Ship route optimization method and system based on environmental spatio-temporal changes

Abstract

This invention relates to a method and system for optimizing ship routes based on spatiotemporal environmental changes. The method includes: acquiring a trajectory density image and performing Mercator projection correction; obtaining a trajectory pixel set by filtering trajectory pixels based on color intensity; identifying global route blocks through density-based spatial clustering; calculating local route center feature points for each global route block; constructing a route map that satisfies edge set constraints using the center feature points of the global route blocks as nodes; generating a node sequence as an initial route under route node constraints through route traversal based on depth-first search and node selection based on greedy heuristics; mapping the initial route to a spatiotemporal trajectory; and performing local optimization on the nodes with the goal of minimizing the overall cost to obtain the final ship route. The overall cost includes the heuristic cost of adjacent nodes and the environmental cost of quantifying the impact of the environment on navigation safety and energy consumption.

Description

Technical Field

[0001] This invention relates to the field of ship route planning technology, and in particular to a method and system for optimizing ship routes based on environmental spatiotemporal changes. Background Technology

[0002] With the widespread application of Automatic Identification Systems (AIS), vessel route information in global waters is becoming increasingly publicly available. How to efficiently identify vessel routes and easily derive route design data has become a key concern in the industry. Clearly, color images based on AIS data, as a visualization medium, can display the movement trajectories of vessels in time and space. This offers a new approach to acquiring vessel route information. Graphical images for AIS data visualization provide a means of extracting information from densely populated vessel route areas.

[0003] In recent years, many scholars both domestically and internationally have conducted in-depth research on information extraction and ship routes based on big data. In studies of maritime routes and other related fields, scholars have primarily focused on utilizing large amounts of raw AIS data to explore innovative methods, models, and address key questions. However, this approach often overlooks the rich scientific questions inherent in color images as another form of AIS data representation and a carrier of underlying geographic information. Current information extraction methods have achieved significant results in route information, particularly in the direct processing of AIS data. Simultaneously, these studies also reflect that information extraction based on color images is emerging as a new trend.

[0004] Chinese patent application publication number CN119006373A discloses a method and storage medium for extracting ship routes from trajectory images. By combining 3D data extraction and bilinear interpolation techniques with edge detection and Mercator projection, it solves the problems of subjective dependence and complexity in existing ship route extraction methods, achieving efficient and automated route extraction and analysis. However, it does not fully consider color information in the image during trajectory pixel search, leaving room for improvement in recognition performance.

[0005] In summary, there is currently a lack of a method and system for optimizing ship routes to solve or partially solve the aforementioned problems. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the existing technology and provide a method and system for optimizing ship routes based on environmental spatiotemporal changes. This addresses or partially addresses the problems of existing automatic ship route design technologies, which mainly focus on mining raw AIS data, neglecting the advantages of color images as information carriers, and lacking adaptability to complex water structures and dynamic environmental changes.

[0007] The objective of this invention can be achieved through the following technical solutions: One aspect of the present invention provides a method for optimizing ship routes based on spatiotemporal environmental changes, comprising the following steps: The trajectory density image is acquired and Mercator projection correction is performed. The trajectory pixel set is obtained by filtering the trajectory pixels based on color intensity. Based on the set of trajectory pixels, global route blocks are identified by density-based spatial clustering, and local route center feature points are calculated for each global route block. Using the central feature point of the global route block as a node, a route graph that satisfies the edge set constraint is constructed. Through route traversal based on depth-first search and node selection based on greedy heuristic, a node sequence is generated under the route node constraint condition, which serves as the initial route. The initial route is mapped to a spatiotemporal trajectory. With the goal of minimizing the overall cost, local optimization is performed on the nodes to obtain the final ship route. The overall cost includes the heuristic cost of adjacent nodes and the environmental cost of quantifying the impact of the environment on navigation safety and energy consumption.

[0008] As a preferred technical solution, the Mercator projection correction includes the following steps: For each pixel in the trajectory density image, the corresponding longitude value is obtained based on a pre-established pixel-geographic coordinate mapping relationship along the longitude direction; For the pixels in the trajectory density image, the latitude coordinates are mapped to the Mercator projection longitudinal coordinates. The Mercator projection coordinates corresponding to the minimum and maximum latitudes are calculated respectively to construct a Mercator projection coordinate space. A linear mapping relationship between the longitudinal pixel coordinates and the projection coordinates is established in the Mercator projection coordinate space. The projection coordinates are then inversely calculated to obtain the corresponding latitude values.

