A Ship Dynamic Clustering Method Based on Complexity

Abstract

The present invention discloses a ship dynamic clustering method based on complexity, which relates to the technical field of ship traffic management and safety. The method includes obtaining ship AIS data in a research area and a time period, and preprocessing the abnormal data in the AIS data; dividing the ship AIS data into multiple time periods at the same time interval according to the time sequence to obtain ship AIS data in different time periods; calculating the traffic complexity between ships based on the ship AIS data in different time periods; constructing a ship topology network graph in different time periods based on the traffic complexity between ships; constructing a Louvain model based on the ship topology network graph in different time periods, dynamically clustering the ships, and identifying the partitions to which each ship belongs; and using modularity to detect the effect of the partitions to which each identified ship belongs, so as to complete the ship dynamic clustering based on complexity. The present invention solves the problem that the existing clustering methods fail to consider the interaction characteristics between ships.

Description

Technical Field

[0001] The present invention relates to the technical field of ship traffic management and safety, and particularly to a ship dynamic clustering method based on complexity. Background Art

[0002] Ship traffic safety has always been a matter of great concern in the shipping industry. Especially in the case of a large volume of ship traffic, it increases the possibility of maritime traffic accidents, accompanied by casualties, environmental pollution, and economic losses. Therefore, strengthening ship traffic safety management, optimizing navigation routes, and enhancing monitoring and early warning capabilities have become important tasks for ensuring the safety and sustainable development of maritime transportation.

[0003] Currently, various methods and models have been proposed in the field of collision risk estimation and control to reduce risks and improve efficiency in maritime transportation. In particular, the widespread deployment of the Automatic Identification System (AIS) can not only provide early collision alerts and assist in anti-collision decision-making to enhance safety management, but also provide rich AIS trajectory data, promoting the accuracy and effectiveness of the analysis and modeling of trajectory behavior and driving the development of this field. However, there are still many challenges to be solved, such as the dynamic interaction characteristics of ships in complex navigation environments and the identification of different ship behavior patterns. In waters with high ship density, ship traffic usually exhibits clustering characteristics, with a high degree of spatio-temporal interaction between ships within the same group and relatively sparse interaction between different groups. The most commonly used clustering method is the density-based DBSCAN algorithm. However, this method only considers the spatial distance between ships and fails to consider the interaction characteristics between ships, such as the influence of the navigation angle, speed, and time dimension between ships. Summary of the Invention

[0004] In view of the above deficiencies in the prior art, a ship dynamic clustering method based on complexity provided by the present invention solves the problem that existing clustering methods fail to consider the interaction characteristics between ships.

[0005] To achieve the above invention objective, the technical solution adopted by the present invention is: a ship dynamic clustering method based on complexity, comprising the following steps:

[0006] S1: Obtain ship AIS data within the research area and time period, and perform preprocessing on the abnormal data in the ship AIS data;

[0007] S2: Based on the preprocessing result, divide the ship AIS data into multiple time periods at the same time interval in chronological order to obtain ship AIS data for different time periods;

[0008] S3: Calculate the traffic complexity between ships based on the ship AIS data for different time periods;

[0009] S4: Construct a ship topological network diagram for different time periods based on the traffic complexity among ships;

[0010] S5: Construct a Louvain model based on the ship topological network diagrams for different time periods, dynamically cluster the ships, and identify the partitions to which each ship belongs;

[0011] S6: Use modularity to detect the effect of the partitions to which each identified ship belongs, and complete the dynamic clustering of ships based on complexity.

[0012] Furthermore, the traffic complexity among ships in S3 includes traffic density complexity and traffic conflict complexity. The traffic density complexity Den ij is:

[0013]

[0014] where λ and α are correction parameters dependent on the navigation environment, both being positive numbers, is the relative distance vector of ships, and SD ij is the minimum safe navigation distance between ship i and ship j;

[0015] The traffic conflict complexity conf ij is:

[0016] conf ij = angle ij · conv ij

[0017]

[0018]

[0019] where angle ij is the complexity of the angle between two ships, conv ij is the convergence complexity between two ships, f(·) is the function between the complexity of the navigation angle and the angle, θ is the navigation angle, L low is the minimum threshold of the distance between two ships, L up is the maximum threshold of the distance between two ships, is the relative velocity vector.

