Ship port intelligent multi-ship encounter risk early warning method and system

Abstract

The invention relates to the technical field of intelligent shipping, and discloses an intelligent multi-ship encounter risk early warning method and system for a ship port, and the method comprises the steps: collecting and fusing ship dynamic data, environment data, channel static data and VHF voice communication data in real time, obtaining a fusion data set, inputting a space-time diagram into a graph neural network, and carrying out the real-time collection of the fusion data set; capturing mutual influence among ships through a graph attention mechanism, learning an evolution law of ship motion through a time sequence convolutional layer, and outputting a depth spatial-temporal feature embedding vector of each node; inputting the depth spatial-temporal feature embedded vector into a Transform model based on an attention mechanism, and identifying the risk level and type by predicting the ship trajectory in the future 5 minutes; according to the determined risk level and type, automatically generating early warning information including early warning reasons, risk influence analysis and recommendation operation suggestions; hierarchical pushing is carried out through a message queue, an API interface and a visual interface, and highlight display is carried out on the electronic chart; the shipping supervision efficiency is obviously improved.

Description

Technical Field

[0001] This invention relates to the field of intelligent shipping technology, specifically to a method and system for early warning of risks of multiple ships encountering each other in ports. Background Technology

[0002] With the growth of international trade demand, the flow of ships in high-density waterways and port areas has surged, leading to frequent accidents such as collisions, groundings, and deviations from the waterway, causing huge property losses and marine pollution. Current mainstream Vessel Traffic Service (VTS) systems mainly rely on radar, AIS, and other equipment to collect data, which has significant limitations: the crude classification of ships leads to a high false alarm rate, such as fishing boat groups being easily misjudged as being too close together; traditional algorithms cannot accurately capture the dynamic interaction relationships and motion evolution patterns between ships, resulting in lagging risk prediction and insufficient accuracy. Summary of the Invention

[0003] The purpose of this invention is to solve the above-mentioned problems by designing a method and system for early warning of risks of multiple ships meeting in ports.

[0004] The first aspect of this invention provides a method for intelligent multi-ship encounter risk early warning in ports, the method comprising the following steps: Real-time collection and fusion of ship dynamic data, environmental data, waterway static data, and VHF voice communication data to obtain a fused dataset; Based on the fused dataset, a spatiotemporal graph is constructed by treating each ship in the water as a graph node and the edges between nodes as spatial interaction relationships between ships. The spatiotemporal graph is input into the graph neural network, and the mutual influence between ships is captured through the graph attention mechanism. The evolution law of ship motion is learned through the temporal convolutional layer, and the deep spatiotemporal feature embedding vector of each node is output. By embedding deep spatiotemporal features into vector inputs to an attention-based Transformer model, risk levels and types can be identified by predicting ship trajectories within the next 5 minutes. Based on the determined risk level and type, an early warning message is automatically generated, which includes the reason for the warning, risk impact analysis, and recommended operational suggestions. Tiered push notifications are delivered via message queues, API interfaces, and visual interfaces, and highlighted on electronic nautical charts.

[0005] Optionally, in a first implementation of the first aspect of the present invention, the real-time acquisition and fusion of ship dynamic data, environmental data, waterway static data, and VHF voice communication data to obtain a fused dataset includes: Collect ship dynamic data, environmental data, waterway static data, and VHF voice communication data, and convert the VHF voice communication data into VHF text data; Outlier removal is performed on the collected dynamic data, environmental data is smoothed, and VHF text data is segmented and semantically cleaned. A weighted fusion algorithm is used to align ship dynamic data and environmental data by timestamp, associate VHF text data with corresponding ship IDs, and embed waterway static data as spatial background information to form a structured fusion dataset.

