,C3,…,C N (27)

[0044] The calculation of distance between clusters is expressed as follows:

[0045]

[0046] Where C m and C n Representing clusters m and n respectively; |C m |and|C n | represents cluster C m and C n The number of ships in the country It is the sum of the time-based collision risk of the corresponding ships in the two clusters;

[0047] During the k-th iteration, a global search is performed on the current danger matrix to determine the maximum inter-group distance:

[0048] [p,q]={p,q|L avg (C p C q )=maxL avg p,q∈N k-1 p≠q,N k-1 =N-2*(k-1)} (29)

[0049] Where [p, q] is the index with the maximum inter-cluster distance, and N k-1 It is the number of clusters in the (k-1)th iteration. The corresponding clusters are merged to form a new cluster.

[0050]

[0051] in, It is a merged cluster;

[0052] The cluster representation updated in the k-th iteration is as follows:

[0053]

[0054] Synchronize and update the similarity matrix in the (k-1)th iteration;

[0055]

[0056] When the number of remaining objects meets the preset termination condition, the iteration process will terminate. At this point, the remaining objects have met the condition of being uniquely assigned to the corresponding category or do not need to be merged further, ensuring the determinism of the classification results and the efficiency of the algorithm.

[0057] |N k |≤2 (33) Further: The improved contour coefficient calculation formula is as follows:

[0058]

[0059] Among them, C m It is a cluster containing ship i, C n It is related to cluster C m Different clusters, a(i) represents ship i in other clusters C m The average time-based collision risk between ships, b(i) is the risk of ship i in relation to other clusters, excluding C. m In addition, the minimum average time collision risk between them;

[0060] For a cluster containing only a single ship (i.e., |C) i |=1), the profile coefficient is set to zero;

[0061]

[0062] For the candidate cluster set generated by hierarchical clustering Calculate the weighted average of the improved S under this candidate hierarchy:

[0063]

[0064] After calculating the weighted improved profile coefficients for all candidate levels, the optimal number of levels is determined.

[0065]

[0066] Furthermore: The comprehensive collision hazard index expression is as follows:

[0067] r = w s ·r s +w t ·r t (40)

[0068] Where r is the comprehensive collision hazard index, w s wt The corresponding spatial collision hazard level r s Time-based collision risk level r t The weight value;

[0069] Since the collision risk index is a negative indicator, it is processed using a negative indicator standardization method during the data standardization phase.

[0070]

[0071] Among them, w1 and w2 are dynamically determined by the entropy weight method;

[0072] Establish a proportional matrix based on standardized data;

[0073]

[0074] The expression for the information entropy of the indicator is as follows:

[0075]

[0076] The final weight expression for the indicator is obtained as follows:

[0077]

[0078] Furthermore: the process of enabling the vessel to operate according to collision avoidance priorities based on the improved speed barrier method is as follows:

[0079] The relative velocity vectors of the two ships, defined from point P O Along V OT The emitted rays are represented as:

[0080] λ(P O (t),V OT (t))={P O (t)+V OT (t)*t|t≥0} (27)

[0081] Where, φ O (t), P O (t), V O (t)}、φ O (t), P O (t) and V O (t) represent the heading, position, and speed of the ship's OS at time t, respectively. T (t), P T (t), V T (t)}、φ T (t), P T (t) and V T (t) represents the heading, position, and speed of the other ship at time t;

[0082] When it falls into two tangent lines and When the area between the target and target ships is considered to pose a collision risk, the set of relative velocities V that would cause a collision between the two ships is defined. OT RCC (Relative Collision Zone) OT (Relative collision zone),RCC OT Represented as:

[0083]

[0084] When multiple obstacles exist, the RCC of each obstacle needs to be converted to a unified velocity space representation, and the relative velocity space needs to be converted to the ship's absolute velocity space. Therefore, the speed obstacles of the ship that will collide with the target ship are:

[0085]

[0086] in The Minkowski operation, which moves the relative collision cone (RCC) based on the velocity vector of the other vessel, is used to perform the operation. OT The ship's velocity vector cannot be in the V position. OS|TS regional;

[0087] Considering the actual maneuvering characteristics of the giving-way vessel, and based on navigational practice, the course change range for the giving-way vessel to take evasive action is [-π / 3, π / 3]. If this vessel is the giving-way vessel, the actual permissible course range (ACR) is:

[0088]

[0089] in, and These represent the upper and lower limits of the ship's heading angle in ACR, respectively;

[0090] The course obstacle zone relative to other vessels is defined as follows:

[0091]

[0092] in, and Corresponding to tangents respectively and The left and right directional boundaries.

[0093] Similarly, compared to COR OT The safe course region (SCR) of this vessel relative to other vessels is defined as follows:

[0094]

[0095] The navigable area for ships is represented as follows:

[0096]

[0097] Furthermore, the expression for the resumption timing of the nearest encounter point is as follows:

[0098] t r =TCPA+n*t s (34)

[0099] Among them, t r The timing for the ship to resume navigation, where n is a positive integer set by the navigator; t s For time step.

[0100] A method for dynamic association and identification of collision hazards under complex encounter situations includes:

[0101] A collision risk calculation model module based on domain intrusion parameters is constructed: It is used to construct a collision risk calculation model based on domain intrusion parameters through a three-dimensional evaluation framework of spatial intrusion state, time dimension prediction and quantitative dimension characterization, combined with an asymmetric Gaussian distribution fitting algorithm, so as to realize the collision risk calculation of the ship and other target ships in complex scenarios based on spatiotemporal coupling characteristics.

[0102] High-risk vessel screening module: This module is used to screen high-risk vessels by evaluating the collision risk of the vessel and other target vessels in complex scenarios based on the spatiotemporal coupling characteristics, as assessed by the collision risk assessment model. It uses a dual dynamic threshold mechanism of space and time to screen dangerous combinations.

[0103] Hazard cluster segmentation module: This module is used to segment high-risk vessels after screening. Based on an improved hierarchical clustering algorithm, it incorporates the time-collision hazard index into the improved profile coefficient. By quantifying the dynamic balance between intra-cluster compactness and inter-cluster separation, it determines the optimal clustering level and the optimal number of levels, thereby enabling the segmentation of clusters with similar hazard levels.

