Marine navigation event processing scheme recommendation method and system based on prediction intention

Abstract

The invention discloses a prediction intention-based marine navigation event processing scheme recommendation method and system, and the method comprises the steps: firstly building an intention prediction model based on a machine learning method, and obtaining the prediction intention of a ship through the intention prediction model according to the historical marine navigation data of the ship; according to the matching relation between intentions and events, the events corresponding to the prediction intentions are recognized, and the events are abnormal marine navigation events; and finally, calculating the similarity between the event and the historical event, and selecting the processing scheme corresponding to the historical event with the highest similarity for recommendation. The method can be widely applied to multiple scenes such as maritime supervision, port management and maritime search and rescue, and greatly improves the shipping safety and management efficiency.

Description

Technical Field

[0001] The present invention relates to the field of maritime navigation incident management, and in particular to a method and system for recommending maritime navigation incident handling solutions based on prediction intentions. Background Art

[0002] Traditional maritime navigation incident management systems mainly rely on manual monitoring and operation. Faced with the growing global maritime traffic, existing systems often face problems such as untimely response and insufficient information processing when handling maritime incidents. Current maritime monitoring methods mostly rely on automatic identification systems (AIS), radars, and manual visual monitoring. At present, some preliminary applications of ship trajectory analysis and behavior detection based on big data and machine learning, such as abnormal navigation detection, are often limited to specific scenarios, but lack intelligent behavior prediction and automated processing capabilities. The shortcomings of existing technologies are 1) delayed event response: the current system usually responds after the event occurs, lacking the ability to warn of potential risks in advance; 2) excessive manual intervention: a large number of maritime navigation event processing relies on the judgment of operators, which is prone to human errors, especially when dealing with complex or emergency events, it is difficult to make accurate and timely decisions. Summary of the invention

[0003] Purpose of the invention: The purpose of the present invention is to provide a method and system for recommending maritime navigation incident handling solutions based on predicted intentions, which can identify maritime navigation incidents through intention prediction and recommend corresponding handling solutions, thereby solving the problems of delayed response and excessive manual intervention in the prior art.

[0004] Technical solution: The method for recommending a marine navigation incident handling solution based on predicted intention of the present invention comprises the following steps:

[0005] (1) Establishing an intention prediction model based on a machine learning method, and using the intention prediction model to obtain the predicted intention of the ship based on the historical maritime navigation data of the ship;

[0006] (2) identifying an event corresponding to the predicted intention based on the matching relationship between the intention and the event, where the event is an abnormal maritime navigation event;

[0007] (3) Calculate the similarity between the event and historical events, and select the processing solution corresponding to the historical event with the highest similarity for recommendation.

[0008] Furthermore, in step (1), historical maritime navigation data of the ship is collected from a number of data sources, including automatic identification system data, satellite remote sensing data, radar data, ship communication data, weather and ocean data, historical behavior pattern data, and sensor data in the ship.

[0009] Furthermore, in step (1), the historical maritime navigation data of the ship is preprocessed, and the preprocessing includes data cleaning, time synchronization, and data normalization.

[0010] Furthermore, in step (1), feature engineering methods are used to extract key features from historical maritime navigation data and train the intention prediction model.

[0011] Furthermore, in step (2), after identifying the event corresponding to the predicted intention based on the matching relationship between the intention and the event, the method further includes: verifying the predicted intention based on the current position and speed of the ship, and eliminating events that are not related to the current position and time of the ship using a spatiotemporal filtering algorithm.

[0012] Furthermore, in step (2), for the identified events, the order of recommended processing solutions is determined according to the event priority. The event priority calculation method is:

[0013] ;

[0014] Among them, P is the event priority, S is the severity of the event, and U is the urgency of the event. and is the weight coefficient.

[0015] Furthermore, the event in step (3) and historical events The similarity between them is:

[0016] ;

[0017] in, For events The most similar historical event, is the similarity function, is the maximum value function.

[0018] The marine navigation incident handling solution recommendation system based on prediction intention of the present invention comprises:

[0019] An intention prediction unit, used to establish an intention prediction model based on a machine learning method, and obtain the predicted intention of the ship using the intention prediction model according to the historical maritime navigation data of the ship;

[0020] An event recognition unit, configured to recognize an event corresponding to the predicted intention according to a matching relationship between the intention and the event, wherein the event is an abnormal marine navigation event;

[0021] The processing solution recommendation unit is used to calculate the similarity between the event and the historical event, and select the processing solution corresponding to the historical event with the highest similarity for recommendation.

[0022] The electronic device described in the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded into the processor, the method for recommending a solution for handling marine navigation events based on predicted intentions is implemented.