[0006] Optionally, in a second implementation of the first aspect of the present invention, the construction of a spatiotemporal graph based on a fused dataset, using each vessel in the waterway as a graph node and the edges between nodes representing spatial interaction relationships between vessels, includes: Extract all ship ID information from the fused dataset, define each ship as an independent node in the spatiotemporal graph, and assign a unique numerical number to each node for differentiation; Each node is assigned static features, dynamic features, and environmental features. For any two nodes, the spatial straight-line distance is calculated based on the position information in the dynamic features. The interaction strength is calculated by combining the relative speed and relative heading. When the distance is less than the preset safety threshold and the interaction strength exceeds the set value, an undirected edge is established between the two nodes, and the weight of the edge is the interaction strength value. Update the dynamic and environmental characteristics of each node, delete edges that exceed the safety threshold, and create new edges between nodes that meet the new interaction conditions to form a spatiotemporal graph that evolves over time.

[0007] Optionally, in a third implementation of the first aspect of the present invention, assigning static features, dynamic features, and environmental features to each node includes: Ship type, gross tonnage, length, and draft are extracted from the fused dataset as static features of the nodes and associated with the corresponding node numbers; The ship's current position, speed, heading, and turning rate are extracted from the fused dataset and synchronously updated to the corresponding nodes according to the timestamp, serving as the dynamic features of the nodes; Environmental parameters corresponding to the ship's current position are extracted from the environmental data of the fused dataset, and associated with the corresponding nodes through spatial coordinate matching, serving as the environmental features of the nodes.

[0008] Optionally, in a fourth implementation of the first aspect of the present invention, the step of inputting the spatiotemporal graph into a graph neural network, capturing the mutual influence between ships through a graph attention mechanism, learning the evolution law of ship motion through a temporal convolutional layer, and outputting the deep spatiotemporal feature embedding vector of each node includes: The constructed spatiotemporal graph is input into a graph neural network, and the attention weight of each node to its neighboring nodes is calculated through a graph attention mechanism. The features of neighboring nodes are weighted and summed based on attention weights, and then fused with the features of the current node to obtain the spatial interaction features of the node. Spatial interaction features at different time steps are arranged in chronological order and input into a temporal convolutional layer. Multiple convolutional kernels slide along the time dimension to capture the short-term trends and long-term patterns of ship motion, and output a feature vector containing temporal evolution information. By concatenating feature vectors of spatial interaction features and temporal evolution information, and compressing the dimensionality through a fully connected layer, a deep spatiotemporal feature embedding vector is generated.

[0009] Optionally, in a fifth implementation of the first aspect of the present invention, the step of embedding depth spatiotemporal features into a vector input to an attention-based Transformer model, and identifying risk levels and types by predicting ship trajectories within the next 5 minutes, includes: The deep spatiotemporal feature embedding vectors are sorted according to time series to form an input sequence containing ship motion features over the past 30 minutes, which is then used as input to the Transformer model. The encoder calculates the association weights of features at different time points in the input sequence through a multi-head self-attention mechanism, and performs a non-linear transformation on the attention-weighted features through a feedforward neural network to output encoded features containing temporal dependencies. The decoder uses the encoded features output by the encoder as a condition to gradually generate predicted values ​​of the ship's position and speed for the next 5 minutes through a masked self-attention mechanism, forming a complete future trajectory sequence. The predicted trajectory is compared with the preset safe waterway to calculate the trajectory deviation value. The nearest meeting point and the time to reach the nearest meeting point are calculated for the predicted trajectories of multiple ships. The risk level and type are determined by combining the trajectory deviation value, the nearest meeting point and the time to reach the nearest meeting point.

[0010] Optionally, in a sixth implementation of the first aspect of the present invention, the step of automatically generating warning information containing the warning reason, risk impact analysis, and recommended operation suggestions based on the determined risk level and type includes: Based on the determined risk level and type, the historical case library is invoked to match the impact scope and potential consequences of similar risk events, and impact analysis content is generated by combining the current environmental characteristics. Based on risk type and impact analysis, operational suggestions are matched from a pre-set response rule base to generate early warning information.