[0009] As a preferred technical solution, the process of selecting trajectory pixels based on color intensity includes the following steps: For each pixel in the trajectory density image after Mercator projection correction, the red, green and blue channel intensities are extracted. Based on preset weighting coefficients representing the contribution of different color channels to trajectory information, a trajectory saliency parameter is calculated. Based on the trajectory saliency parameter and a preset trajectory discrimination threshold, it is determined whether to add the current pixel to the trajectory pixel set.

[0010] As a preferred technical solution, the process of identifying global route blocks based on density-based spatial clustering and calculating local route center feature points for each global route block includes the following steps: The trajectory pixel set is globally clustered by density-based spatial clustering, and a global cluster set is constructed through density reachability relationship. Each cluster corresponds to a global route block, representing a macroscopic route block or intersection region. For each global route block, local refined clustering is performed using MiniBatch KMeans to obtain the local route center feature points corresponding to each global route block.

[0011] As a preferred technical solution, the route traversal based on depth-first search and the node selection based on greedy heuristics include the following steps: On the route map, a path traversal is performed using depth-first search to obtain a set of potential route branches that satisfy the route node constraints; During the node expansion process of the path traversal, the heuristic cost is the great circle distance from the current node to the target node, and the neighboring node with the smallest heuristic cost is selected for search.

[0012] As a preferred technical solution, the route node constraints include spatial legality constraints, turning angle constraints, path non-repetition constraints, and minimum segment length constraints.

[0013] As a preferred technical solution, the overall cost is: in, , , These are the overall cost, the heuristic cost of adjacent nodes, and the environmental cost. This represents the great circle distance between two nodes. This represents the number of nodes in the initial flight path. This represents the wind field influence function. This represents the wave field influence function. Represents the flow field influence function. These are the weighting coefficients. Adjacent nodes The sailing time between For the planned speed of the ship, For a moment The ship is located at the node In the corresponding environmental state, Indicates the first Each route node Indicates wind field, Indicates the significant wave height. This represents the ocean current vector field.

[0014] As a preferred technical solution, the process of performing local optimization on the nodes includes the following steps: When the environmental cost of a certain segment of the route exceeds a preset threshold, the corresponding node is locally adjusted, a neighborhood search space is constructed for the current node, the environmental cost of the node at the future arrival time is recalculated within the neighborhood, and the node with the smaller overall cost is selected to replace the original node, thereby achieving local optimization of the route.

[0015] As a preferred technical solution, the edge set constraint conditions include distance constraints, route connectivity constraints, and heading continuity constraints.

[0016] Another aspect of the present invention provides a ship route optimization system based on spatiotemporal environmental changes, for implementing the aforementioned ship route optimization method, the system comprising: The image correction and trajectory pixel recognition module is used to acquire trajectory density images and perform Mercator projection correction, and obtain a trajectory pixel set by filtering trajectory pixels based on color intensity; The global route block identification module is used to identify global route blocks based on the trajectory pixel set by density-based spatial clustering, and to calculate local route center feature points for each global route block. The initial route generation module is used to construct a route map that satisfies the edge set constraints by using the central feature point of the global route block as a node. It generates a node sequence as the initial route by traversing the route based on depth-first search and selecting nodes based on greedy heuristics under the route node constraints. The optimization module is used to map the initial route into a spatiotemporal trajectory and perform local optimization on the nodes with the goal of minimizing the overall cost to obtain the final ship route. The overall cost includes the heuristic cost of adjacent nodes and the environmental cost of quantifying the impact of the environment on navigation safety and energy consumption.

[0017] Compared with the prior art, the present invention has at least one of the following beneficial effects: (1) Fully consider the color information in the image in the trajectory pixel search: This invention obtains the trajectory density image and performs Mercator projection correction, and obtains the trajectory pixel set by screening the trajectory pixels based on color intensity. It fully considers the characteristic that the color intensity of the ship's historical trajectory is positively correlated with the trajectory density in the unit area in the AIS trajectory density image, thereby accurately extracting the effective trajectory pixels in the image and eliminating the interference of background areas and low-frequency noise pixels.

[0018] (2) Realize route optimization based on environmental spatiotemporal changes: After obtaining the initial route, the present invention maps the initial route to a spatiotemporal trajectory. With the goal of minimizing the overall cost, the nodes are locally optimized to obtain the final ship route. The invention fully considers the impact of the synchronous changes of the marine environment where the ship is located on the safety and economy of the ship's navigation by time and space. The time dimension of the environmental field is introduced to dynamically evaluate and locally adjust the route.