[0023] The computer-readable storage medium of the present invention stores a computer program, and when the computer program is executed by a processor, the method for recommending a marine navigation event handling solution based on predicted intention is implemented.

[0024] Beneficial effects: Compared with the prior art, the advantages of the present invention are: (1) The present invention automatically predicts the future behavior intentions of ships by analyzing multi-source data, thereby improving the ability to predict future intentions, discovering potential risks earlier and taking countermeasures in advance; (2) The present invention matches maritime navigation events based on intentions and sorts them according to event priorities to ensure that high-risk events are handled first and reduce the burden of manual intervention; (3) The present invention can be widely used in multiple scenarios such as maritime supervision, port management, and maritime search and rescue, greatly improving shipping safety and management efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 The present invention is a flowchart of a method for recommending a solution for handling marine navigation incidents. DETAILED DESCRIPTION

[0026] The technical solution of the present invention is further described below in conjunction with the accompanying drawings.

[0027] like Figure 1 As shown, the method for recommending a solution for handling marine navigation events based on predicted intentions includes the following steps.

[0028] Step 1: Data collection and preprocessing.

[0029] Step 1-1: Collect ship-related maritime navigation data from multiple data sources in real time and perform data preprocessing to provide accurate input for subsequent model predictions. The innovation of this step lies in the integration of heterogeneous data, which can better integrate data from different sources and improve the accuracy and real-time performance of predictions. Data sources include but are not limited to automatic identification systems, radars, meteorological sensors, etc., as shown in Table 1 below:

[0030] Table 1 Data sources

[0031]

[0032] Step 1-2: Preprocess the collected data, including data cleaning, time synchronization, data normalization and other operations to ensure data quality and consistency.

[0033] 1) Data cleaning: By setting predetermined thresholds or rules, detect and remove outliers and noise data in the original data, including missing data, unreasonable speed or position changes, abnormal signals caused by sensor failure, etc. For data beyond the reasonable range, interpolation or historical data is used to fill in to ensure the integrity and accuracy of the data;

[0034] 2) Time synchronization: For the collection time of different data sources, the timestamps of each data source are calibrated based on a unified time base (such as UTC time). Data with different timestamps are aligned through interpolation or linear extrapolation to make the data of each data source consistent at the same time scale. For example, for data collection systems with different frequencies (such as AIS updated every 2 seconds and radar updated every 1 second), the radar data is time interpolated at 2-second intervals to synchronize it with the AIS data;

[0035] 3) Normalization: Normalize data of different dimensions to ensure that data of different dimensions can be calculated and processed uniformly. Normalization uses the following formula:

[0036] ;

[0037] in, is the standardized data, is the original data, is the mean of the data, is the standard deviation of the data.

[0038] Step 2: Intent prediction.

[0039] Step 2-1: Collect the ship's real-time dynamic data (such as current position, speed, heading, etc.) and historical behavior data (such as past tracks, speed change patterns, etc.) to build a ship behavior database.

[0040] Step 2-2: Use feature engineering methods to extract representative and predictive key features from the ship behavior database, such as speed change rate, heading change rate, and relative distance, etc. These features not only reflect the current dynamic state of the ship, but also reveal its behavior patterns and trend changes.

[0041] Step 2-3: Use the extracted key features to train the prediction model. The prediction model can be a long short-term memory network (LSTM) or a hidden Markov model (HMM). The feature data needs to be input into the prediction model to learn the historical behavior pattern of the ship.

[0042] Taking the long short-term memory network as an example, the input sequence is , the output is the ship's intention , the update formula of the long short-term memory network is:

[0043] ;

[0044] ;

[0045] in is the current time step, is the length of the input sequence (i.e. the length of the historical data window), is the time step length of the prediction (i.e. the number of steps to predict in the future), is the hidden state at time step t (the internal state of the LSTM unit, representing the accumulation of timing information), is the activation function, usually a Sigmoid function, which is used to limit the output range to (0, 1). is the weight matrix between hidden states (indicating the influence of the previous hidden state on the current hidden state), is the weight matrix input to the hidden state (indicating the impact of the input features on the current hidden state), is a bias term used to adjust the output of the model. is the weight matrix from hidden state to output (indicating the influence of hidden state on the final output).

[0046] Step 2-4: Use the prediction model to predict the ship’s intentions by analyzing the ship’s historical trajectory and current status to predict the ship’s future intentions, including normal sailing, port calls, etc.

[0047] By training on historical behavior data, the model can capture the temporal dependencies and underlying patterns of ship behavior, and thus dynamically adjust the prediction algorithm with the assistance of real-time data to improve the accuracy and real-time performance of the prediction.