[0020] Furthermore, the traffic complexity among ships in S3 is:

[0021]

[0022] where complex ij (·) is the traffic complexity function among ships, θ ijis the navigation angle between ship i and ship j, Den ij (·) is the traffic density complexity function, conf ij (·) is the traffic conflict complexity function.

[0023] Furthermore, in S4, the traffic complexity between ships is used as the weight of the edge, and the ships are used as vertices to construct the ship topology network diagram in different time periods. The formula is:

[0024] G Ti =(V Ti ,E Ti ,w Ti )

[0025] Among them, G Ti is the ship topology network, V Ti is the vertex set in the topological network in Ti period, E Ti is the set of edges in the topological network in time period Ti, w Ti is the weight set in the topological network in the Ti period.

[0026] Furthermore, the S5 comprises the following sub-steps:

[0027] S51: Ship topology network G in time period Ti Ti Assign a unique partition number to the ships in The set of partition numbers to which the ship initially belongs in the Ti time period;

[0028] S52: The ship topology network G in the Ti time period Ti and partition number Input the Louvain model and output the ship topology network G Ti The ship belongs to the zone C Ti =(V Ti ,c ti ), c ti The set of partition numbers to which the ship belongs in the Ti time period;

[0029] S53: Extract the ship topology network G in the Ti time period Ti The ship set and the ship topology network G in the Ti+1 time period Ti+1 The same ship in the ship set V Ti ∩V Ti+1 , V Ti+1 is the vertex set in the topological network in the Ti+1 period;

[0030] S54: V Ti ∩V Ti+1 The ships in the set are assigned the partition number of the ship in the Ti time period, which is V Ti+1Assign a unique partition number to each of the remaining ships to obtain the initial partition to which the ships belong in the time period Ti+1 is the set of initial partition numbers to which the ships belong in the time period Ti+1;

[0031] S55: Input the ship topological network G Ti+1 in the time period Ti+1 and the initial partition to which the ships belong in the time period Ti+1 into the Louvain model, and output the partition C Ti+1 to which the ships in the ship topological network G belong, where C Ti+1 = (V Ti+1 , c ti+1 ), and c ti+1 is the set of partition numbers to which the ships belong in the time period Ti+1;

[0032] S56: Repeat the above steps S51 - S55 until the AIS data for all time periods are analyzed to complete the identification of the partitions to which each ship belongs.

[0033] Furthermore, the modularity Q in the S6 is:

[0034]

[0035] where m is the total weight of all edges, w ij is the weight of the edge between ship i and ship j, k i is the number of edges connected to ship i, k j is the number of edges connected to ship j, δ(·) is the Krill function, c i is the partition to which ship i belongs, and c j is the partition to which ship j belongs.

[0036] The beneficial effects of the present invention are as follows: The present invention can more effectively capture and represent the interaction characteristics of the navigation angle, speed, and time dimension effects between ships, and then cluster the ships to identify a set of highly interactive ships, helping to quickly establish dynamic constraint boundaries, that is, identify the neighboring ships that need to be considered in the conflict avoidance strategy. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 is a flowchart of a ship dynamic clustering method based on complexity.

[0038] Figure 2 are graphs of ship partition results in different time periods. DETAILED DESCRIPTION OF THE INVENTION

[0039] The present invention will be further described below with reference to the drawings and specific embodiments.