[0019] (3) Achieving accurate calculation of ship routes: Based on the trajectory pixel set, this invention identifies global route blocks through density-based spatial clustering. For each global route block, it calculates the local route center feature point, achieving accurate extraction of route features in complex waters with route intersections, parallel navigation, and the coexistence of discrete noise. Using the center feature point of the global route block as a node, a route map satisfying the edge set constraint is constructed. Through route traversal based on depth-first search and node selection based on greedy heuristics, a node sequence is generated under the route node constraint condition as the initial route, realizing the accurate construction of a continuous and feasible route from the origin port to the destination port. Attached Figure Description

[0020] Figure 1 This is a flowchart of the ship route optimization method based on environmental spatiotemporal changes in the embodiment; Figure 2 This is a schematic diagram showing the results of trajectory point information extraction based on graphic images in the embodiment; Figure 3 This is a schematic diagram of the pixel clustering results based on trajectory point information in the embodiment; Figure 4 This is a schematic diagram of the route feature extraction results based on double-layer nested clustering in the embodiment; Figure 5 This is a route design example from the embodiments; Figure 6 This is another example of route design in the embodiments; Figure 7 This is a schematic diagram of a ship route optimization system based on environmental spatiotemporal changes in the embodiment. Detailed Implementation

[0021] 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, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0022] Example 1 Existing automated ship route design technologies primarily focus on mining raw AIS data, neglecting the advantages of color images as information carriers. Furthermore, traditional route design methods lack adaptability to complex waterway structures and dynamic environmental changes. This embodiment provides a ship route optimization method based on spatiotemporal environmental variations. This method utilizes ship trajectory density image features to construct a global route topology network through global-local two-layer clustering, using this network as the bottom-layer solution space. This solves the technical problems of discontinuous route extraction and poor environmental adaptability in complex sea areas, achieving automated route generation that balances familiar routes with dynamic environmental constraints.

[0023] See Figure 1 This method includes the following steps: Step S1: Pixel geospatial matching.

[0024] Because the Mercator projection exhibits a characteristic where the latitudinal scale varies with latitude—meaning the actual length corresponding to a unit latitude differs at different latitudes—directly employing a linear mapping between pixel coordinates and latitude would increase spatial positioning errors in high-latitude regions, affecting the accuracy of route extraction, clustering, and flight path optimization results. Considering these issues, this step provides a pixel latitude-longitude matching method that takes into account the latitudinal asymptotic rate characteristic of the Mercator projection.

[0025] Obtain the spatial parameters of the map sheet and standardize the angle system. Obtain the geographical latitude and longitude range covered by the Mercator chart, including the minimum longitude. Maximum longitude Minimum latitude and maximum latitude And obtain the horizontal pixel count of the corresponding image. With vertical pixel count The latitude and longitude coordinates are converted from degrees to radians as follows: (1) (1) Establish the pixel-geographic coordinate mapping relationship in the longitude direction.

[0026] Because the Mercator projection maintains a linear scaling relationship in the longitude direction, for any lateral pixel coordinate... Its corresponding longitude Determined using linear interpolation: (2) The longitude is then converted from radians to degrees. (3) (2) Latitude direction nonlinear matching based on Mercator projection.

[0027] To address the scale variation characteristic of the Mercator projection along the latitudinal direction, the latitudinal coordinates are first mapped to the Mercator projection longitudinal coordinates. The calculation formula is as follows: (4) Calculate the minimum latitude respectively With maximum latitude Corresponding Mercator projection coordinates: (5) (6) Establish vertical pixel coordinates within the Mercator projection coordinate space. With projected coordinates Linear mapping relationship between (7) Subsequently, the projected coordinates are inversely calculated to obtain the corresponding latitude values: (8) (3) Trajectory pixel filtering based on color intensity.

[0028] In AIS trajectory density images, historical ship trajectories are typically encoded using high-brightness or high-saturation colors, with color intensity positively correlated with trajectory density per unit area. To eliminate interference from background areas and low-frequency noise pixels, color thresholding is first applied to the image. Let the original trajectory density image be... (9) in Here are the pixel coordinates, and R, G, and B represent the intensity of the red, green, and blue channels, respectively.

[0029] definition The weighting coefficients, representing the contributions of different color channels to the trajectory information, are then used to define the trajectory significance function: (10) Set trajectory discrimination threshold When the following conditions are met: When the specified value is reached, the pixel is determined to be a valid trajectory pixel. After threshold filtering, the set of trajectory pixels is obtained: (11) in These are the latitude and longitude coordinates of the corresponding pixel.

[0030] Step S2: Extract route features based on double-layer nested clustering.

[0031] To address the challenges of intersecting routes, parallel navigation, and the coexistence of discrete noise in complex waterways, this step proposes a global-local two-layer nested clustering method to extract stable route structure features.

[0032] (1) Global route block identification based on DBSCAN.

