Ship trajectory recovery and clustering method

Through the correction and cluster analysis of ship trajectory data, the problem of ship navigation trajectory deviation in complex environments is solved, and the accurate correction and optimization of trajectory is achieved, providing effective data support for route planning and operating status analysis.

CN119089221AActive Publication Date: 2024-12-06THREE GORGES ELECTRIC POWER JINZHOU ENERGY (HUBEI) CO LTD

Patent Information

Application Number
CN202411099704.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-12
Publication Date
2024-12-06
Estimated Expiration
2044-08-12

AI Technical Summary

Technical Problem

The ship's navigation trajectory is difficult to accurately record and analyze in complex environments, resulting in deviation from the expected trajectory and affecting subsequent flight route planning and operating status analysis.

Method used

By obtaining ship operation and route environment data, identifying abnormal nodes, establishing route networks and ship trajectory expressions, correcting trajectory nodes, fusing trajectory data, and achieving trajectory integration and optimization through clustering analysis.

Benefits of technology

Effectively correct abnormal nodes, reduce error data, provide optimized trajectory analysis tools, and support route planning and ship status prediction.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119089221A_ABST
    Figure CN119089221A_ABST
Patent Text Reader

Abstract

The invention belongs to the technical field of ship route analysis methods, and particularly relates to a ship trajectory recovery and clustering method. The method specifically comprises the following steps: collecting ship operation data and route environment data in to-be-analyzed route historical data; extracting ship operation data and route environment data corresponding to each node, determining nodes which are not matched with ship parameters and route environment facts, establishing a route network and ship trajectory expression to position suspected segments, calculating duplex affiliation probability, and further screening to determine an actual route; according to the ship trajectory restoration and clustering method, restoration and correction of abnormal node data in the ship trajectory can be effectively achieved, error data are effectively reduced, an optimization tool is provided for ship trajectory analysis data processing, integration analysis of the ship trajectory is achieved through optimization clustering, and support is provided for ship trajectory prediction and planning.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of ship route analysis methods, and in particular relates to a ship track restoration and clustering method. Background Art

[0002] With the further development and utilization of the country's inland rivers and marine resources, and the rapid development of shipping production based on inland rivers and oceans, shipping has become an important boost to the sustainable economic development of riverside and coastal areas. Unlike the traditional transportation industry, which can achieve precise positioning based on a large number of land-based precise positioning and navigation measures, the current shipping system mainly relies on long-range GPS and latitude and longitude monitoring equipment to achieve positioning and navigation. Its accuracy and real-time performance cannot be effectively guaranteed. Even if the same ship performs the same transportation task along the same route, the combined influence of the ever-changing route environment and unpredictable immediate disposal methods may cause the navigation trajectory of the ship to change each time, and generate a large number of different operation trajectories in the ship's navigation log and other records. Although these trajectories may generally conform to the basic form of each heading, when it is necessary to further analyze the operating status and data of the ship in each section, a large amount of trajectory data that deviates from expectations and actual situations will be generated, making it difficult to execute subsequent steps. Summary of the invention

[0003] The purpose of the present invention is to provide a ship trajectory restoration and clustering method for a solution for restoring ship trajectories that deviate from the actual situation during route and ship trajectory analysis, and further perform trajectory clustering analysis to provide a basic data optimization approach for route planning and ship operation status analysis.

[0004] To achieve the above purpose, the present invention adopts the following technical solution.

[0005] The ship track restoration and clustering method of the present invention is mainly used to analyze and extract ship track data, restore and correct abnormal estimation nodes according to the actual route and ship track data, so as to obtain correct and effective real data, and realize cluster analysis of ship tracks on this basis, providing an effective solution for unifying the historical tracks of multiple voyages. It specifically includes the following steps:

[0006] Step 1: Get the original data set

[0007] Collect ship operation data and route environment data from the historical data of the route to be analyzed; shipping data includes ship static parameters and ship dynamic navigation parameters; route environment data includes water temperature environment parameters and weather environment parameters;