[0104] Collision avoidance priority ranking module: used to calculate the information entropy weight of the comprehensive collision hazard index based on the entropy weight method, and to rank the collision avoidance priority of ships after they are divided into clusters;

[0105] Operation Module I: Used to enable the vessel to operate according to collision avoidance priorities based on the improved speed barrier method;

[0106] Operation Module II: Used to enable the vessel to return to its original trajectory after performing operations according to the collision avoidance priority, based on the re-entry timing of the nearest encounter point.

[0107] A computer device includes: a processor and a memory, the memory storing a program module that runs on the processor to implement any of the methods described above.

[0108] This invention provides a dynamic correlation identification method for collision hazards in complex encounter situations. Addressing the critical aspect of collision hazard assessment in conflict analysis, this invention comprehensively considers the degree of vessel intrusion and the remaining time of intrusion, integrating the four-stage collision avoidance theory, a circular vessel domain model, and Gaussian fitting to construct a collision hazard calculation model based on domain intrusion parameters, significantly improving the accuracy of collision hazard identification. Furthermore, addressing the shortcomings of existing methods in the dynamic analysis of multi-vessel conflict network topology and the determination of conflict target correlation, this invention employs methods such as average linking and contour coefficients to achieve adaptive judgment of encounter vessel clustering levels. In addition, based on the entropy weight calculation method, an entropy weight calculation model for collision hazard index information is constructed to scientifically divide the vessel avoidance priorities within the same cluster, effectively improving the efficiency of collision avoidance decision-making and the collaborative capability for maritime traffic safety in complex multi-vessel scenarios.

[0109] Compared with the prior art, the present invention has the following beneficial effects:

[0110] This invention focuses on the collision risk assessment problem in multi-ship encounter conflict analysis and constructs a real-time calculation model based on domain intrusion parameters. This model integrates the four-stage collision avoidance theory, the circular ship domain model, and Gaussian fitting techniques. By quantifying the degree of domain intrusion and remaining time, it achieves a dynamic and refined assessment of collision risk, significantly improving the accuracy of hazard identification and the reliability of early warning in complex situations.

[0111] To address the shortcomings of traditional methods in the dynamic analysis and correlation determination of multi-ship conflict networks, this invention integrates the average link algorithm and the silhouette coefficient method to construct an adaptive clustering hierarchy determination mechanism. This mechanism quantifies the inter-cluster connection strength and separation degree to achieve dynamic analysis of the conflict network topology and adaptive determination of the optimal hierarchy, significantly improving the accuracy and computational efficiency of target correlation identification.

[0112] This invention constructs a collision hazard index weight calculation model based on information entropy theory. It dynamically quantifies the information utility value of the index through the entropy weight method and combines conflict level and motion situation data to effectively solve the limitations of the traditional subjective weighting method and achieve objective division of ship avoidance priorities within the same cluster. Attached Figure Description

[0113] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the 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 based on these drawings without creative effort.

[0114] Figure 1 It is a ship collision avoidance parameter identifier;

[0115] Figure 2 It is average link hierarchy clustering;

[0116] Figure 3 It is a comparison of the order of avoidance actions;

[0117] Figure 4 It is a navigable area calculation model;

[0118] Figure 5 This is the first round of clustering phylogenetic chart;

[0119] Figure 6 It is an improved profile coefficient diagram;

[0120] Figure 7 This is the second round of clustering phylogenetic chart;

[0121] Figure 8 The diagram shows the relative motion trajectories of each give-way vessel to the other vessels: (a) Spatiotemporal characteristics of OS relative motion trajectory, (b) Spatiotemporal characteristics of TS1 relative motion trajectory, (c) Spatiotemporal characteristics of TS5 relative motion trajectory, (d) Spatiotemporal characteristics of TS4 relative motion trajectory, (e) Spatiotemporal characteristics of TS2 relative motion trajectory, and (f) Spatiotemporal characteristics of TS7 relative motion trajectory.

[0122] Figure 9 These are the relative distance change curves: (a) OS relative distance time series evolution pattern, (b) TS1 relative distance time series evolution pattern, (c) TS5 relative distance time series evolution pattern, (d) T4S relative distance time series evolution pattern, (e) TS2 relative distance time series evolution pattern, and (f) TS7 relative distance time series evolution pattern.

[0123] Figure 10 These are curves showing the relative heading changes: (a) the dynamic change pattern of OS relative to heading, (b) the dynamic change pattern of TS1 relative to heading, (c) the dynamic change pattern of TS5 relative to heading, (d) the dynamic change pattern of TS4 relative to heading, (e) the dynamic change pattern of TS2 relative to heading, and (f) the dynamic change pattern of TS7 relative to heading.

[0124] Figure 11The avoidance effects of two hazard identification methods are shown: (a) initial situation setting for 10 ships, and (b) initial situation setting for encountering 8 ships.

[0125] Figure 12 This is clustering and decision-making process optimization. (a) Clustering results for 10 ship encounter scenarios; (b) Clustering results for 8 ship encounter scenarios. Detailed Implementation

[0126] It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0127] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the present invention or its application or use. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0128] A method for dynamic association and identification of collision hazards under complex encounter situations includes the following steps:

[0129] S1: Based on the dual parameters DDV and TDV of the spatial intrusion characteristics of the ship domain, a three-dimensional evaluation framework is constructed through spatial intrusion state, time dimension prediction and quantitative dimension characterization. Combined with the asymmetric Gaussian distribution fitting algorithm, a collision risk calculation model based on domain intrusion parameters is constructed to realize the calculation of collision risk between the ship and other target ships in complex scenarios based on spatiotemporal coupling characteristics.

[0130] S2: The collision risk assessment model evaluates the collision risk between the vessel and other target vessels in complex scenarios based on spatiotemporal coupling characteristics. The model then uses a dual dynamic threshold mechanism of space and time to screen for dangerous combinations and select high-risk vessels.

[0131] S3: For high-risk vessels after screening, based on the improved hierarchical clustering algorithm, the time collision risk index is integrated into the improved profile coefficient. The optimal clustering level is determined by quantifying the dynamic balance between intra-cluster compactness and inter-cluster separation, and the optimal number of levels is determined to achieve the division of clusters with similar risk levels.