[0033] First, the density-based spatial clustering algorithm DBSCAN is used to cluster the trajectory pixel set. Global clustering is performed to eliminate discrete noise and divide potential route blocks.

[0034] Define arbitrary trajectory points of - Neighborhood is: (12) in This represents the great circle distance based on latitude and longitude. If it satisfies... Then the point This is identified as a core point. A global cluster set is constructed based on density reachability relationships. (13) Each cluster It represents a macroscopic route block or intersection area.

[0035] (2) Local route center extraction based on MiniBatch KMeans.

[0036] In each global route block To further extract the route center structure, the MiniBatchKMeans algorithm is introduced for local refined clustering. Let the blocks... Includes Given _trace points, the number of clusters is set to _t_. The objective function is: (14) in For the first j Local cluster centers. MiniBatch KMeans updates the clusters iteratively using mini-batch samples: (15) While ensuring computational efficiency, stable cluster centers are obtained. The final set of all cluster centers is defined as follows: (16) Each node Characterizes a central feature point of a flight path.

[0037] Step S3: Design of ship routes based on node constraints.

[0038] After completing the route feature extraction, a set of route center feature points representing the historical route structure has been obtained. To achieve the construction of a continuous and feasible route from the origin port to the destination port, this step provides a ship route design method based on route node constraints. This method comprehensively adopts depth-first search (DFS), a greedy heuristic strategy, and route node constraints to achieve automatic route generation.

[0039] (1) Construction of route node diagram model.

[0040] The set of route center feature points obtained during the route feature extraction stage is represented as: (17) Each node This refers to a route node that has physical navigational significance.

[0041] Construct an undirected graph based on the spatial adjacency relationships between nodes. The construction of the edge set E satisfies the following constraints: 1. Distance constraints.

[0042] Any two nodes Great circle distance between satisfy 2. Route connectivity constraints.

[0043] Node connections must be located within an identified route block to avoid crossing low-density or non-navigable areas; 3. Heading continuity constraints.

[0044] The turning angle of adjacent sides does not exceed the set threshold. .

[0045] Node pairs that meet the above conditions This forms a candidate route.

[0046] (2) Route traversal mechanism based on depth-first search.

[0047] On the route chart G, a depth-first search (DFS) strategy is used to traverse the routes to ensure connectivity and integrity.

[0048] Let the starting node be Target node Define the path status as: (18) The DFS search rules are as follows: 1. Starting from the starting node Depart and push it onto the path stack; 2. At the current node Among all reachable adjacent nodes, a recursive search is performed sequentially; 3. If the current path contains the target node If so, it is determined to be a candidate route; 4. If no feasible expansion node is found, backtrack to the previous node and continue the search.

[0049] DFS search can traverse all potential route branches that satisfy connectivity constraints.

[0050] (3) Node selection strategy based on greedy heuristic.

[0051] To improve search efficiency and guide the flight path towards the target direction, a greedy heuristic function is introduced during the DFS search process to prioritize candidate nodes. The current node is defined. To the target node The heuristic cost function is: (19) in The great circle distance between two nodes is represented by the Haversine formula: (20) During node expansion, the neighboring node with the lowest heuristic cost is selected for search first, thereby achieving a greedy search behavior that approaches the target.

[0052] (4) Comprehensive determination of route node constraints.

[0053] During the DFS and greedy search processes, the addition of each node must simultaneously satisfy the following route node constraints: 1. Spatial legality constraints.

[0054] The node is located in a high-density area of ​​a historical air route or within its safe buffer zone; 2. Steering angle constraint.

[0055] Suppose three consecutive nodes Its heading angle satisfy ; 3. Path non-repetition constraint.

[0056] Avoid repeated node visits and prevent path loops; 4. Minimum segment length constraint.

[0057] The distance between adjacent nodes is not less than a set threshold to ensure the smoothness and maneuverability of the flight path.

[0058] A candidate node is allowed to join the current route only if it meets all of the above constraints.

[0059] (5) Output the route generation results.

[0060] Through a joint search mechanism combining DFS, greedy algorithm, and node constraints, the final output route can be represented as a sequence of nodes: (twenty one) Step S4: Route optimization based on environmental spatiotemporal changes.

[0061] Ship navigation is essentially a spatial displacement process that progresses over time. During this process, the marine environment in which the ship is located, including wind fields, wave fields, and current fields, changes synchronously with time and space. If route design is carried out only under static environmental assumptions, it will be difficult to reflect real navigation conditions, thus affecting navigation safety and economy.

[0062] Therefore, this step provides a route optimization method based on environmental spatiotemporal changes, which dynamically evaluates and locally adjusts the route by introducing the time dimension of the environmental field.