[0008] Step 2: ship track node correction, including steps B1 to B5;

[0009] B1. Determine the unmatched nodes step: extract the ship operation data and route environment data corresponding to each node, and determine the nodes that do not match the ship's own parameters and route environment facts;

[0010] B2. Establishing route network and ship trajectory expression: Based on the current route construction situation, a route network directed graph G(V, E) is established, where V = (v a |a=1,2...A) represents the set of route nodes, E=(e b |b=1,2...B) represents the set of flight segments, where a flight segment refers to a route segment between two route nodes; based on the specific size of the route network and the computing power of the computing platform, it is divided into several network grids, and the intersecting routes in each grid are determined;

[0011] Based on the ship navigation data, a ship trajectory expression consisting of trajectory nodes is established. m}, where P m ={p m,i} represents the mth ship trajectory, p m,i ={x m,i ,y m,i , t m,i , v m,i ,u m,i} represents the feature vector of the ith track point in the mth ship track, x m,i and m,i Indicates the latitude and longitude values, t m,i Indicates the time node, v m,i Indicates the ship speed, u m,i Indicates heading;

[0012] B3. Suspected segment positioning: For a certain trajectory point p m,i , taking its network grid as the center, select R m,i ×R m,i The flight segments in the range constitute the suspected flight segment E m,i ;

[0013] Where R m,i is a variable selected based on the actual distance between the current trajectory point and the nearby flight segments. m,i ×R m,i If there is no suspected segment within the range, R needs to be increased. m,i , if R m,i ×R m,i If the number of suspected flight segments exceeds a certain limit, R needs to be reduced. m,i , to ensure R m,i ×R m,i The number of suspected flight segments within the range is not less than 1 and does not exceed the preset maximum number;

[0014] Analysis of Suspected Segment E m,i The direction of each segment in the trajectory is related to the ship at the node p m,i If the angle exceeds 90°, it means that the direction of the segment is inconsistent with the actual heading of the ship and it will be removed;

[0015] B4. Track node positioning:

[0016] Based on the suspected flight segment E after eliminating invalid data m,i ={e m,i,1 , e m,i,2 ...e m,i,c ...e m,i,C}, calculate the trajectory point p respectively m,i The probability of dual belonging to each suspected segment is:

[0017]

[0018] where e m,i,c represents the trajectory point p m,i The cth suspected segment; ρ represents the positioning standard deviation, which is generally 2 to 10; r m,i,c is the trajectory point p m,i To the cth suspected segment e m,i,c The distance; α m,i,c is the ship at the trajectory point p m,i The heading of the cth suspected segment e m,i,c The deviation angle of the heading; γ 1 is the distance weight coefficient, γ 2 is the heading weight coefficient, γ 1 +γ 2 =1;

[0019] B5. Track node correction: According to the calculation results, the one with the highest probability of belonging is selected as the actual route, or several suspected routes with the highest probability of belonging are selected in order to be identified by the expert system or further screened to determine the actual route; after determining the actual route, the ship's track nodes are mapped to the corresponding route using the vertical mapping method; the track point P is mapped according to the feature vectors of the front and rear nodes adjacent to the track point pm,i on the route. m,i The eigenvector p m,i ={x m,i ,y m,i , t m,i , v m,i ,u m,i}Make corrections;

[0020] Step 3: Ship trajectory fusion distance expression, including C1~C4;

[0021] C1. Obtain the characteristic parameters of each node of each ship trajectory, and obtain the ship trajectory expressed by the multi-dimensional characteristic vector of each node;

[0022] C2. For any two multidimensional feature vectors expressing the trajectory {P m} and {Q n}, remove the overlapping common nodes, and connect the remaining nodes in the way of shortest Euclidean distance to obtain the trajectory {P m} and {Q n The fusion trajectory of all nodes in {R m∩n};