[0132] S4: Calculate the information entropy weights of the comprehensive collision hazard index based on the entropy weight method, and rank the collision avoidance priorities of ships after they are grouped.

[0133] S5: Based on the improved speed barrier method, the vessel will operate according to the collision avoidance priority.

[0134] S6: Based on the re-entry timing of the nearest encounter point, after the vessel has performed operations according to the collision avoidance priority, it will return to its original trajectory.

[0135] The steps S1 / S2 / S3 / S4 / S5 / S6 are executed sequentially;

[0136] Based on the dynamic evolution of ship collision risk, the traditional collision avoidance phase can be subdivided into four progressive stages, with the legal effect of avoidance liability commencing in the second stage. At this point, the collision hazard assessment system shifts from the traditional DCPA / TCPA parameter system to a dual-parameter assessment model based on the spatial intrusion characteristics of the ship's domain: DDV (Degree of Domain Violation) / TDV (Time to Domain Violation). This model, through a three-dimensional assessment framework (spatial intrusion state, time-dimensional prediction, and quantitative dimension representation), constructs a collision risk quantification system more aligned with maritime practice. Combined with an asymmetric Gaussian distribution fitting algorithm, it achieves accurate capture of the spatiotemporal coupling characteristics of collision risk in complex scenarios, providing a more practical assessment basis for intelligent collision avoidance decision-making systems.

[0137] The collision risk calculation model based on domain intrusion parameters includes the following:

[0138] Based on a circular ship domain model, the intrusion level (DDV) of a target ship into the ship's security domain is derived and quantified to reflect the interaction between the ship's real-time motion state and its surrounding environment. Figure 1 As shown, for any pair of meeting vessels, the DDV calculation formula is as follows:

[0139] DDV = max(1 - DCPA / D) ij ,0) (1)

[0140] Among them, D ij Radius of the ship's domain;

[0141] The membership function r is established based on the Gaussian equation. s for:

[0142]

[0143] Where a represents the peak collision risk level, b represents the peak location, and c represents the variance, which controls the width of the Gaussian curve;

[0144] For any two-ship encounter situation, the Time to Intrude (TDV) remaining in the domain invasion is calculated as follows:

[0145]

[0146] Among them, V R This represents the relative speed of ship j relative to ship i;

[0147] The time-based collision hazard set is R u Membership function r u for:

[0148]

[0149] in, This indicates the time it takes for the two ships to reach the applicable distance in the situation.

[0150] The process of screening high-risk vessels by using a dual dynamic threshold mechanism of space and time to filter dangerous combinations is as follows:

[0151] In complex multi-ship encounter scenarios, accurately identifying high-risk collision combinations is a crucial prerequisite for initiating effective collision avoidance decisions. Based on a collision hazard assessment model, a dual dynamic threshold mechanism is used to screen hazardous combinations. When the spatial collision hazard level between the target ship and the ship meets r... s ≥r0 (r0 is the preset safety threshold), and the time-based collision risk reaches r t When the value is greater than or equal to r0 + Δr (Δr is a dynamically adjusted parameter), the system will automatically classify the ship pair as a high-risk combination. This method provides a basis for the dynamic analysis of conflict network topology by quantifying the spatiotemporal coupling characteristics of spatial collision risk.

[0152] When the spatial collision risk level between the target ship and the ship meets r s ≥r0, where r0 is a preset safety threshold, and the time-based collision risk level reaches r t When ≥r0+Δr, the ship pair (i,j) will be automatically identified as a high-risk combination;

[0153]

[0154] Where Δr is a dynamically adjusted parameter, satisfying And the set of ship pair sequences where i ≠ j is P, satisfying r t ij The set of ship pair sequences ≥ r0 + Δr and i,j∈P,i≠j is P'.

[0155] Based on a spatiotemporal urgency screening mechanism, only vessel pairs meeting the risk threshold proceed to the next stage, effectively identifying key target vessels and filtering out safe interactions. Similarity measurement, as the core of the clustering algorithm, quantifies the dispersion of encountering vessels through temporal similarity. The temporal similarity function is:

[0156] d(i,j)=1-r t ij(i,j∈P) (46)

[0157] Where: r t ij For the time dimension, it represents the membership function of collision risk.

[0158] A quantitative model of the similarity of the motion of encountering ships is constructed based on the collision risk membership function in the time dimension.

[0159] Construct a motion similarity quantification matrix for meeting ship pairs to determine the degree of temporal connection:

[0160]

[0161] M sim The larger the value, the more similar the risk characteristics of the two ships in terms of time urgency. The correlation of time collision risk is calculated only when ship pair (i,j) belongs to the high-risk set P'. Otherwise, a value of 0 is directly assigned, indicating that there is no significant conflict correlation between the two ships, which facilitates subsequent cluster analysis.

[0162] For high-risk vessels after screening, an improved hierarchical clustering algorithm is used. The time-based collision hazard index is incorporated into the improved profile coefficient. The optimal clustering level and the optimal number of levels are determined by quantifying the dynamic balance between intra-cluster compactness and inter-cluster separation. The process of dividing clusters with similar hazard levels is as follows:

[0163] Traditional ship collision risk assessment focuses excessively on quantifying risk while neglecting the correlation between multiple ship interactions. While methods such as density-based spatial clustering (DBSCAN) and K-Means clustering are widely used, they suffer from poor adaptability to density changes and sensitivity to initial parameters. This invention proposes an improved hierarchical clustering algorithm, such as... Figure 2 As shown, the algorithm includes four key steps: dynamically averaging link reconstructing inter-cluster distances, identifying and merging extreme ship combinations, updating the similarity matrix, and adaptively setting the termination threshold.