[0011] A second aspect of the present invention provides a smart multi-ship encounter risk early warning system for ships in ports, the system comprising: The acquisition module is used to collect and fuse ship dynamic data, environmental data, waterway static data and VHF voice communication data in real time to obtain a fused dataset; The building module is used to construct a spatiotemporal graph based on the fused dataset, with each ship in the water as a graph node and the edges between nodes representing the spatial interaction relationships between ships. The output module is used to input the spatiotemporal graph into the graph neural network, capture the mutual influence between ships through the graph attention mechanism, learn the evolution law of ship motion through the temporal convolutional layer, and output the deep spatiotemporal feature embedding vector of each node. The identification module is used to embed deep spatiotemporal features into the vector input of the attention-based Transformer model, and to identify the risk level and type by predicting the ship's trajectory in the next 5 minutes. The generation module is used to automatically generate early warning information that includes the cause of the warning, risk impact analysis, and recommended operation suggestions based on the determined risk level and type. The push module is used to push information in a tiered manner through message queues, API interfaces, and visual interfaces, and to highlight it on electronic nautical charts.

[0012] A third aspect of the present invention provides a ship port intelligent multi-ship encounter risk early warning device, the ship port intelligent multi-ship encounter risk early warning device including a memory and at least one processor, the memory storing instructions; the at least one processor calling the instructions in the memory to cause the ship port intelligent multi-ship encounter risk early warning device to perform the various steps of the ship port intelligent multi-ship encounter risk early warning method as described in any of the preceding claims.

[0013] A fourth aspect of the present invention provides a computer-readable storage medium storing instructions that, when executed by a processor, implement the steps of the intelligent multi-ship encounter risk warning method for ships and ports as described in any of the preceding claims.

[0014] The technical solution provided by this invention involves real-time collection and fusion of ship dynamic data, environmental data, waterway static data, and VHF voice communication data to obtain a fused dataset. Based on this fused dataset, a spatiotemporal graph is constructed, with each ship in the waterway as a graph node and edges between nodes representing spatial interactions between ships. This spatiotemporal graph is then input into a graph neural network, where a graph attention mechanism captures the mutual influence between ships, and a temporal convolutional layer learns the evolution of ship motion, outputting a deep spatiotemporal feature embedding vector for each node. This deep spatiotemporal feature embedding vector is then input into a Transformer model based on an attention mechanism to predict ship trajectories within the next 5 minutes, identifying risk levels and types. Based on the determined risk levels and types, a warning reason and risk impact analysis are automatically generated. This invention provides early warning information with recommended operational suggestions; it pushes information in a tiered manner through message queues, API interfaces, and visual interfaces, and highlights it on electronic nautical charts; it ensures data comprehensiveness through multi-source data fusion, solving the problem of incomplete information from a single data source; it uses spatiotemporal diagrams and graph neural networks to accurately capture the spatial interaction relationships and motion evolution patterns between ships, and combines the Transformer model to achieve accurate prediction of future trajectories, significantly reducing the false alarm rate and improving the timeliness and accuracy of risk identification; the generated early warning information includes the cause, impact, and operational suggestions, providing clear guidance for watchkeepers; tiered push and highlighted display on electronic nautical charts ensure targeted responses to different levels of risk, significantly improving shipping supervision efficiency, effectively avoiding navigation accidents in complex waters, and providing solid technical support for ship navigation safety. Attached Figure Description

[0015] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention.