[0040] As shown in Figure 1As shown in the figure, a complexity-based dynamic clustering method for ships includes the following steps:

[0041] S1: Obtain the ship AIS data within the research area and time period, and preprocess the abnormal data in the ship AIS data;

[0042] S2: Based on the preprocessing results, divide the ship AIS data into multiple time periods at the same time interval in chronological order to obtain the ship AIS data for different time periods;

[0043] S3: Calculate the traffic complexity between ships based on the ship AIS data for different time periods;

[0044] S4: Construct a ship topological network diagram for different time periods based on the traffic complexity between ships;

[0045] S5: Construct a Louvain model based on the ship topological network diagrams for different time periods, dynamically cluster the ships, and identify the partitions to which each ship belongs;

[0046] S6: Use modularity to detect the effect of the partitions to which each identified ship belongs, and complete the dynamic clustering of ships based on complexity.

[0047] The present invention is based on the historical trajectory data of the Automatic Identification System (AIS) of ships, and combines the speed difference, course angle, distance, etc. between ships to calculate the traffic complexity between ships (including traffic density complexity and traffic conflict complexity). A topological network is constructed based on the complexity between ships, that is, ships are vertices, the complexity between ships is greater than 0, there is an edge between ships, and this complexity is the weight of the edge. Based on the constructed topological network, the dynamic louvain algorithm is used to cluster the ships. The Louvain algorithm is very efficient in processing complex graphs, and the time complexity is O(nlogn).

[0048] The traffic complexity between ships in S3 includes traffic density complexity and traffic conflict complexity. The traffic density complexity refers to the traffic complexity caused by ship density. Define the traffic density complexity Den between ships ij as a function of the distance between ships. The traffic density complexity Den ij is:

[0049]

[0050] where λ and α are correction parameters depending on the navigation environment, both are positive numbers, is the relative distance vector of ships, SD ij is the minimum safe navigation distance between ship i and ship j, SD ij The value of depends on the type of ship and the navigation environment;

[0051] Traffic conflict complexity refers to the traffic complexity caused by navigation conflicts between ships. The factors affecting conflicts are the navigation angle and whether there is convergence or not.

[0052] When the navigation angle θ between two ships is small, such as less than 20 degrees, the two ships are almost parallel, the minimum collision distance is short, and the ship has more time to take avoidance measures. Therefore, this situation is relatively safe. When θ is 180 degrees (under the meeting condition), the two ships will be in a head-on meeting state. Although the relative speed of the ships is the highest, according to the International Regulations for Preventing Collisions at Sea (COLREGs, 1972), the responsibilities of both sides are clear. Therefore, this situation is not the most dangerous. When θ is about 120 degrees, the minimum collision distance is the longest, and the ship needs to take action in advance. When θ is about 120 degrees and the two ships meet, the relative speed of the ships is relatively high, and the action becomes more complex. This situation is the most dangerous. Define the functional relationship between the navigation angle complexity and the angle as:

[0053]

[0054] Also consider that when the distance between two ships is smaller, the complexity between the ships is higher, and the complexity value reaches the maximum. This means that when other ships approach one's own ship, the potential collision risk increases. However, when the distance between the two ships is very small, their navigation trajectories and angles become highly correlated, and there is almost no room for any change or adjustment. Therefore, the influence of the change in the angle on the collision risk decreases and becomes insignificant. When the distance between the ships is greater than a certain threshold, there is no potential risk between the ships. Therefore, the angle complexity angle between the two ships ij only exists when the distance between the two ships is less than a certain threshold L up and greater than a certain threshold L low Therefore, define the angle complexity angle between the two ships ij as:

[0055]

[0056] Assume that when two ships sail at a certain angle, when they move away from each other, the collision risk and traffic complexity will be reduced. When the two ships gradually approach and meet, the collision risk and navigation complexity will increase. The traffic complexity caused by whether the two ships converge or not is the convergence complexity. The convergence complexity between the two ships is:

[0057]

[0058] When it means that the two ships approach each other and meet. Conversely, the two ships gradually move away from each other. Then the traffic conflict complexity between the two ships is:

[0059] conf ij = angle ij · conv ij

[0060] where angle ij is the complexity of the included angle between two ships, conv ij is the complexity of convergence between two ships, f(·) is the function between the complexity of the navigation included angle and the included angle, θ is the navigation included angle, L low is the minimum threshold of the distance between two ships, L up is the maximum threshold of the distance between two ships, is the relative velocity vector.