[0164] The improved hierarchical clustering algorithm includes the following process:

[0165] First, each ship in P' is considered as a cluster (C1, C2, C3, ..., C N Initial cluster state C 0 The description is as follows:

[0166] C 0 =[C1,C2,C3,…,C N (48)

[0167] The calculation of distance between clusters is expressed as follows:

[0168]

[0169] Where C m and C n Representing clusters m and n respectively; |C m |and|C n | represents cluster C m and C n The number of ships in the country It is the sum of the time-based collision risk of the corresponding ships in the two clusters;

[0170] During the k-th iteration, a global search is performed on the current danger matrix to determine the maximum inter-group distance:

[0171] [p,q]={p,q|L avg (C p C q )=maxL avg p,q∈N k-1 p≠q,N k-1 =N-2*(k-1)} (50)

[0172] Where [p, q] is the index with the maximum inter-cluster distance, and N k-1 It is the number of clusters in the (k-1)th iteration. The corresponding clusters are merged to form a new cluster.

[0173]

[0174] in, It is a merged cluster;

[0175] Accordingly, the cluster representation updated in the k-th iteration is as follows:

[0176]

[0177] Synchronize and update the similarity matrix in the (k-1)th iteration;

[0178]

[0179] When the number of remaining objects meets the preset termination condition, the iteration process will terminate. At this point, the remaining objects have met the condition of being uniquely assigned to the corresponding category or do not need to be merged further, ensuring the determinism of the classification results and the efficiency of the algorithm.

[0180] |N k |≤2 (54)

[0181] To optimize ship clustering and prevent excessive merging, this study proposes an adaptive termination mechanism based on an improved profile coefficient. Addressing the problem that traditional clustering algorithms easily generate fragmented clusters or excessive merging in multi-ship encounter scenarios, this paper innovatively integrates a time-based collision hazard index into the profile coefficient evaluation system. The optimal clustering level is determined by quantifying the dynamic balance between intra-cluster compactness and inter-cluster separation. The improved profile coefficient S(i) maintains monotonicity using a linear mapping (0 for worst-case, 1 for best-case). For ship i, the improved profile coefficient calculation formula is as follows:

[0182]

[0183] Among them, C m It is a cluster containing ship i, C n It is related to cluster C m Different clusters, a(i) represents ship i in other clusters C m The average time-based collision risk between ships, b(i) is the risk of ship i in relation to other clusters, excluding C. m In addition, the minimum average time collision risk between them;

[0184] For a cluster containing only a single ship (i.e., |C) i |=1), the profile coefficient is set to zero;

[0185]

[0186] For the candidate cluster set generated by hierarchical clustering Calculate the weighted average of the improved S under this candidate hierarchy:

[0187]

[0188] After calculating the weighted improved profile coefficients for all candidate levels, the optimal number of levels is determined.

[0189]

[0190] In complex encounter situations, collision avoidance priority allocation is the core mechanism for collaborative collision avoidance among ships. The key lies not in individual ship intelligent optimization, but in achieving collaborative transformation of group decision-making through a dynamic priority framework. Since the spatiotemporal urgency of each ship varies in different clusters, their spatiotemporal hazard function values ​​also differ. Based on this, an information entropy weight calculation model for a comprehensive collision hazard index is constructed using the entropy weight method, providing theoretical support for priority allocation and ensuring a balance between safety and efficiency in complex encounter scenarios. The expression for the comprehensive collision hazard index is as follows:

[0191] r = w s ·r s +w t ·r t(61)

[0192] Where r is the comprehensive collision hazard index, w s w t The corresponding spatial collision hazard level r s Time-based collision risk level r t The weight value.

[0193] Because the collision risk index is a negative indicator (i.e., the higher the index value, the higher the risk), it is detrimental to overall safety assessment. Therefore, a negative index standardization method is used during the data standardization phase.

[0194]

[0195] Among them, w1 and w2 are dynamically determined by the entropy weight method;

[0196] Establish a proportional matrix based on standardized data;

[0197]

[0198] The expression for the information entropy of the indicator is as follows:

[0199]

[0200] The final weight expression for the indicator is obtained as follows:

[0201]

[0202] To accurately calculate the decision-making timing and path options for ship collision avoidance and rerouting, this invention proposes a real-time optimization method based on discretized dynamic programming. For example... Figure 3 As shown, the entire decision-making process includes five basic steps: (1) multi-dimensional collision risk assessment; (2) adaptive hierarchical clustering based on average links; (3) priority action ranking based on entropy weight method; (4) avoidance scheme optimization based on improved speed barrier method; and (5) determination of rendezvous timing based on the nearest encounter point. The process is repeated iteratively until all ships can safely leave each other.

[0203] To improve the practical application effectiveness of ship collision avoidance decision models, this invention proposes a dynamic extraction strategy for navigable areas. This strategy integrates the Velocity Obstacle Method with good seamanship requirements, transforming the relative motion state of two ships into a velocity vector field through velocity space transformation, and constructing a navigable area boundary model that includes safety distance constraints, ship maneuverability constraints, and steering amplitude constraints.

[0204] like Figure 4 As shown, the process of enabling the vessel to operate according to collision avoidance priorities based on the improved speed barrier method is as follows:

[0205] The relative velocity vectors of the two ships, defined from point P O Along V OT The emitted rays are represented as:

[0206] λ(P O (t),V OT (t))={P O (t)+V OT (t)*t|t≥0} (27)

[0207] Where, φ O (t), P O (t), V O (t)}、φ O (t), P O (t) and V O (t) represent the heading, position, and speed of the ship's OS at time t, respectively. T (t), P T (t), V T (t)}、φ T (t), P T (t) and V T (t) represents the heading, position, and speed of the other ship at time t;

[0208] When it falls into two tangent lines and When the area between the target and target ships is considered to pose a collision risk, the set of relative velocities V that would cause a collision between the two ships is defined. OT RCC (Relative Collision Zone) OT (Relative collision zone),RCC OT Represented as:

[0209]

[0210] When multiple obstacles exist, the RCC of each obstacle needs to be converted to a unified velocity space representation. For practical application, the relative velocity space needs to be converted to the ship's absolute velocity space. Therefore, the speed obstacles of the ship that will collide with the target ship are:

[0211]

[0212] in The Minkowski operation, which moves the relative collision cone (RCC) based on the velocity vector of the other vessel, is used to perform the operation. OT The ship's velocity vector cannot be in the V position. OS|TS regional;

[0213] Considering the actual maneuvering characteristics of the giving-way vessel, and based on navigational practice, the course change range for the giving-way vessel to take evasive action is [-π / 3, π / 3]. If this vessel is the giving-way vessel, the actual allowable course region (ACR) is:

[0214]

[0215] in, and These represent the upper and lower limits of the ship's heading angle in ACR, respectively;

[0216] The course obstacle zone relative to other vessels is defined as follows:

[0217]

[0218] in, and Corresponding to tangents respectively and The left and right directional boundaries.