[0016] Figure 1 A flowchart of the intelligent multi-ship encounter risk early warning method for ships in ports provided in this embodiment of the invention; Figure 2 This is a schematic diagram of the structure of the intelligent multi-ship encounter risk early warning system for ships in ports provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of the intelligent multi-ship encounter risk early warning device for ports provided in an embodiment of the present invention. Detailed Implementation

[0017] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” or “having,” and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0018] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 The flowchart of the intelligent multi-ship encounter risk early warning method for ships in ports provided by the embodiments of the present invention includes the following steps: Step 101: Collect and fuse ship dynamic data, environmental data, waterway static data, and VHF voice communication data in real time to obtain a fused dataset; In this embodiment, ship dynamic data, environmental data, waterway static data, and VHF voice communication data are collected, and the VHF voice communication data is converted into VHF text data. Outlier removal is performed on the collected dynamic data, smoothing is performed on the environmental data, and word segmentation and semantic cleaning are performed on the VHF text data. A weighted fusion algorithm is used to align the ship dynamic data and environmental data by timestamp, associate the VHF text data with the corresponding ship ID, and embed the waterway static data as spatial background information to form a structured fusion dataset.

[0019] In this embodiment, dynamic data such as the ship's latitude and longitude position, speed, heading, and turning rate are collected in real time through an AIS receiving device. Environmental data such as wind speed, visibility, tidal height, and water flow speed are collected through meteorological monitoring stations and hydrological sensors. Static data such as water depth, restricted area boundaries, navigation mark positions, and waterway grades are retrieved from the waterway management database. Audio of conversations between ships is collected through a VHF voice recorder. Subsequently, deep learning-based speech recognition technology is used to transcribe the VHF audio data. The ship call sign recognition model is used to match the ship IDs of the two parties in the conversation to generate VHF text data containing timestamps, ship IDs, and dialogue content. Dynamic data outlier removal involves setting reasonable thresholds, such as speeds exceeding the maximum design speed of the vessel type or position jumps exceeding normal navigation distances, to filter out outliers. A sliding window algorithm is then used to identify consecutive outlier sequences. Isolated outliers are replaced by interpolation of data from the preceding and following moments, while consecutive outlier sequences are marked as missing data. Environmental data smoothing employs a moving average method, filtering time-series data such as wind speed and tides in 10-minute windows to eliminate noise caused by short-term sensor fluctuations. VHF text data is first segmented using a Chinese word segmentation tool, then modal particles, repetitive words, and meaningless noise are removed, retaining core semantic information including navigation intentions and position reports. Weights are assigned to dynamic data from different sources based on data reliability. Ship dynamic data and environmental data at the same moment are accurately matched using timestamps to generate time-labeled spatiotemporal data pairs. Based on the ship ID in VHF text data, dialogue content is bound to the corresponding ship's dynamic data to achieve the association between ship behavior and interaction intent. The static waterway data is rasterized according to spatial coordinates, and the waterway attributes such as water depth and no-navigation signs of the grid where the ship is located are matched with the ship's real-time position and added to the ship data with the corresponding timestamp as spatial background features. Finally, a structured fusion dataset containing timestamps, ship IDs, dynamic features, environmental features, interaction intents, and spatial background is formed.

[0020] Step 102: Based on the fused dataset, construct a spatiotemporal graph by treating each ship in the water as a graph node and the edges between nodes as spatial interaction relationships between ships. In this embodiment, all ship ID information is extracted from the fused dataset, and each ship is defined as an independent node in the spatiotemporal graph. A unique numerical number is assigned to each node for identification. Static features, dynamic features, and environmental features are assigned to each node. For any two nodes, the spatial straight-line distance is calculated based on the position information in the dynamic features. The interaction strength is calculated by combining the relative speed and relative heading. When the distance is less than a preset safety threshold and the interaction strength exceeds a set value, an undirected edge is established between the two nodes, and the weight of the edge is the interaction strength value. The dynamic features and environmental features of each node are updated. Edges that exceed the safety threshold are deleted, and new edges are created between nodes that newly meet the interaction conditions, forming a spatiotemporal graph that evolves over time.

[0021] In this embodiment, ship type, gross tonnage, length, and draft are extracted from the fused dataset as static features of the node and associated with the corresponding node number; the current position, speed, heading, and turning rate of the ship are extracted from the fused dataset and synchronously updated to the corresponding node according to the timestamp as dynamic features of the node; the environmental parameters corresponding to the current position of the ship are extracted from the environmental data of the fused dataset and associated with the corresponding node through spatial coordinate matching as environmental features of the node.