[0063] (1) Spatiotemporal parameterization of flight routes.

[0064] Let the initial route generated based on node constraints be... ,in Indicates the first i Each route node. Based on the ship's planned speed. V The flight path is parameterized in time, mapping it to a spatiotemporal trajectory: (twenty two) any adjacent nodes The sailing time between them is: (twenty three) The arrival time of the node is calculated recursively as follows: (twenty four) (2) Construction and mapping of the environmental spatiotemporal field.

[0065] set up Indicates wind field; Indicates the significant wave height; Let the ocean current vector field be represented, then the spatiotemporal field of the marine environment is: (25) At any time The ship is located at the node The corresponding environmental state is: Spatiotemporal interpolation methods are used to map discrete environmental data to the corresponding time and spatial locations of flight route nodes.

[0066] (3) Construction of the environmental cost function.

[0067] To quantify the impact of the environment on navigation safety and energy consumption, a comprehensive environmental cost function is constructed: (26) in, Represents the wind field influence function; This represents the wave field influence function; Represents the flow field influence function; These are the weighting coefficients.

[0068] (4) Comprehensive route optimization objective function.

[0069] Taking environmental factors into account, a comprehensive objective function for route optimization is constructed: (27) (5) Local route adjustment based on spatiotemporal environmental constraints.

[0070] When the environmental cost of a segment of the flight path exceeds a set threshold, a local adjustment is made to the corresponding node. Let the current node be... The neighborhood search space is: (28) Recalculate the future arrival time of nodes within the neighborhood. Given the environmental costs, choose to satisfy The original node is replaced by the new node to achieve local optimization of the route.

[0071] The method will be described below with specific examples.

[0072] Step 1: Obtaining ship route information based on trajectory density images.

[0073] A ship track density image was generated based on historical AIS data, selecting major international shipping lanes in the Indian Ocean (covering the Gulf of Aden-Arabian Sea-Bay of Bengal region). The image resolution was set to 2000×1500 pixels, with a latitude and longitude coverage of 0°-135°E and 40°S-40°N. Ship track density exhibits a significant bright banded distribution within the major international shipping lanes, with high-density areas closely coinciding with the actual main shipping routes. This image is a ship track density map with continuous color transitions, gradually shifting from green and blue to yellow and orange, culminating in a deep red highlight on the busiest shipping routes, visually reflecting the spatial distribution of ship track density from low to high. Spatial matching of pixels was performed on the original ship track density image to obtain the spatial distribution of pixels in latitude and longitude.

[0074] Further color thresholding was performed on the trajectory density image to extract pixels that met the trajectory saliency condition. The results are as follows: Figure 2 As shown in the diagram, the Strait of Malacca-Indian Ocean-Suez Canal constitutes the most prominent, continuous, and densest main shipping corridor in the overall trajectory distribution. This corridor converges significantly from the South China Sea towards the Strait of Malacca at its eastern end, forming a highly concentrated trajectory convergence area near Singapore. After entering the Indian Ocean, the route extends through the Bay of Bengal and south of Sri Lanka to the Arabian Sea, with a clear overall direction and stable width. At its western end, it further passes through the Gulf of Aden into the Red Sea and connects to the Suez Canal, forming a global high-density shipping corridor connecting Asia and Europe. It exhibits outstanding spatial continuity and high traffic intensity, and can be considered a typical high-density habitual shipping route. Based on this main route, several secondary high-density regional shipping routes are also distributed within the Indian Ocean. These are particularly prominent in the waters surrounding the western coast of the Indian Peninsula and Sri Lanka, where multiple routes fan out northwestward from the south of Sri Lanka, connecting major Indian ports with Middle Eastern and East African coastal countries. A significant route convergence area is formed in southern Sri Lanka, demonstrating its pivotal and diversionary role in the Indian Ocean shipping system. Meanwhile, the shipping route from the Persian Gulf through the Strait of Hormuz to the Arabian Sea also exhibits high density characteristics, with concentrated tracks and consistent directions. It primarily serves oil tanker and liquefied natural gas transport. Although the spatial range is limited, the traffic intensity per unit area is extremely high, making it a typical high-intensity narrow-channel route. In contrast, the track density in the open waters south of the equator to approximately 30°S in the southern Indian Ocean is generally lower. Tracks are distributed in a wide band with low overlap, reflecting more the detours and direct routes chosen by ships during long-distance voyages based on weather avoidance or route optimization, and has not yet formed a significantly concentrated, habitual route structure.

[0075] Step 2: Global-local trajectory point clustering and route structure extraction.