[0061] The traffic complexity between ships in S3 is as follows:

[0062]

[0063] where complex ij (·) is the traffic complexity function between ships, θ ij is the navigation included angle between ship i and ship j, Den ij (·) is the traffic density complexity function, conf ij (·) is the traffic conflict complexity function.

[0064] In S4, taking the traffic complexity between ships as the edge weight and ships as vertices, a ship topological network diagram for different time periods is constructed. The formula is:

[0065] G Ti = (V Ti , E Ti , w Ti )

[0066] where G Ti is the ship topological network, V Ti is the vertex set in the topological network in the Ti time period, E Ti is the edge set in the topological network in the Ti time period, w Ti is the weight set in the topological network in the Ti time period.

[0067] S5 includes the following sub-steps:

[0068] S51: Assign a unique partition number to the ships in the ship topological network G Ti in the Ti time period is the set of initial partition numbers of ships in the Ti time period;

[0069] S52: The ship topological network G Ti in the Ti time period and the partition number Input into the Louvain model, and output the ship topological network G Ti of the ship's affiliated partition C Ti =(V Ti , c ti ), where c ti is the set of the numbers of the partitions to which the ships belong in the Ti time period;

[0070] S53: Extract the ships in the ship topological network G Ti in the Ti time period and the ships in the ship topological network G Ti+1 in the Ti+1 time period that have the same ship set V Ti ∩V Ti+1 , where V Ti+1 is the set of vertices in the topological network in the Ti+1 time period;

[0071] S54: Assign the numbers of the partitions to which the ships in the V Ti ∩V Ti+1 set belong in the Ti time period, and assign a unique partition number to each of the remaining ships in V Ti+1 to obtain the initial partitions to which the ships belong in the Ti+1 time period is the set of the numbers of the initial partitions to which the ships belong in the Ti+1 time period;

[0072] S55: Input the ship topological network G Ti+1 in the Ti+1 time period and the initial partitions to which the ships belong in the Ti+1 time period into the Louvain model, and output the ship's affiliated partition C Ti+1 of the ship topological network G Ti+1 =(V Ti+1 , c ti+1 ), where c ti+1 is the set of the numbers of the partitions to which the ships belong in the Ti+1 time period;

[0073] S56: Repeat the above steps S51 - S55 until all AIS data in all time periods are analyzed to complete the identification of the partitions to which each ship belongs.

[0074] The modularity Q in the above S6 is as follows:

[0075]

[0076] where m is the total weight of all edges, w ij is the weight of the edge between ship i and ship j, that is, the traffic complexity complex ij between the ships, k i is the degree of ship i, that is, the number of edges connected to ship i, k jis the degree of vessel j, i.e., the number of edges connected to vessel j. δ(·) is the Kronecker delta function. When c i = c j , it takes the value of 1, otherwise 0. c i is the partition to which vessel i belongs, and c j is the partition to which vessel j belongs.

[0077] The modularity value usually ranges from -1 to 1. The higher the value, the more obvious the partition structure. When the value of Q is greater than 0.3, the partition structure is usually considered obvious. A value between 0.3 and 0.7 usually indicates a relatively good community structure. When Q is close to 0 or negative, it indicates that the partition structure is not obvious.