[0022] Step 103: Input the spatiotemporal graph into the graph neural network, capture the mutual influence between ships through the graph attention mechanism, learn the evolution law of ship motion through the temporal convolutional layer, and output the deep spatiotemporal feature embedding vector of each node. In this embodiment, the constructed spatiotemporal graph is input into a graph neural network, and the attention weight of each node to its neighboring nodes is calculated through a graph attention mechanism. Based on the attention weights, the features of the neighboring nodes are weighted and summed, and then fused with the features of the current node to obtain the spatial interaction features of the node. The spatial interaction features at different time steps are arranged in chronological order and input into a temporal convolutional layer. Multiple convolutional kernels slide along the time dimension to capture the short-term trends and long-term patterns of ship motion, and output a feature vector containing temporal evolution information. The feature vector containing the spatial interaction features and the feature vector containing temporal evolution information is concatenated, and the dimensionality is compressed through a fully connected layer to generate a deep spatiotemporal feature embedding vector.

[0023] Step 104: Embed the deep spatiotemporal features into the vector input of the Transformer model based on the attention mechanism, and identify the risk level and type by predicting the ship trajectory in the next 5 minutes; In this embodiment, the deep spatiotemporal feature embedding vectors are sorted according to time series to form an input sequence containing ship motion features from the past 30 minutes, which serves as the input to the Transformer model. The encoder calculates the correlation weights of features at different time points in the input sequence through a multi-head self-attention mechanism, and performs a nonlinear transformation on the attention-weighted features through a feedforward neural network to output encoded features containing temporal dependencies. The decoder uses the encoded features output by the encoder as a condition to gradually generate predicted ship positions and speeds for the next 5 minutes through a masked self-attention mechanism, forming a complete future trajectory sequence. The predicted trajectories are compared with the preset safe waterway to calculate the trajectory deviation value, and the nearest meeting point and the time of arrival at the nearest meeting point for the predicted trajectories of multiple ships are calculated. Combining the trajectory deviation value, the nearest meeting point, and the time of arrival at the nearest meeting point, the risk level and type are determined.

[0024] Step 105: Based on the determined risk level and type, automatically generate early warning information that includes the cause of the warning, risk impact analysis, and recommended operational suggestions; In this embodiment, based on the determined risk level and type, the historical case library is invoked to match the impact scope and potential consequences of similar risk events, and combined with the current environmental characteristics to generate impact analysis content; based on the risk type and impact analysis, operational suggestions are matched from the preset response rule library to form early warning information.

[0025] In this embodiment, a set of cases consistent with the current risk type, such as collision, grounding, and deviation from the channel, is retrieved from a historical case database. This case database contains detailed records of each historical event, including the risk level, environmental characteristics at the time of occurrence, types and numbers of vessels involved, actual impact range, and potential consequences. By calculating the similarity between the current risk and historical cases, including risk level matching degree, environmental characteristic overlap degree, and vessel type consistency, 3-5 of the most similar cases are selected as references. Common features of their impact range and potential consequences are extracted. Then, combined with the specific characteristics of the current environment, such as whether the current channel is a busy waterway, whether there are ecological protection zones nearby, and whether the vessel involved is a dangerous goods vessel, targeted adjustments are made to finally generate the impact analysis content. The pre-defined response rule base is divided into sub-bases based on risk type, such as collision risk, grounding risk, and deviation from the shipping lane risk. Each sub-base contains standardized operational suggestions for different scenarios, and each suggestion is associated with key parameters in the impact analysis, such as the size of the impact range, whether special vessels are involved, and the degree of environmental hazard. Based on the current risk type, the system locates the corresponding sub-base, and then selects suitable operational suggestions based on the core information in the impact analysis, prioritizing measures with the highest degree of matching to the current scenario, while supplementing with additional suggestions for special situations. Finally, the selected operational suggestions are sorted by execution priority and integrated into a complete early warning information.