[0219] Similarly, compared to COR OT The safe course region (SCR) of this vessel relative to other vessels is defined as follows:

[0220]

[0221] To enable a vessel traveling in a straight line to quickly and accurately determine the intention of a yielding vessel to take evasive action, the change in the relative heading angle between the two vessels should be greater than π / 18. This maneuver is considered to conform to the spirit of good seamanship. Therefore, the navigable area for a vessel is represented as follows:

[0222]

[0223] The expression for the timing of the return journey at the nearest meeting point after the yielding vessel has passed is as follows:

[0224] t r =TCPA+n*t s (34)

[0225] Among them, t r The timing for the ship to resume navigation, where n is a positive integer set by the navigator; t s For time step.

[0226] Multi-ship collision avoidance decision-making is essentially a multi-stage dynamic optimization problem. Its decision-making process can be divided into several related stages along the time dimension. Each stage requires an adaptive decision based on the current situation, ultimately forming a complete dynamic decision sequence. To systematically evaluate the impact of ship collision hazard correlation identification on decision-making effectiveness, this study constructs a complex simulation scenario involving uncoordinated avoidance behavior and multi-ship group interaction, focusing on quantitative analysis and visualization verification of multi-ship encounter situations. As shown in Table 1, the simulation is configured with initial state parameters for 10 ships, covering key elements such as position coordinates, speed, heading, and ship territory radius. OS, TS1, and TS2 form an encounter situation with similar intersection angles.

[0227] Table 1

[0228]

[0229] At step 30 of the simulation, TS3 suddenly turns left, a maneuver that could lead to a chain reaction of encounters with the continuously approaching TS2 and TS4-TS8. Simultaneously, TS6, traveling from south to north, may also encounter OS, TS4, and TS5 in a secondary crossover scenario. By comparing the differences in decision-making effectiveness of different correlation analysis methods, the system outputs two types of simulation results: a quantitative evaluation module, covering ship group relationship map construction, improved profile coefficient calculation, and avoidance priority ranking; and a visualization analysis module, including multi-dimensional situational awareness elements such as relative motion trajectory spatiotemporal feature extraction, relative distance temporal evolution pattern recognition, and analysis of relative heading dynamic change patterns. This system constructs a closed-loop process covering collision hazard analysis to decision-making effectiveness evaluation, systematically verifying the core value of collision hazard correlation identification in complex encounter scenarios, and providing theoretical support for improving the real-time decision-making capabilities of intelligent navigation systems.

[0230] To fully demonstrate the feasibility and practicality of the proposed collision hazard correlation identification method, a collision hazard analysis was first conducted on a multi-ship encounter scenario. The collision hazard correlation results are shown in Table 2.

[0231] Table 2

[0232]

[0233] At the 35th time step, OS and TS1 meet the preset threshold conditions for crossover (rs>0.5, rt>0.5). By calculating the similarity matrix and inter-cluster distance, a similarity matrix is ​​constructed as follows: Figure 5 The genealogical diagram shown visually presents the hierarchical structure of ship clustering and encounter situations. By calculating the profile coefficient, ships are divided into two major clusters, such as... Figure 6As shown, based on the clustering results, the proportion matrix and index weights are calculated to determine the avoidance action sequence corresponding to each cluster. The priority order for ship avoidance in the first cluster is determined to be OS>TS1>TS2, and the priority order in the second cluster is TS5>TS4>TS7. When the simulation reaches the 81st time step, the collision detection module identifies new collision hazard combinations between TS2 and TS3, and TS7 and TS8. The similarity matrix and inter-cluster distances are recalculated, and an updated phylogenetic tree is constructed as shown. Figure 7 As shown, the process of ship clustering and the hierarchical relationship between ships are presented. Similarly, the second clustering result also contains two large clusters: cluster 1 contains TS2 and TS3, and cluster 2 contains TS7 and TS8. Since each cluster contains only two ships, the action sequence is determined according to the International Regulations for Preventing Collisions at Sea. It should be noted that in cluster 1, TS2 is given the obligation to give way because TS3 took uncoordinated actions that disrupted the original cross formation and triggered the automatic collision avoidance agreement. In cluster 2, TS7 is given higher priority according to the cross encounter clause.

[0234] Based on the collision avoidance priority classification results in the conflict analysis, the overall action plan for each vessel is determined using a navigable area calculation model. Simulation results include optimized trajectories, relative motion trajectories, and relative heading distance variation curves, such as... Figure 8-10 As shown, from Figure 8 The given motion trajectories of each give-way vessel relative to other vessels are: (a) Spatiotemporal characteristics of OS relative motion trajectory, (b) Spatiotemporal characteristics of TS1 relative motion trajectory, (c) Spatiotemporal characteristics of TS5 relative motion trajectory, (d) Spatiotemporal characteristics of TS4 relative motion trajectory, (e) Spatiotemporal characteristics of TS2 relative motion trajectory, and (f) Spatiotemporal characteristics of TS7 relative motion trajectory.

[0235] It can be seen that the yielding vessel's avoidance action includes the entire avoidance process, including both the avoidance itself and the resumption of navigation.

[0236] at the same time, Figure 9 These are the relative distance change curves: (a) OS relative distance time series evolution pattern, (b) TS1 relative distance time series evolution pattern, (c) TS5 relative distance time series evolution pattern, (d) T4S relative distance time series evolution pattern, (e) TS2 relative distance time series evolution pattern, and (f) TS7 relative distance time series evolution pattern.

[0237] from Figure 9 The given relative distance change curves show that the giving-way vessel did not infringe on the other vessel's domain during the entire avoidance process, meaning that the minimum encounter distance was greater than the corresponding domain radius, thus meeting the safety requirements.