[0023] C3, the trajectory {P m}={P 1 , P 2 ...P m ...P M Each node P in m Along the fusion trajectory {R m∩n} after being connected to form a local fusion trajectory {P′ m}; Set the trajectory {Q n}={Q 1 ,Q 2 ...Q n ...Q N Each node Q in n Along the fusion trajectory {R m∪n} after being connected to form a local fusion trajectory {Q′ n};

[0024] C4. Calculate the fusion distance expression between nodes of the local fusion trajectory

[0025]

[0026] Among them, L m∪n represents the fusion trajectory {R m∪n}; d(a, b) represents the Euclidean distance between nodes a and b; S′ m Denotes the fusion trajectory {P′ m The time complexity of the trajectory consisting of the 1st to Mth nodes in}, L′ n Denotes the fusion trajectory {Q′ n The time complexity of the trajectory consisting of the 1st to nth nodes in};

[0027] Step 4: Ship trajectory fusion clustering;

[0028] Based on the feature fusion distance of each node in any two trajectories obtained in step 3, the trajectory nodes are clustered using the fusion distance as a metric, specifically including D1 to D5:

[0029] D1. Determine the minimum number of trajectory points J required for each trajectory classification result 0 , initialize the cluster radius R 0 ;

[0030] D2, each trajectory node point i Centered by node i, search for nodes whose fusion distance with node i does not exceed R 0 All nodes constitute a valid node set {point i,1 , point i,2 ...point i,T}, count the number of nodes in the valid node set T;

[0031] D3. Extract T≥J 0 The valid node set corresponding to the trajectory nodes constitutes the central node set {point′ i}; with the central node set {point′ i} an element point′ i.1 Traverse the central node set for the initial node, searching for a fusion distance no greater than R 0 The other elements form the central node sequence {point′ i.1 , point′ i.2 ...point′ i.x};

[0032] D4, connect the central nodes in sequence to form a classification trajectory {Line i}, for the node {point′ in the classification trajectory i.1 , point′ i.2 ...point′ i.x}, with its radius R 0 All nodes within the range are a cluster;

[0033] D5. Determine the classification results, delete duplicate clustering results, filter out node elements that have not been classified into any cluster, and perform the following verification:

[0034] D50 verifies whether the total number of undivided node elements or the proportion of the total number of nodes exceeds the preset noise threshold. If so, return to step D1 and increase the clustering radius R 0 If the value is not exceeded, go to step D51;

[0035] D51 verifies whether the total number of clustering results obtained exceeds the preset upper limit of the total number of clusters. If it exceeds, return to step D1 to increase the cluster radius R 0 The value is reclassified, if it does not exceed, go to step D52;

[0036] D52 verifies whether the total number of clustering results obtained is lower than the preset lower limit of the total number of clusters. If it exceeds, return to step D1 to reduce the cluster radius R 0 The value of is reclassified, and if it does not exceed, the clustering result is output.

[0037] As a further supplement to the aforementioned ship trajectory restoration and clustering method or as a preferred implementation method, the static parameters of the ship include route information, time node information, and ship parameter information; the dynamic navigation parameters include ship speed and heading information, engine operating status information, and fuel consumption information; the water temperature environment parameters include tide, flow rate, and water level; the weather environment parameters include wind direction and speed, air pressure, and weather conditions.

[0038] As a further supplement to the aforementioned ship trajectory restoration and clustering method or as a preferred implementation, the situations in step B1 that do not match the ship's own parameters include: the node speed and heading exceed the maximum speed of the ship, the heading exceeds the allowable heading, the engine parameters are abnormal, and the fuel consumption information is abnormal; the situations that do not match the route environment facts include: the coordinate data is obviously abnormal, the hydrological environment parameters are obviously inconsistent with the real-time records, and the weather environment is obviously inconsistent with the meteorological monitoring data.

[0039] As a further supplement to the above-mentioned ship trajectory restoration and clustering method or as a preferred implementation, for ships at a relatively close distance, r m,i,c The straight-line distance is used for calculation. For the longer distance, r m,i,c Great circle distance is used for calculation and all flight segments should use the same distance calculation method.