[0026] Step 106: Push messages in a tiered manner through message queues, API interfaces, and visualization interfaces, and highlight them on the electronic nautical chart.

[0027] In this embodiment, based on the risk level determined by the risk identification module, such as no risk, low risk, medium risk, medium-high risk, high risk, and extremely high risk, a tiered push operation is performed: For low-risk events, they are only recorded and archived through the system's backend logs in the visual interface for later retrospective query; medium-risk events are simultaneously pushed to the duty officer's PC-side visual monitoring interface, displaying core warning information in the form of pop-up reminders; medium-high risk and above events initiate multi-channel collaborative push, achieving high-concurrency data transmission through message queues, pushing structured warning instructions to the maritime regulatory backend system, and connecting to the ship's bridge terminal and the duty officer's mobile APP through API interfaces to ensure that front-line operators receive them in real time; at the same time, all risk events are differentiated and highlighted on the electronic nautical chart, with low risk marked in yellow, medium risk marked in orange, and high risk and above marked in red. The marking content includes the ship ID, risk type, and warning time. Clicking on the marking will display a pop-up window to view the complete warning reason, risk impact analysis, and recommended operation suggestions. It supports dynamic playback of risk trajectories controlled by the time axis, helping regulatory personnel to intuitively grasp the risk situation and respond quickly.

[0028] Please see Figure 2A schematic diagram of the structure of the intelligent multi-ship encounter risk early warning system for ports provided in this embodiment of the invention. The system includes: The acquisition module is used to collect and fuse ship dynamic data, environmental data, waterway static data and VHF voice communication data in real time to obtain a fused dataset; The building module is used to construct a spatiotemporal graph based on the fused dataset, with each ship in the water as a graph node and the edges between nodes representing the spatial interaction relationships between ships. The output module is used to input the spatiotemporal graph into the graph neural network, capture the mutual influence between ships through the graph attention mechanism, learn the evolution law of ship motion through the temporal convolutional layer, and output the deep spatiotemporal feature embedding vector of each node. The identification module is used to embed deep spatiotemporal features into the vector input of the attention-based Transformer model, and to identify the risk level and type by predicting the ship's trajectory in the next 5 minutes. The generation module is used to automatically generate early warning information that includes the cause of the warning, risk impact analysis, and recommended operation suggestions based on the determined risk level and type. The push module is used to perform tiered push notifications via message queues, API interfaces, and visual interfaces, and to highlight them on electronic nautical charts. Figure 3 This is a schematic diagram of the structure of a smart multi-ship encounter risk warning device for ships and ports provided in an embodiment of the present invention. The smart multi-ship encounter risk warning device 300 for ships and ports can vary considerably due to different configurations or performance. It may include one or more central processing units (CPUs) 310 (e.g., one or more processors) and a memory 320, and one or more storage media 330 (e.g., one or more mass storage devices) for storing application programs 333 or data 332. The memory 320 and storage media 330 can be temporary or persistent storage. The program stored in the storage media 330 may include one or more modules (not shown in the diagram), each module may include a series of instruction operations on the smart multi-ship encounter risk warning device 300 for ships and ports. Furthermore, the processor 310 may be configured to communicate with the storage media 330 and execute the series of instruction operations in the storage media 330 on the smart multi-ship encounter risk warning device 300 for ships and ports to implement the method provided in the above embodiment.

[0029] The intelligent multi-ship encounter risk early warning device 300 for ships and ports may also include one or more power supplies 340, one or more wired or wireless network interfaces 350, one or more input / output interfaces 360, and / or one or more operating systems 331, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 3 The structure of the intelligent multi-ship encounter risk early warning device for ports shown does not constitute a limitation on the computer equipment provided by the present invention. It may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0030] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the various steps of the intelligent multi-ship encounter risk warning method for ships and ports provided in the above embodiments.