[0076] First, based on the spatial matching and route pixel information extraction of the above images, the DBSCAN algorithm is used to perform global clustering of trajectory points, setting the neighborhood radius parameter ε to 5 nautical miles and the minimum number of points MinPts=50. For example... Figure 3 As shown, the clustering results divide the trajectory points in the Indian Ocean into several macroscopic route blocks and effectively remove discrete noise points.

[0077] Based on global clustering, MiniBatch KMeans is used for further localized clustering of each route block. The number of clusters is adaptively set according to the block size, with K set to 100. This extracts the route center nodes, with a node spacing of approximately 10–20 nautical miles. The results are as follows: Figure 4As shown, all cluster centers together constitute the route skeleton structure, forming a route boundary constraint zone on its outer side. The results indicate that the extracted route width is basically consistent with the actual international waterway width, with the main route width concentrated in the range of 20–40 nautical miles.

[0078] Step 3: Route design based on route node constraints like Figure 5 As shown, Case 1 selects the route cluster center node number 52 as the starting point, geographically located west of the Strait of Malacca and near the entrance to the Andaman Sea, belonging to a typical outbound convergence area from Southeast Asia to the Indian Ocean; the ending point is the cluster center node number 63, located in the southwestern southern Indian Ocean, close to the intersection of shipping routes southeast of South Africa, corresponding to an important channel node for the southward extension of the Asia-Europe-Africa route. Both starting and ending points are located in the central areas of habitually used shipping routes with high trajectory density, possessing good route representativeness. In this case, based on the route node graph model, a depth-first search (DFS) combined with a greedy heuristic strategy is used to generate the Fletcher route. During the search process, the maximum connectivity distance between nodes is set to approximately 30 nautical miles, the turning angle constraint is set to no more than 30°, and a greedy criterion of a constant heading line towards the target point is introduced. The generated route extends along the high-density shipping route zone in the central Indian Ocean, with gentle changes in heading, avoiding crossing low-density trajectory areas. Compared with the theoretical shortest constant heading line, the total distance of the route generated in Case 1 increases by approximately 2.5%–3.0%. The difference stems primarily from the fact that the algorithm prioritizes route legality and navigation safety constraints during the node selection phase, rather than simply shortening the geometric distance. This makes the route more closely aligned with historically frequent traffic channels, resulting in higher engineering feasibility and navigation stability.

[0079] like Figure 6 As shown, Case 2 starts at cluster center node 87, located in the western part of the southern Indian Ocean, near the southeastern coast of Africa, corresponding to one of the starting nodes of major ocean shipping routes in the Southern Hemisphere; the ending point is cluster center node 106, located east of the Strait of Malacca, near the Sunda Islands, a crucial shipping exit node connecting the Indian Ocean and the western Pacific. This combination of start and end points represents a typical transoceanic navigation scenario in the southern Indian Ocean-Southeast Asia direction. During route generation, a greedy criterion of a great circle towards the target point is introduced, causing the algorithm to be more frequently constrained by heading continuity and turning angle constraints during the node expansion stage, resulting in a slightly curved overall shape for the generated route. The route actively approaches the high-density distribution zone of shipping route nodes south of Sri Lanka and near the equator, avoiding entering the sparsely populated waters of the southern region. Results show that this route is approximately 3%–4% longer than the shortest great circle route. This is mainly because in the mid-to-low latitude transition region, the algorithm selects multiple secondary route nodes for transition in order to satisfy route continuity and node connectivity, thereby significantly improving the stability and controllability of the route while sacrificing a small amount of travel distance.

[0080] Step 4: Route optimization based on environmental spatiotemporal changes.

[0081] The spatiotemporal distribution of the Indian Ocean environment in the generated case waters from January to June 2024 reflects the coordinated changes in wind, wave, and current fields under a typical monsoon background. As the months progressed, the overall Indian Ocean environment exhibited a significant temporal evolution pattern: at the beginning of the year, wave and wind fields were relatively uniformly distributed south of the equator; however, in late spring and early summer, influenced by the strengthening monsoon, continuous zonal distributions of high wind speeds and high wave heights gradually formed in the Arabian Sea, the central Indian Ocean, and areas near the equator, significantly increasing spatial environmental differences. Particularly within the latitude range of 10°–25°S, high-wave areas and strong countercurrent areas exhibited strong spatial stability and temporal persistence, significantly impacting transoceanic navigation safety and energy consumption.