[0040] As a further supplement to the above-mentioned ship trajectory restoration and clustering method or as a preferred implementation, the narrower the channel is and the more complicated the channel curvature is, the more γ 1 The smaller γ 2 The bigger.

[0041] The beneficial effects are as follows: the ship trajectory restoration and clustering method of the present invention can effectively realize the restoration and correction of abnormal node data in the ship trajectory, effectively reduce erroneous data, provide an optimization tool for ship trajectory analysis data processing, and realize the integrated analysis of ship trajectories through optimized clustering, providing support for ship trajectory prediction and planning. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a flow chart of the ship trajectory restoration and clustering method.

[0043] Figure 2 It is a schematic diagram of ship trajectory node positioning;

[0044] Figure 3 is a schematic diagram of the fusion trajectory;

[0045] Figure 4 It is a schematic diagram of fusion distance clustering. DETAILED DESCRIPTION

[0046] The present invention is described in detail below in conjunction with specific embodiments.

[0047] like Figure 1 As shown, the ship trajectory restoration and clustering method of the present invention is mainly used to analyze and extract ship trajectory data, restore and correct abnormal estimation nodes according to the actual route and ship trajectory data, so as to obtain correct and effective real data, and realize cluster analysis of ship trajectories on this basis, providing a technical solution for unifying the historical trajectories of multiple voyages. It specifically includes the following steps:

[0048] Step 1: Get the original data set

[0049] Collect ship operation data from the historical data of the route to be analyzed. Shipping data includes ship operation data: static ship parameters (route information, time node information, ship parameter information), dynamic navigation parameters (ship speed and heading information, engine operation status information, fuel consumption information), route environment data: water temperature environment parameters (tide, flow rate, water level), weather environment parameters (wind direction and speed, air pressure, weather conditions), obtain the original data set for preprocessing, and eliminate invalid and redundant data;

[0050] Step 2: ship track node correction. This step is mainly used to correct the phenomenon that the track node does not match the route during continuous dynamic navigation, including steps B1 to B5;

[0051] B1. Determine the steps of mismatching nodes, including:

[0052] Extract the ship operation data and route environment data corresponding to each node, and determine the nodes that do not match the ship's own parameters and route environment facts;

[0053] The situations that do not match the ship's own parameters include: the node speed and heading exceed the maximum speed of the ship, the heading exceeds the allowed heading, the engine parameters are abnormal, and the fuel consumption information is abnormal;

[0054] The situations that do not match the route environment facts include: obvious abnormalities in coordinate data, obvious mismatches between hydrological environment parameters and real-time records, and obvious mismatches between weather environment and meteorological monitoring data;

[0055] B2. Establishing route network and ship trajectory expression, including:

[0056] Based on the current route construction situation, a route network directed graph G(V, E) is established, where V = (v a|a=1,2...A) represents the set of route nodes, E=(e b |b=1,2...B) represents the set of flight segments, where a flight segment refers to a route segment between two route nodes; based on the specific size of the route network and the computing power of the computing platform, it is divided into several network grids, and the intersecting routes in each grid are determined;

[0057] Based on the ship navigation data, a ship trajectory expression consisting of trajectory nodes is established. m}, where P m ={p m,i} represents the mth ship trajectory, p m,i ={x m,i ,y m,i , t m,i , v m,i ,u m,i} represents the feature vector of the ith track point in the mth ship track, x m,i and m,i Indicates the latitude and longitude values, t m,i Indicates the time node, v m,i Indicates the ship speed, u m,i Indicates heading;

[0058] B3. Suspected segment positioning

[0059] For a trajectory point p m,i , taking its network grid as the center, select R m,i ×R m,i The flight segments in the range constitute the suspected flight segment E m,i ;

[0060] Where R m,i According to the trajectory point p m,i A variable selected by the actual distance between the nearby flight segments. m,i ×R m,i If there is no suspected segment within the range, R needs to be increased. m,i , if R m,i ×R m,i If the number of suspected flight segments exceeds a certain limit, R needs to be reduced. m,i , to ensure R m,i ×R m,i The number of suspected flight segments within the range is not less than 1 and does not exceed the preset maximum number;