[0238] Figure 10These are curves showing the relative heading changes: (a) the dynamic change pattern of OS relative to heading, (b) the dynamic change pattern of TS1 relative to heading, (c) the dynamic change pattern of TS5 relative to heading, (d) the dynamic change pattern of TS4 relative to heading, (e) the dynamic change pattern of TS2 relative to heading, and (f) the dynamic change pattern of TS7 relative to heading.

[0239] In addition, from Figure 10 The relative bearing change curves show that the relative heading changes of each yielding vessel are greater than π / 18, meeting the large-angle avoidance requirements of the International Regulations for Preventing Collisions at Sea (IRCS). Therefore, the decision-making method proposed in this patent can quickly and accurately generate cooperative collision avoidance decision schemes that meet the requirements of the IRCS.

[0240] To verify the impact of hazard association identification on decision-making effectiveness, this study compared the avoidance performance of two hazard identification methods in encounter scenarios involving 8 and 10 ships. Figure 11 ), Figure 11 The collision avoidance effects of two hazard identification methods are compared: (a) initial situation settings for 10 ships, and (b) initial situation settings for 8 ships encountering each other; the parameter configurations of the initial eight ships are kept consistent. The first round of clustering results shows that the dynamic clustering mechanism has significant advantages in collision avoidance decision-making.

[0241] Figure 12 This is clustering and decision-making process optimization. (a) Clustering results for 10 ship encounter scenarios; (b) Clustering results for 8 ship encounter scenarios.

[0242] (1) The first cluster retains only OS, TS1 and TS2. Thanks to the adaptive adjustment of the stopping threshold by the profile coefficient method, the decision complexity and avoidance time span are reduced, thereby reducing the probability of abnormal optimization behavior.

[0243] (2) The second cluster retains the composition of TS5, TS4 and TS7, but the action sequence is accurately sorted by the entropy weight method (TS5>TS4>TS7 replaces TS4>TS5>TS7). This method ensures the accuracy of the action sequence within the cluster by standardizing the collision risk index and assigning reasonable weights.

[0244] The improvement in decision-making performance is shown in Table 3:

[0245] Table 3

[0246]

[0247] (1) Timeliness: The average start time of the first round of collision avoidance actions was reduced from 61.4 steps to 38.0 steps (an improvement of 38.2%), and the parallel decision-making of small-scale clusters eliminated action delays, and the response speed of key vessels (such as OS) was improved by 34%; the average navigation recovery time was reduced from 231.2 steps to 137.3 steps (a reduction of 41.0%), which effectively alleviated route interruption and reduced fuel consumption.

[0248] (2) Accuracy: The dynamic hazard reassessment mechanism monitors the state transition of independent clusters in real time. For example, TS3 was initially marked as an independent cluster, but was reorganized into the TS2-TS3 conflict cluster in step 81 based on real-time risk assessment, reducing the collision avoidance time from 169.8 steps to 99.25 steps (a reduction of 41.5%).

[0249] Improved computing architecture;

[0250] This application adopts a progressive load distribution approach instead of the traditional global optimization framework, reducing the maximum cluster size in the first round from five ships to three ships. It also avoids the computational bottleneck of serial decision-making in large-scale clusters through a phased strategy, significantly improving the system's real-time performance and scalability.

[0251] A method for dynamic association and identification of collision hazards under complex encounter situations includes:

[0252] A collision risk calculation model module based on domain intrusion parameters is constructed: It is used to construct a collision risk calculation model based on domain intrusion parameters through a three-dimensional evaluation framework of spatial intrusion state, time dimension prediction and quantitative dimension characterization, combined with an asymmetric Gaussian distribution fitting algorithm, so as to realize the collision risk calculation of the ship and other target ships in complex scenarios based on spatiotemporal coupling characteristics.

[0253] High-risk vessel screening module: This module is used to screen high-risk vessels by evaluating the collision risk of the vessel and other target vessels in complex scenarios based on the spatiotemporal coupling characteristics, as assessed by the collision risk assessment model. It uses a dual dynamic threshold mechanism of space and time to screen dangerous combinations.

[0254] Hazard cluster segmentation module: This module is used to segment high-risk vessels after screening. Based on an improved hierarchical clustering algorithm, it incorporates the time-collision hazard index into the improved profile coefficient. By quantifying the dynamic balance between intra-cluster compactness and inter-cluster separation, it determines the optimal clustering level and the optimal number of levels, thereby enabling the segmentation of clusters with similar hazard levels.

[0255] Collision avoidance priority ranking module: used to calculate the information entropy weight of the comprehensive collision hazard index based on the entropy weight method, and to rank the collision avoidance priority of ships after they are divided into clusters;

[0256] Operation Module I: Used to enable the vessel to operate according to collision avoidance priorities based on the improved speed barrier method;

[0257] Operation Module II: Used to enable the vessel to return to its original trajectory after performing operations according to the collision avoidance priority, based on the re-entry timing of the nearest encounter point.

[0258] A computer device includes: a processor and a memory, the memory storing a program module, characterized in that the program module runs on the processor to implement the method as described in any one of the claims.