[0082] In this context, the initial route generated in step three was time-parameterized. Assuming the ship travels at a constant speed of 14 knots, the estimated arrival times of each node on the route were calculated segment by segment, and the node times were matched with the corresponding environmental spatiotemporal fields. Analysis results show that the initial route significantly overlaps with high-wave, strong-wind, and counter-current areas in several key segments, resulting in high exposure to adverse environments and significant potential risks and energy costs. After introducing an environmental cost function, local optimization adjustments were implemented for the route nodes corresponding to the aforementioned high-risk segments. This optimization process, while maintaining the continuity of the original route structure and habitual channel constraints, only made minor adjustments to the positions of local nodes, ensuring the route avoids areas of concentrated environmental risk in both time and space. The optimization results show that the cumulative exposure time of the route in high-wave and counter-current areas was reduced by approximately 15%–25%, the overall environmental cost was reduced by approximately 18%, and the change in the total route distance was controlled within 2%.

[0083] Based on the optimization and verification logic of a single route described above, this embodiment further extends the method to the entire Indian Ocean, constructing a global route topology network oriented towards environmental spatiotemporal optimization. This network is not a superposition of individual routes, but rather a set of safe routes retained after extracting cluster centers and removing high-risk routes that have been long-term affected by harsh environments. This mesh framework serves as the underlying reference solution space for environmental and spatiotemporal optimization; any new route planning can be generated and dynamically adjusted based on this environmentally adaptable network framework.

[0084] Example 2 Based on Example 1, this example provides a ship route optimization system based on environmental spatiotemporal changes, used to implement the ship route optimization method of Example 1. See [link to example]. Figure 7 The system includes: (1) Image correction and trajectory pixel recognition module. Used to acquire trajectory density images and perform Mercator projection correction, and obtain a trajectory pixel set by filtering trajectory pixels based on color intensity.

[0085] (2) Global route block identification module. It is used to identify global route blocks based on the trajectory pixel set by density-based spatial clustering, and calculate the local route center feature point for each global route block.

[0086] (3) Initial route generation module. It is used to construct a route map that satisfies the edge set constraint condition by using the central feature point of the global route block as the node. It generates a node sequence as the initial route under the route node constraint condition by traversing the route based on depth-first search and selecting the node based on greedy heuristic.

[0087] (4) Optimization module. It is used to map the initial route to a spatiotemporal trajectory, and to perform local optimization on the nodes with the goal of minimizing the overall cost, so as to obtain the final ship route. The overall cost includes the heuristic cost of adjacent nodes and the environmental cost of quantifying the impact of the environment on navigation safety and energy consumption.

[0088] In summary, the present invention has the following characteristics: (1) This invention is based on the habitual routes chosen by ships over a long period of time, and uses ship trajectory density images to represent historical traffic intensity, thereby naturally introducing group navigation experience and implicit safety redundancy in the route design stage. Compared with traditional pure path optimization algorithms that are prone to traversing low-traffic waters, this invention prioritizes generating routes along high-density channels, improving route navigability and safety reliability.

[0089] (2) Compared with the approach of starting from the ship's AIS trajectory, the route driven by the image is expressed in a continuous spatial form, making the route boundary, intersection area and passage corridor present an explicit visual structure. Combined with threshold segmentation and clustering, features such as the route center and width can be stably extracted from the image, improving the efficiency and consistency of route recognition.

[0090] (3) For high-density complex waterways with intersections, parallelism and noise, this method transforms the image phenomenon of multiple overlapping routes into a structured set of nodes through global-local two-layer clustering, avoiding the problems of misconnection, breakage or deviation in the intersection area caused by traditional single clustering or pure geometric optimization, thereby improving the correctness of route topology and engineering usability.

[0091] (4) This method constructs a global route topology network optimized for environmental spatiotemporal conditions, which serves as the underlying reference solution space after avoiding high-risk routes. Based on this backbone network, ship routes can be automatically designed and dynamically adjusted, significantly improving environmental adaptability and engineering availability while taking into account habitual routes. The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for optimizing ship routes based on spatiotemporal environmental changes, characterized in that, Includes the following steps: The trajectory density image is acquired and Mercator projection correction is performed. The trajectory pixel set is obtained by filtering the trajectory pixels based on color intensity. Based on the set of trajectory pixels, global route blocks are identified by density-based spatial clustering, and local route center feature points are calculated for each global route block. Using the central feature point of the global route block as a node, a route graph that satisfies the edge set constraint is constructed. Through route traversal based on depth-first search and node selection based on greedy heuristic, a node sequence is generated under the route node constraint condition, which serves as the initial route. The initial route is mapped to a spatiotemporal trajectory. With the goal of minimizing the overall cost, local optimization is performed on the nodes to obtain the final ship route. The overall cost includes the heuristic cost of adjacent nodes and the environmental cost of quantifying the impact of the environment on navigation safety and energy consumption.