[0031] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described equipment or apparatus / unit can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0032] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0033] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent multi-ship encounter risk early warning in ports, characterized in that, The method includes the following steps: Real-time collection and fusion of ship dynamic data, environmental data, waterway static data, and VHF voice communication data to obtain a fused dataset; Based on the fused dataset, a spatiotemporal graph is constructed by treating each ship in the water as a graph node and the edges between nodes as spatial interaction relationships between ships. The spatiotemporal graph is input into the graph neural network, and the mutual influence between ships is captured through the graph attention mechanism. The evolution law of ship motion is learned through the temporal convolutional layer, and the deep spatiotemporal feature embedding vector of each node is output. By embedding deep spatiotemporal features into vector inputs to an attention-based Transformer model, risk levels and types can be identified by predicting ship trajectories within the next 5 minutes. Based on the determined risk level and type, an early warning message is automatically generated, which includes the reason for the warning, risk impact analysis, and recommended operational suggestions. Tiered push notifications are delivered via message queues, API interfaces, and visual interfaces, and highlighted on electronic nautical charts.

2. The intelligent multi-ship encounter risk early warning method for ships in ports as described in claim 1, characterized in that, The real-time collection and fusion of ship dynamic data, environmental data, waterway static data, and VHF voice communication data yields a fused dataset, including: Collect ship dynamic data, environmental data, waterway static data, and VHF voice communication data, and convert the VHF voice communication data into VHF text data; Outlier removal is performed on the collected dynamic data, environmental data is smoothed, and VHF text data is segmented and semantically cleaned. A weighted fusion algorithm is used to align ship dynamic data and environmental data by timestamp, associate VHF text data with corresponding ship IDs, and embed waterway static data as spatial background information to form a structured fusion dataset.

3. The intelligent multi-ship encounter risk early warning method for ships in ports as described in claim 1, characterized in that, The method, based on a fused dataset, constructs a spatiotemporal graph by treating each vessel in the water as a graph node and the edges between nodes as spatial interaction relationships between vessels, including: Extract all ship ID information from the fused dataset, define each ship as an independent node in the spatiotemporal graph, and assign a unique numerical number to each node for differentiation; Each node is assigned static features, dynamic features, and environmental features. For any two nodes, the spatial straight-line distance is calculated based on the position information in the dynamic features. The interaction strength is calculated by combining the relative speed and relative heading. When the distance is less than the preset safety threshold and the interaction strength exceeds the set value, an undirected edge is established between the two nodes, and the weight of the edge is the interaction strength value. Update the dynamic and environmental characteristics of each node, delete edges that exceed the safety threshold, and create new edges between nodes that meet the new interaction conditions to form a spatiotemporal graph that evolves over time.

4. The intelligent multi-ship encounter risk early warning method for ships in ports as described in claim 2, characterized in that, The process of assigning static features, dynamic features, and environmental features to each node includes: Ship type, gross tonnage, length, and draft are extracted from the fused dataset as static features of the nodes and associated with the corresponding node numbers; The ship's current position, speed, heading, and turning rate are extracted from the fused dataset and synchronously updated to the corresponding nodes according to the timestamp, serving as the dynamic features of the nodes; Environmental parameters corresponding to the ship's current position are extracted from the environmental data of the fused dataset, and associated with the corresponding nodes through spatial coordinate matching, serving as the environmental features of the nodes.