[0061] Analysis of Suspected Segment E m,i The direction of each segment in the trajectory is related to the ship at the node p m,i If the angle exceeds 90°, it means that the direction of the segment is inconsistent with the actual heading of the ship and it will be removed;

[0062] refer to Figure 2 As shown, we select the suspected flight segments within the 3×3 grid range of the trajectory point p, and obtain 1 , l 2 , l 3 Three suspected flight segments, of which l 3 If the heading angle between the flight segment and the trajectory point p exceeds 90°, then the trajectory point p cannot be along l 3 The flight segments should be eliminated, and finally the possible suspicious segments l 1 , l 2 The suspected segment is the segment closest to the actual position or adjacent position of the inferred ship trajectory node, and the specific segment needs to be further determined.

[0063] The following are the steps for further positioning:

[0064] B4. Trajectory node positioning

[0065] This application combines the actual operation characteristics of the ship and considers the joint judgment based on the attribution rate of the two key characteristic parameters, namely, the distance between the ship's trajectory node and the suspected segment and the heading, to determine the segment to which the ship's trajectory node is most likely to belong. Specifically:

[0066] Based on the suspicious flight segment set E after eliminating invalid data m,i ={e m,i,1 , e m,i,2 ...e m,i,c ...e m,i,C}, calculate the trajectory point p respectively m,i The probability of dual belonging to each suspected segment

[0067]

[0068] where e m,i,c represents the trajectory point p m,i The cth suspected segment of the ship; ρ represents the standard deviation of the ship's positioning, which is generally 2 to 10. The higher the positioning accuracy, the smaller the standard deviation; r m,i,c is the trajectory point p m,i To the cth suspected segment e m,i,c For a shorter distance, the straight-line distance can be used for calculation, and for a longer distance, the great-circle distance should be used for calculation, and all flight segments should use the same distance calculation method; m,i,c is the ship at the trajectory point p m,i The heading of the cth suspected segment e m,i,c The deviation angle of the heading; γ 1 is the distance weight coefficient, γ 2 is the heading weight coefficient, γ1+γ 2=1, and the narrower the channel is, the more complicated the channel curvature is, then γ 1 The smaller γ 2 The bigger;

[0069] B5. Track node correction

[0070] According to the calculation results, the one with the highest probability of belonging is selected as the actual route, or several suspected routes with the highest probability of belonging are selected in order to be identified by the expert system or further screened to determine the actual route; after determining the actual route, the vertical mapping method is used to map the ship trajectory nodes to the corresponding route; the trajectory point p is mapped according to the feature vectors of the front and rear nodes adjacent to the trajectory point pm,i on the route. m,i The eigenvector p m,i ={x m,i ,y m,i , t m,i , v m,i ,u m,i}Make corrections;

[0071] Determine the trajectory point p m,i After the specific position on the flight segment, the trajectory point p is corrected by the adjacent nodes before and after or the endpoint of the flight segment (when there are no other trajectory nodes in the flight segment) m,i ,For the various parameters in the eigenvector, the correction is mainly performed through numerical fitting or interpolation, so as to ensure that the changing trends of the various parameters are continuous and average when the ship continuously travels in the voyage.

[0072] Step 3: Ship trajectory fusion distance expression

[0073] C1. Obtain the characteristic parameters of each node of each ship trajectory, and obtain the ship trajectory expressed by the multi-dimensional characteristic vector of each node;

[0074] C2. For any two multidimensional feature vectors expressing the trajectory {P m} and {Q n}, remove the overlapping common nodes, and connect the remaining nodes in the way of shortest Euclidean distance to obtain the trajectory {P m} and {Q n The fusion trajectory of all nodes in {R m∪n};

[0075] Simple expression Figure 3 As shown, the black dots represent the trajectory nodes expressed by the multidimensional feature vector, the two lines on the left and right represent the two trajectory paths, and the thick dotted line represents the fused trajectory connecting all the nodes.