[0259] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for dynamic association and identification of collision hazards under complex encounter situations, characterized in that: Includes the following steps: Based on the dual parameters DDV and TDV of the spatial intrusion characteristics of the ship domain, a three-dimensional evaluation framework is constructed through spatial intrusion state, time dimension prediction and quantitative dimension characterization. Combined with the asymmetric Gaussian distribution fitting algorithm, a collision risk calculation model based on domain intrusion parameters is constructed to realize the calculation of collision risk between the ship and other target ships in complex scenarios based on spatiotemporal coupling characteristics. The collision risk assessment model evaluates the collision risk of the vessel and other target vessels in complex scenarios based on spatiotemporal coupling characteristics. A dual dynamic threshold mechanism of space and time is used to screen dangerous combinations and identify high-risk vessels. For high-risk vessels after screening, an improved hierarchical clustering algorithm is used to integrate the time collision risk index into the improved profile coefficient. The optimal clustering level is determined by quantifying the dynamic balance between the compactness within the cluster and the separation between the clusters, and the optimal number of levels is determined to achieve the division of clusters with similar time collision risks. The information entropy weights of the comprehensive collision hazard index are calculated using the entropy weight method, and the collision avoidance priorities of ships after being grouped into clusters are ranked. Based on the improved speed barrier method, this vessel can operate according to collision avoidance priorities. Based on the timing of the resumption of navigation at the nearest encounter point, the vessel will return to its original course after performing operations according to the collision avoidance priority. The process of screening high-risk vessels by using a dual dynamic threshold mechanism of space and time to filter hazardous combinations is as follows: When the spatial collision risk level between the target ship and this ship meets the requirements r s ≥ r 0, r 0 represents the preset safety threshold, and the time-based collision risk level reaches [a certain threshold]. r t ≥ r 0+Δ r At that time, the ship will be automatically calibrated to ( i,j) This is a high-risk portfolio; Where Δ r To dynamically adjust parameters and meet the requirements and The set of ship pair sequences is ,satisfy and The set of ship pair sequences is ; The improved hierarchical clustering algorithm includes the following process: First, each ship in P' is considered a cluster ( Initial cluster state The description is as follows: The calculation of the average time collision risk among ships in a group is expressed as follows: in, C m and C n Representing clusters m and n ;| C m | and | C n | represent clusters respectively C m and C n The number of ships in the country; L avg ( C m , C n This represents the average time-based collision risk for all cross-cluster vessel pairs between two clusters. in, i , j Clusters C m , C n The ship index within, d ( i , j ) indicates a ship i With ships j The degree of dispersion of the risk of collision over time; In the k During the next iteration, a global search is performed on the current hazard matrix to determine the maximum average time collision hazard between ships in the cluster: in, [ p , q [ ] is an index with the highest average time collision risk between ship pairs in the cluster. N k-1 It is the cluster number in the (k-1)th iteration. The corresponding clusters are merged to form a new cluster. Among them, C l k Let C represent the new cluster formed in the k-th iteration. p k-1 With C q k-1 Let p and q represent the two clusters selected for merging in the (k-1)th iteration; No. k The cluster representation updated in the next iteration is as follows: Synchronize and update the similarity matrix in the (k-1)th iteration; The iteration process will terminate when the number of remaining objects meets the preset termination condition. At this point, the remaining objects have met the condition of being uniquely assigned to the corresponding category or do not require further merging, ensuring the determinism of the classification results and the efficiency of the algorithm. The improved formula for calculating the profile coefficient is as follows: Among them, C m It is a cluster containing ship i, C n It is related to cluster C m Different clusters, a ( i ) is a ship i With cluster C m The average time-based collision risk among other vessels in the vessel. b ( i ) is a ship i In addition to other clusters, C m In addition, the minimum average time collision risk between them; For a cluster containing only a single vessel, i.e. | C i |=1, profile coefficient is set to zero; For the candidate cluster set generated by hierarchical clustering Calculate the improved [structure] under this candidate hierarchy. S Weighted average: After calculating the weighted improved profile coefficients for all candidate levels, determine the optimal number of levels.

2. The method for dynamic association and identification of collision hazards under complex encounter situations according to claim 1, characterized in that: The collision risk calculation model based on domain intrusion parameters includes the following: Based on the circular ship domain model, the intrusion value (DDV) of a target ship into the security domain of the ship is derived and the formula for DDV calculation is as follows: (1) Among them, D ij The radius of the ship's domain; the spatial collision hazard r is established based on the Gaussian equation. s The membership function, with DDV as input, is used to normalize the degree of intrusion into the ship's domain space, resulting in the space collision hazard level r. s The greater the intrusion of the target vessel into the vessel's domain, the higher the risk of a spatial collision. s The higher; For any two-ship encounter situation, the Time to Intrude (TDV) calculation formula is as follows: in, TDV This indicates the remaining time from the current moment until the target vessel first reaches the boundary of this vessel's territory; TCPA Indicates the remaining time for the two ships to reach the nearest meeting point; DCPA Indicates the closest possible meeting distance between the two ships; D ij Indicates the radius of the ship's domain; V R This represents the relative speed of ship j relative to ship i; Membership function of time collision hazard r t for: in, The time it takes for the two ships to reach the applicable distance in the situation is represented by 'a', which represents the peak value of the collision risk over time, 'b' represents the peak position, and 'c' represents the variance, which controls the width of the Gaussian curve.

3. The method for dynamic association and identification of collision hazards under complex encounter situations according to claim 1, characterized in that: The degree of dispersion of meeting ships is quantified by using a time similarity function, which is defined as follows: (1) in: For ships i and ships j The membership function of the collision risk between them in the time dimension; A quantitative model of the motion similarity of encountering ships is constructed based on the membership function of collision risk in the time dimension. Construct a motion similarity quantification matrix for meeting ship pairs to determine the degree of temporal connection: Among them The larger the value, the more similar the risk characteristics of the two ships in terms of time urgency; only when the ships are similar in terms of (…). i , j This belongs to the high-risk group. If the time-related collision risk correlation is calculated, otherwise the value is directly assigned to 0, indicating that there is no significant conflict correlation between the two ships.

4. The method for dynamic association and identification of collision hazards under complex encounter situations according to claim 1, characterized in that: The comprehensive collision hazard index is expressed as follows: in r To comprehensively assess collision hazard indicators, , The corresponding spatial collision hazard level Time-based collision risk The weight value.