[0078] In an embodiment of the present invention, the AIS data of vessels in the Qiongzhou Strait area from 09:00:00 to 09:09:00 on October 9, 2023 is selected. Through steps S1 and S2, the AIS data is preprocessed and divided into three time periods of data at intervals of 3 minutes, with time tags T0, T1, and T2; the traffic complexity between vessels in different time periods is calculated through step S3; through step S4, a topological network diagram of the traffic complexity of vessels in different time periods is constructed; through step S5, the vessel partition results for the three time periods are obtained. As Figure 2 shown, the vessel partition results for different time periods (the circled numbers are partition numbers) show that the vessel partition structure is significant (the modularity values are 0.706, 0.643, and 0.626 respectively), with high-density interactions among vessels within the partitions and relatively sparse interactions between partitions.

[0079] Those of ordinary skill in the art will realize that the embodiments described herein are for helping readers understand the principles of the present invention and should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various other specific deformations and combinations that do not depart from the essence of the present invention based on the technical revelations disclosed in the present invention, and these deformations and combinations are still within the protection scope of the invention.

Claims

1. A ship dynamic clustering method based on complexity, characterized in that: The following steps are involved: S1: Obtain ship AIS data within the study area and time period, and perform abnormal data preprocessing on the ship AIS data; S2: Based on the preprocessing results, the ship AIS data is divided into multiple time periods in chronological order at the same time interval to obtain the ship AIS data of different time periods; S3: Calculate the traffic complexity between ships based on the ship AIS data in different time periods; The traffic complexity between ships in S3 includes traffic density complexity and traffic conflict complexity. for: in, and are correction parameters that depend on the navigation environment and are all positive numbers. is the relative distance vector of the ship, For ships and ships The minimum safe sailing distance between The traffic conflict complexity for: in, is the complexity of the angle between the two ships, is the convergence complexity between the two ships, is the function between navigation angle complexity and angle, is the navigation angle, is the minimum threshold of the distance between two ships, is the maximum threshold of the distance between two ships, is the relative velocity vector; The traffic complexity between ships in S3 is: in, is the traffic complexity function between ships, For ships and ships The navigation angle, is the traffic density complexity function, is the traffic conflict complexity function; S4: Based on the traffic complexity between ships, construct the ship topology network diagram in different time periods; S5: Based on the ship topology network diagram in different time periods, the Louvain model is constructed to dynamically cluster ships and identify the partition to which each ship belongs; S6: Modularity is used to detect the effectiveness of the partitions to which each identified ship belongs, and dynamic clustering of ships based on complexity is completed.

2. A complexity-based ship dynamic clustering method according to claim 1, characterized in that: In S4, the traffic complexity between ships is used as the edge weight, and the ships are used as vertices to construct the ship topology network diagram in different time periods. The formula is: in, For the ship topology network, for The set of vertices in the time period topology network, for The set of edges in a time-slot topology network, for The set of weights in a time-slot topology network.

3. A complexity-based ship dynamic clustering method according to claim 2, characterized in that: The S5 comprises the following sub-steps: S51: Ship topology network in time period Assign unique partition numbers to ships in , for The set of partition numbers to which the ship initially belongs in the time period; S52: Ship topology network in time period and partition number Input Louvain model, output ship topology network The ship's zone , for The set of partition numbers to which the ships belong in the time period; S53: Extraction Ship topology network in time period Ship collection and Ship topology network in time period Ships with the same ship set , for The set of vertices in a time period topology network; S54: Ship assignment in the collection The zone number to which the ship belongs in the time period is Each of the remaining ships in the The zone to which the ship initially belongs during the time period , for The set of partition numbers to which the ship initially belongs in the time period; S55: Ship topology network in time period and Time period The initial zone to which the ship belongs Input Louvain model, output ship topology network The ship's zone , for The set of partition numbers to which the ships belong in the time period; S56: Repeat the above steps S51-S55 until the AIS data of all time periods are analyzed and the partition to which each ship belongs is identified.

4. A complexity-based ship dynamic clustering method according to claim 1, characterized in that: The modularity of S6 for: in, is the total weight of all edges, For ships and ships The weight of the edge between For ships The number of connected edges, For ships The number of connected edges, is the Creel function, For ships The partition to which it belongs, For ships The partition to which it belongs.

Images