[0076] C3, the trajectory {P m}={P 1 , P 2 ...Pm ...P M Each node P in m Along the fusion trajectory {R m∪n} after being connected to form a local fusion trajectory {P′ m}; Set the trajectory {Q n}={Q 1, ,Q 2 ...Q n ...Q N Each node Q in n Along the fusion trajectory {R m∪n} after being connected to form a local fusion trajectory {Q′ n};

[0077] C4. Calculate the fusion distance expression between nodes of the local fusion trajectory

[0078] Among them, L m∪n represents the fusion trajectory {R m∪n}; d(a, b) represents the Euclidean distance between nodes a and b; S′ m Denotes the fusion trajectory {P′ m The time complexity of the trajectory consisting of the 1st to Mth nodes in}, L′ n Denotes the fusion trajectory {Q′ n The time complexity of the trajectory consisting of the 1st to nth nodes in};

[0079] Step 4: Ship trajectory fusion clustering

[0080] like Figure 4 As shown in Figure 1, based on step 3, the feature fusion distance of each node in any two trajectories is obtained, and the trajectory nodes are clustered using the fusion distance as a metric, where the black points represent the ship trajectories, and the distance between the points is the fusion distance;

[0081] The specific steps include:

[0082] D1. Determine the minimum number of trajectory points J required for each trajectory classification result 0 , initialize the cluster radius R 0 ;

[0083] D2, each trajectory node point i Centered by node i, search for nodes whose fusion distance with node i does not exceed R 0 All nodes constitute a valid node set {point i,1 , point i,2 ...point i,T}, count the number of nodes in the valid node set T;

[0084] D3. Extract T≥J 0 The valid node set corresponding to the trajectory nodes constitutes the central node set {point′ i}; with the central node set {point′ i} an element point′ i.1 Traverse the central node set for the initial node, searching for a fusion distance no greater than R 0 The other elements form the central node sequence {point′ i.1 , point′ i.2 ...point′ i.x};

[0085] D4, connect the central nodes in sequence to form a classification trajectory {Line i}, for the node {point′ in the classification trajectory i.1 , point′ i.2 ...point′ i.x}, with its radius R 0 All nodes within the range are a cluster;

[0086] D5. Determine the classification results, delete duplicate clustering results, filter out node elements that have not been classified into any cluster, and perform the following verification:

[0087] D50 verifies whether the total number of undivided node elements or the proportion of the total number of nodes exceeds the preset noise threshold. If so, return to step D1 and increase the clustering radius R 0 If the value is not exceeded, go to step D51;

[0088] D51 verifies whether the total number of clustering results obtained exceeds the preset upper limit of the total number of clusters. If it exceeds, return to step D1 to increase the cluster radius R 0 The value is reclassified, if it does not exceed, go to step D52;

[0089] D52 verifies whether the total number of clustering results obtained is lower than the preset lower limit of the total number of clusters. If it exceeds, return to step D1 to reduce the cluster radius R 0 The value of is reclassified, and if it does not exceed, the clustering result is output.

[0090] By adjusting the cluster radius R 0 , the number of clustering types and clustering results of different total amounts can be obtained. In practical applications, clustering results with different similarities under the aforementioned multidimensional features can be determined according to different clustering radii, which provides support for further analysis of the impact of multidimensional features on ship trajectories, as well as classification, management and screening of ship trajectories.

[0091] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention, rather than to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the essence and scope of the technical solution of the present invention.