5. The method for dynamic association and identification of collision hazards under complex encounter situations according to claim 1, characterized in that: The process by which this vessel operates according to collision avoidance priorities based on the improved speed barrier method is as follows: A relative motion ray is constructed based on the ship's position and the relative velocity vectors of the two ships. This relative motion ray is defined by the ship's position at time [time value missing]. t Location P O ( t Starting from ) along V OT ( t ) Extending in a direction, represented as: (1) in, Φ O ( t ), P O ( t )and V O ( t ) represent the ship's OS at time . t The course, position, and speed of the ship. Φ T ( t ), P T ( t )and V T ( t (The following appears to be a separate, unrelated sentence:) These are the times of other ships. t The course, position, and speed; τ For ray parameters whose values ​​are not less than zero; V OT ( t (time) t The relative velocity vector of this vessel with respect to the target vessel; λ ( P O ( t ), V OT ( t )) represents the ray of relative motion. When the relative motion ray is located at two tangents and When the area between the target and target ships is considered to pose a collision risk, the set of relative velocities that would cause a collision between the two ships is defined. V OT relative collision area (Relative collision zone) Represented as: The two tangents are respectively from the position of the ship. P O ( t The left tangent line drawn towards the boundary of the target vessel's safety zone. L O 1 ( t ) and right tangent L O 2 ( t The two tangents respectively form relative collision cones. RCC OT The left and right boundaries; (2) When there are multiple obstacles, each obstacle needs to be... RCC Converting to a unified velocity space, and transforming the relative velocity space into the ship's absolute velocity space, the velocity obstacle that would cause a collision with the target ship is: (3) in The Minkowski operation represents the function of moving the relative collision zone based on the velocity vector of the other vessel. The ship's velocity vector cannot be in the range of area; Considering the actual maneuvering characteristics of the giving-way vessel, and based on navigational practice, the course change range for the giving-way vessel to take evasive action is [-π / 3, π / 3]. If this vessel is the giving-way vessel, the actual permissible course range (ACR) is: (4) in, Φ ACR 1 and Φ ACR 2 These represent the actual permissible heading range. ACR The lower and upper limits; The course obstacle zone relative to other vessels is defined as follows: (5) in, Φ COR 1 and Φ COR 2 They represent the left tangent line respectively. L O 1 and right tangent L O 2 Determined left and right heading boundaries; Compared to The safe course region (SCR) of this vessel relative to other vessels is defined as follows: (6) The navigable area for ships is represented as follows: (7)。 6. The method for dynamic association and identification of collision hazards under complex encounter situations according to claim 1, characterized in that: The expression for the resumption timing at the nearest encounter point is as follows: (1) in, t r For the time when ships can resume sailing, n A positive integer set by the driver. t s For time step.

7. A device for dynamic correlation identification of collision hazards under complex encounter situations, characterized in that, Including the following: A collision risk calculation model module based on domain intrusion parameters is constructed: It is used to construct a collision risk calculation model based on domain intrusion parameters through a three-dimensional evaluation framework of spatial intrusion state, time dimension prediction and quantitative dimension characterization, combined with an asymmetric Gaussian distribution fitting algorithm, so as to realize the collision risk calculation of the ship and other target ships in complex scenarios based on spatiotemporal coupling characteristics. High-risk vessel screening module: This module is used to evaluate the collision risk of the vessel and other target vessels in complex scenarios based on the spatiotemporal coupling characteristics, as assessed by the collision risk calculation model. It uses a dual dynamic threshold mechanism of space and time to screen for dangerous combinations and thus screen for high-risk vessels. Hazard cluster segmentation module: This module is used to segment high-risk vessels after screening. Based on an improved hierarchical clustering algorithm, it incorporates the time collision hazard index into the improved profile coefficient. By quantifying the dynamic balance between the compactness within a cluster and the separation between clusters, it determines the optimal clustering level and the optimal number of levels, thereby enabling the segmentation of clusters with similar time collision hazards. Collision avoidance priority ranking module: used to calculate the information entropy weight of the comprehensive collision hazard index based on the entropy weight method, and to rank the collision avoidance priority of ships after they are divided into clusters; Operation Module I: Used to enable the vessel to operate according to collision avoidance priorities based on the improved speed barrier method; Operation Module II: Used to enable the vessel to return to its original trajectory after performing operations according to the collision avoidance priority based on the re-entry timing of the nearest encounter point; The process of screening high-risk vessels by using a dual dynamic threshold mechanism of space and time to filter hazardous combinations is as follows: When the spatial collision risk level between the target ship and this ship meets the requirements r s ≥ r 0, r 0 represents the preset safety threshold, and the time-based collision risk level reaches [a certain threshold]. r t ≥ r 0+Δ r At that time, the ship will be automatically calibrated to ( i,j) This is a high-risk portfolio; Where Δ r To dynamically adjust parameters and meet the requirements and The set of ship pair sequences is ,satisfy and The set of ship pair sequences is ; The improved hierarchical clustering algorithm includes the following process: First, each ship in P' is considered a cluster ( Initial cluster state The description is as follows: The calculation of the average time collision risk among ships in a group is expressed as follows: in, C m and C n Representing clusters m and n ;| C m | and | C n | represent clusters respectively C m and C n The number of ships in the country; L avg ( C m , C n This represents the average time-based collision risk for all cross-cluster vessel pairs between two clusters. in, i , j Clusters C m , C n The ship index within, d ( i , j ) indicates a ship i With ships j The degree of dispersion of the risk of collision over time; In the k During the next iteration, a global search is performed on the current hazard matrix to determine the maximum average time collision hazard between ships in the cluster: in, [ p , q [ ] is an index with the highest average time collision risk between ship pairs in the cluster. N k-1 It is the cluster number in the (k-1)th iteration. The corresponding clusters are merged to form a new cluster. Among them, C l k Let C represent the new cluster formed in the k-th iteration. p k-1 With C q k-1 Let p and q represent the two clusters selected for merging in the (k-1)th iteration; No. k The cluster representation updated in the next iteration is as follows: Synchronize and update the similarity matrix in the (k-1)th iteration; The iteration process will terminate when the number of remaining objects meets the preset termination condition. At this point, the remaining objects have met the condition of being uniquely assigned to the corresponding category or do not require further merging, ensuring the determinism of the classification results and the efficiency of the algorithm. The improved formula for calculating the profile coefficient is as follows: Among them, C m It is a cluster containing ship i, C n It is related to cluster C m Different clusters, a ( i ) is a ship i With cluster C m The average time-based collision risk among other vessels in the vessel. b ( i ) is a ship i In addition to other clusters, C m In addition, the minimum average time collision risk between them; For a cluster containing only a single vessel, i.e. | C i |=1, profile coefficient is set to zero; Candidate sets generated by hierarchical clustering are grouped together. Calculate the improved [structure] under this candidate hierarchy. S Weighted average: in, n Indicates candidate level k The total number of ships involved in the contour factor calculation; N k This indicates the number of clusters at that level; the two have different meanings. After calculating the weighted improved profile coefficients for all candidate levels, determine the optimal number of levels.

8. A computer device, comprising: A processor and a memory, wherein the memory stores a program module, characterized in that the program module runs on the processor to implement the method as described in any one of claims 1-6.

Images