5. The intelligent multi-ship encounter risk early warning method for ships in ports as described in claim 1, characterized in that, The process involves inputting the spatiotemporal graph into a graph neural network, capturing the interactions between ships through a graph attention mechanism, learning the evolution of ship motion through temporal convolutional layers, and outputting a deep spatiotemporal feature embedding vector for each node, including: The constructed spatiotemporal graph is input into a graph neural network, and the attention weight of each node to its neighboring nodes is calculated through a graph attention mechanism. The features of neighboring nodes are weighted and summed based on attention weights, and then fused with the features of the current node to obtain the spatial interaction features of the node. Spatial interaction features at different time steps are arranged in chronological order and input into a temporal convolutional layer. Multiple convolutional kernels slide along the time dimension to capture the short-term trends and long-term patterns of ship motion, and output a feature vector containing temporal evolution information. By concatenating feature vectors of spatial interaction features and temporal evolution information, and compressing the dimensionality through a fully connected layer, a deep spatiotemporal feature embedding vector is generated.

6. The intelligent multi-ship encounter risk early warning method for ships in ports as described in claim 1, characterized in that, The process of embedding deep spatiotemporal features into a vector input to an attention-based Transformer model, and identifying risk levels and types by predicting ship trajectories within the next 5 minutes, includes: The deep spatiotemporal feature embedding vectors are sorted according to time series to form an input sequence containing ship motion features over the past 30 minutes, which is then used as input to the Transformer model. The encoder calculates the association weights of features at different time points in the input sequence through a multi-head self-attention mechanism, and performs a non-linear transformation on the attention-weighted features through a feedforward neural network to output encoded features containing temporal dependencies. The decoder uses the encoded features output by the encoder as a condition to gradually generate predicted values ​​of the ship's position and speed for the next 5 minutes through a masked self-attention mechanism, forming a complete future trajectory sequence. The predicted trajectory is compared with the preset safe waterway to calculate the trajectory deviation value. The nearest meeting point and the time to reach the nearest meeting point of the predicted trajectories of multiple ships are also calculated. The risk level and type are determined by combining the trajectory deviation value, the nearest meeting point and the time to reach the nearest meeting point.

7. The intelligent multi-ship encounter risk early warning method for ships in ports as described in claim 1, characterized in that, The system automatically generates warning information based on the determined risk level and type, including the warning reason, risk impact analysis, and recommended operational suggestions, among other things: Based on the determined risk level and type, the historical case library is invoked to match the impact scope and potential consequences of similar risk events, and impact analysis content is generated by combining the current environmental characteristics. Based on risk type and impact analysis, operational suggestions are matched from a pre-set response rule base to generate early warning information.

8. A smart multi-ship encounter risk early warning system for ships in ports, characterized in that, The system includes: The acquisition module is used to collect and fuse ship dynamic data, environmental data, waterway static data and VHF voice communication data in real time to obtain a fused dataset; The building module is used to construct a spatiotemporal graph based on the fused dataset, with each ship in the water as a graph node and the edges between nodes representing the spatial interaction relationships between ships. The output module is used to input the spatiotemporal graph into the graph neural network, capture the mutual influence between ships through the graph attention mechanism, learn the evolution law of ship motion through the temporal convolutional layer, and output the deep spatiotemporal feature embedding vector of each node. The identification module is used to embed deep spatiotemporal features into the vector input of the Transformer model based on the attention mechanism, and to identify the risk level and type by predicting the ship trajectory within the next 5 minutes. The generation module is used to automatically generate early warning information that includes the cause of the warning, risk impact analysis, and recommended operation suggestions based on the determined risk level and type. The push module is used to push information in a tiered manner through message queues, API interfaces, and visual interfaces, and to highlight it on electronic nautical charts.

9. A smart multi-ship encounter risk early warning device for ships in ports, characterized in that, The intelligent multi-ship encounter risk warning device for ships and ports includes a memory and at least one processor. The memory stores instructions. The at least one processor invokes the instructions in the memory to cause the intelligent multi-ship encounter risk warning device for ships and ports to perform the various steps of the intelligent multi-ship encounter risk warning method for ships and ports as described in any one of claims 1-7.

10. A computer-readable storage medium storing instructions thereon, characterized in that, When the instructions are executed by the processor, they implement each step of the intelligent multi-ship encounter risk warning method for ships and ports as described in any one of claims 1-7.

Images