Claims

1. A ship track restoration method, characterized in that: The following steps are included Step 1: Get the original data set Collect ship operation data and route environment data from the historical data of the route to be analyzed; shipping data includes ship static parameters and ship dynamic navigation parameters; route environment data includes water temperature environment parameters and weather environment parameters; Step 2: ship track node correction, including steps B1 to B5; B1. Determine the unmatched nodes step: extract the ship operation data and route environment data corresponding to each node, and determine the nodes that do not match the ship's own parameters and route environment facts; B2. Establishing route network and ship trajectory expression: Based on the current route construction situation, a route network directed graph G(V, E) is established, where V = (v a |a=1,2...A) represents the set of route nodes, E=(e b |b=1,2...B) represents the set of flight segments, where a flight segment refers to a route segment between two route nodes; based on the specific size of the route network and the computing power of the computing platform, it is divided into several network grids, and the intersecting routes in each grid are determined; Based on the ship navigation data, a ship trajectory expression consisting of trajectory nodes is established. m }, where P m ={p m,i } represents the mth ship trajectory, p m,i ={x m,i ,y m,i ,t m,i ,v m,i ,u m,i } represents the feature vector of the ith track point in the mth ship track, x m,i and m,i Indicates the latitude and longitude values, t m,i Indicates the time node, v m,i Indicates the ship speed, u m,i Indicates heading; B3. Suspected segment positioning: For a certain trajectory point p m,i , taking its network grid as the center, select R m,i ×R m,i The flight segments in the range constitute the suspected flight segment E m,i ; Where R m,i is a variable selected based on the actual distance between the current trajectory point and the nearby flight segments. m,i ×R m,i If there is no suspected segment within the range, R needs to be increased. m,i , if R m,i ×R m,i If the number of suspected flight segments exceeds a certain limit, R needs to be reduced. m,i , to ensure R m,i ×R m,i The number of suspected flight segments within the range is not less than 1 and does not exceed the preset maximum number; Analysis of Suspected Segment E m,i The direction of each segment in the trajectory is related to the ship at the node p m,i If the angle exceeds 90°, it means that the direction of the segment is inconsistent with the actual heading of the ship and it will be removed; B4. Track node positioning: Based on the suspected flight segment E after eliminating invalid data m,i ={e m,i,1 ,e m,i,2 ...e m,i,c ...e m,i,C }, calculate the trajectory point p respectively m,i The probability of dual belonging to each suspected segment is: where e m,i,c represents the trajectory point p m,i The cth suspected segment; ρ represents the standard deviation of ship positioning; r m,i,c is the trajectory point p m,i To the cth suspected segment e m,i,c The distance; α m,i,c is the ship at the trajectory point p m,i The heading of the cth suspected segment e m,i,c The deviation angle of the heading; γ1 is the distance weight coefficient, γ2 is the heading weight coefficient, γ1+γ2=1; B5. Track node correction: According to the calculation results, the one with the highest probability of belonging is selected as the actual route, or several suspected routes with the highest probability of belonging are selected in order to be identified by the expert system or further screened to determine the actual route; after determining the actual route, the ship's track node is mapped to the corresponding route using the vertical mapping method; according to the route segment with the track point p m,i The feature vectors of the adjacent front and rear nodes are related to the trajectory point p m,i The eigenvector p m,i ={x m,i ,y m,i ,t m,i ,v m,i ,u m,i } to make corrections.

2. The ship track restoration method according to claim 1, characterized in that: The static parameters of the ship include route information, time node information, and ship parameter information; the dynamic navigation parameters include ship speed and heading information, engine operating status information, and fuel consumption information; the water temperature environment parameters include tides, flow rate, and water level; the weather environment parameters include wind direction and speed, air pressure, and weather conditions.

3. The ship track restoration method according to claim 1, characterized in that: The situations in step B1 that do not match the ship's own parameters include: the node speed and heading exceed the maximum speed of the ship, the heading exceeds the allowed heading, the engine parameters are abnormal, and the fuel consumption information is abnormal; the situations that do not match the route environment facts include: the coordinate data is obviously abnormal, the hydrological environment parameters are obviously inconsistent with the real-time records, and the weather environment is obviously inconsistent with the meteorological monitoring data.

4. The ship track restoration method according to claim 1, characterized in that: ρ is 2 to 10; for closer distances, r m,i,c The straight-line distance is used for calculation. For the longer distance, r m,i,c Great circle distance is used for calculation and all flight segments should use the same distance calculation method.

5. The ship track restoration method according to claim 1, characterized in that: The narrower the channel and the more complex the channel curvature, the smaller γ1 and the larger γ2.