2. The method for optimizing ship routes based on spatiotemporal environmental changes according to claim 1, characterized in that, The Mercator projection correction includes the following steps: For each pixel in the trajectory density image, the corresponding longitude value is obtained based on a pre-established pixel-geographic coordinate mapping relationship along the longitude direction; For the pixels in the trajectory density image, the latitude coordinates are mapped to the Mercator projection longitudinal coordinates. The Mercator projection coordinates corresponding to the minimum and maximum latitudes are calculated respectively to construct a Mercator projection coordinate space. A linear mapping relationship between the longitudinal pixel coordinates and the projection coordinates is established in the Mercator projection coordinate space. The projection coordinates are then inversely calculated to obtain the corresponding latitude values.

3. The method for optimizing ship routes based on spatiotemporal environmental changes according to claim 1, characterized in that, The process of selecting trajectory pixels based on color intensity includes the following steps: For each pixel in the trajectory density image after Mercator projection correction, the red, green and blue channel intensities are extracted. Based on preset weighting coefficients representing the contribution of different color channels to trajectory information, a trajectory saliency parameter is calculated. Based on the trajectory saliency parameter and a preset trajectory discrimination threshold, it is determined whether to add the current pixel to the trajectory pixel set.

4. The method for optimizing ship routes based on spatiotemporal environmental changes according to claim 1, characterized in that, The density-based spatial clustering identification of global route blocks, and the process of calculating local route center feature points for each global route block, includes the following steps: The trajectory pixel set is globally clustered by density-based spatial clustering, and a global cluster set is constructed through density reachability relationship. Each cluster corresponds to a global route block, representing a macroscopic route block or intersection region. For each global route block, local refined clustering is performed using MiniBatch KMeans to obtain the local route center feature points corresponding to each global route block.

5. The method for optimizing ship routes based on spatiotemporal environmental changes according to claim 1, characterized in that, The route traversal based on depth-first search and the node selection based on greedy heuristics include the following steps: On the route map, a path traversal is performed using depth-first search to obtain a set of potential route branches that satisfy the route node constraints; During the node expansion process of the path traversal, the heuristic cost is the great circle distance from the current node to the target node, and the neighboring node with the smallest heuristic cost is selected for search.

6. The method for optimizing ship routes based on spatiotemporal environmental changes according to claim 1, characterized in that, The constraints on the route nodes include spatial legality constraints, turning angle constraints, path non-repetition constraints, and minimum segment length constraints.

7. The method for optimizing ship routes based on spatiotemporal environmental changes according to claim 1, characterized in that, The overall cost is: in, , , These are the overall cost, the heuristic cost of adjacent nodes, and the environmental cost. This represents the great circle distance between two nodes. This represents the number of nodes in the initial flight path. This represents the wind field influence function. This represents the wave field influence function. This represents the flow field influence function. These are the weighting coefficients. Adjacent nodes The sailing time between For the planned speed of the ship, For a moment The ship is located at the node In the corresponding environmental state, Indicates the first Each route node Indicates wind field, Indicates the significant wave height. This represents the ocean current vector field.

8. The method for optimizing ship routes based on spatiotemporal environmental changes according to claim 1, characterized in that, The process of performing local optimization on nodes includes the following steps: When the environmental cost of a certain segment of the route exceeds a preset threshold, the corresponding node is locally adjusted, a neighborhood search space is constructed for the current node, the environmental cost of the node at the future arrival time is recalculated within the neighborhood, and the node with the smaller overall cost is selected to replace the original node, thereby achieving local optimization of the route.

9. A method for optimizing ship routes based on spatiotemporal environmental changes according to claim 1, characterized in that, The edge set constraints include distance constraints, route connectivity constraints, and heading continuity constraints.

10. A ship route optimization system based on spatiotemporal environmental changes, characterized in that, The system for implementing the ship route optimization method as described in any one of claims 1-9 includes: The image correction and trajectory pixel recognition module is used to acquire trajectory density images and perform Mercator projection correction, and obtain a trajectory pixel set by filtering trajectory pixels based on color intensity; The global route block identification module is used to identify global route blocks based on the trajectory pixel set by density-based spatial clustering, and to calculate local route center feature points for each global route block. The initial route generation module is used to construct a route map that satisfies the edge set constraints by using the central feature point of the global route block as a node. It generates a node sequence as the initial route by traversing the route based on depth-first search and selecting nodes based on greedy heuristics under the route node constraints. The optimization module is used to map the initial route into a spatiotemporal trajectory and perform local optimization on the nodes with the goal of minimizing the overall cost to obtain the final ship route. The overall cost includes the heuristic cost of adjacent nodes and the environmental cost of quantifying the impact of the environment on navigation safety and energy consumption.

Images