6. The ship trajectory clustering method according to claim 1, characterized in that: The following steps are also included: Step 3: Ship trajectory fusion distance expression, including C1~C4; C1. Obtain the characteristic parameters of each node of each ship trajectory, and obtain the ship trajectory expressed by the multi-dimensional characteristic vector of each node; C2. For any two multidimensional feature vectors expressing the trajectory {P m } and {Q n }, remove the overlapping common nodes, and connect the remaining nodes in the way of shortest Euclidean distance to obtain the trajectory {P m } and {Q n The fusion trajectory of all nodes in {R m∪n }; C3, the trajectory {P m }={P1,P2...P m ...P M Each node P in m Along the fusion trajectory {R m∪n } after being connected to form a local fusion trajectory {P' m }; Set the trajectory {Q n }={Q1,,Q2...Q n ...Q N Each node Q in n Along the fusion trajectory {R m∪n } after being connected to form a local fusion trajectory {Q' n }; C4. Calculate the fusion distance expression between nodes of the local fusion trajectory Among them, L m∪n represents the fusion trajectory {R m∪n }; d(a, b) represents the Euclidean distance between nodes a and b; S' m represents the fusion trajectory {P' m The time complexity of the trajectory consisting of the 1st to Mth nodes in}, L' n represents the fusion trajectory {Q' n The time complexity of the trajectory consisting of the 1st to nth nodes in}; Step 4: Ship trajectory fusion clustering Based on the feature fusion distance of each node in any two trajectories obtained in step 3, the trajectory nodes are clustered using the fusion distance as a metric, specifically including D1 to D5: D1, determine the minimum number of trajectory points J0 required for a valid trajectory in each trajectory classification result, and initialize the clustering radius R0; D2, each trajectory node point i As the center, search for all nodes whose fusion distance with node i does not exceed R0 to form a valid node set {point i,1 ,point i,2 ...point i,T }, count the number of nodes in the valid node set T; D3, extract the valid node set of T≥J0 corresponding to the trajectory nodes to form the central node set {point' i }; with the central node set {point' i } in an element point' i.1 Traverse the central node set for the initial node, search for other elements with a fusion distance no greater than R0 to form a central node sequence {point' i.1 ,point' i.2 ...point' i.x }; D4, connect the central nodes in sequence to form a classification trajectory {Line i }, for the node {point' i.1 ,point' i.2 ...point' i.x }, all nodes within its radius R0 are considered a cluster; D5. Determine the classification results, delete duplicate clustering results, filter out node elements that have not been classified into any cluster, and perform the following verification: D50 verifies whether the total number of undivided node elements or the proportion of the total number of nodes exceeds the preset noise threshold. If so, return to step D1 and increase the value of the clustering radius R0 to reclassify; if not, go to step D51; D51 verifies whether the total number of clustering results obtained exceeds the preset upper limit of the total number of clusters. If so, return to step D1 to increase the value of the cluster radius R0 and reclassify. If not, go to step D52; D52 verifies whether the total number of clustering results obtained is lower than the preset lower limit of the total number of clusters. If it exceeds, return to step D1 to reduce the value of the clustering radius R0 and reclassify. If it does not exceed, output the clustering results.

Citation Information

Patent Citations

  • Channel traffic information prediction method and application based on multidata fusion

    CN104091470A

  • Track generation and optimization method and system based on space-time probability road tree

    CN114625141A

  • Multi-agent collaborative route planning method and system considering radar threat in three-dimensional environment

    CN117850471A

  • Sea area ship trajectory prediction method and system based on Beidou system

    CN118259316A

  • Ship image trajectory tracking and prediction method based on ship heading recognition

    US20240013402A1

Cited By

  • Rapid simplification method for ship wandering AIS track

    CN119829691A

  • Multi-source data fusion-based radar-assisted camera ship positioning method and device

    CN119936869A

  • Intelligent scheduling method and system applied to marine transportation assembly box

    CN119990947A