[0048] The data obtained in real time can also be used for training after being saved and preprocessed (called historical data).

[0049] Step 3: Event identification and priority assessment.

[0050] Step 3-1, event identification: First, by analyzing the prediction results of ship intentions, matching predefined events, Table 2 shows the matching relationship between some intentions and events. In fact, one intention may correspond to multiple events, identifying potential maritime navigation events. Event types include but are not limited to collision risks, route deviations, etc. Then, combined with multi-source data from AIS, radar, meteorological sensors, etc., the identified events are verified to ensure the accuracy and reliability of event identification. Finally, the spatiotemporal filtering algorithm is applied to eliminate events that are not related to the current ship position and time, thereby improving the accuracy of event identification.

[0051] Specifically, the identified events are verified by analyzing the match between the predicted results of the ship's intention and the current position of the ship in the AIS data (expressed in geographic coordinates). The current position refers to the geographic coordinates of the ship obtained through the AIS system, including longitude and latitude. This information is combined with dynamic data such as speed (SOG) and heading (COG) to determine the real-time position and movement trend of the ship.

[0052] The current position of the ship is compared according to the intention category, for example: (1) Navigation intention: The system determines whether the navigation is normal by analyzing the degree of deviation of the current position of the ship from the scheduled route. If the position coordinates deviate from the route range by more than the set threshold, it is considered that the navigation intention does not match the actual behavior. (2) Berthing intention: Determine whether the berthing process is normal based on the distance between the current position of the ship and the port position, as well as the speed and heading trends. If the ship approaches the port but the speed does not decrease, there may be a risk. (3) Collision avoidance intention. By calculating the relative distance and heading trend between ships, it is determined whether the collision avoidance behavior is effective. (4) Determine whether the ship is anchoring as intended based on whether the current position is in the anchoring area and whether the speed is close to zero. If the ship deviates from the anchoring area or still maintains the sailing speed, it is determined that there is a deviation.

[0053] The matching relationship between some intents and events in this embodiment is as follows:

[0054] Table 2 The matching relationship between intent and events

[0055]

[0056] Step 3-2, event classification: First, analyze the attributes of the identified events according to their nature, scope of impact, and urgency, including event type, severity level, and urgency level. Secondly, use machine learning classification algorithms (such as decision trees, support vector machines, etc.) to automatically classify events into corresponding event libraries. Each event library corresponds to a specific processing solution. Finally, assign a priority to each classified event. The priority calculation formula is as follows:

[0057] ;

[0058] Among them, P is the event priority, S is the severity of the event, and U is the urgency of the event. and is the weight coefficient, and the best combination is determined through experiments to reflect the actual application requirements.

[0059] Step 3-3, event library maintenance: Dynamically update the event types and classification standards in the event library based on the feedback information collected during system operation to ensure that the event library always reflects the latest maritime navigation safety requirements. Introduce a self-learning algorithm so that the event library can autonomously adjust classification parameters and priority weights based on newly identified events and processing solutions, thereby improving its adaptive capabilities and intelligence level.

[0060] Step 4: Generate recommended processing solutions.

[0061] Step 4-1: Based on the identified events, determine the processing order of each event according to the event priority. Retrieve historical events that are similar to the current event in feature dimensions from the historical case database. The matching process includes but is not limited to keyword matching, attribute matching, and context association analysis.

[0062] Step 4-2: Use the predefined similarity function to compare the current event and historical events The similarity between them is quantified:

[0063] ;

[0064] in, is the historical event with the highest similarity to the current event. is the similarity function, is a mathematical operator that means finding the independent variable that maximizes the objective function value.

[0065] Step 4-3: Based on the similarity calculation results, select the processing plan corresponding to the historical event with the highest similarity to the current event and recommend it to the user. At the same time, the user can make appropriate adjustments to the processing plan as needed to better adapt to the specific circumstances of the current event.

[0066] In this embodiment, the processing scheme corresponding to the event is shown in Table 3:

[0067] Table 3 Matching relationship between events and processing solutions

[0068]

[0069] Step 5: Feedback and adaptive optimization.

[0070] Step 5-1: After each event is processed, the corresponding feedback data is automatically collected. The feedback data includes but is not limited to the processing success rate, processing time, abnormal information, actual behavior and environmental change data.

[0071] Step 5-2: Use a predefined loss function to evaluate the accuracy of the intent prediction model and treatment plan recommendation. Calculate the results based on the loss function:

[0072] ;

[0073] Where L is the loss function, For actual results, For the prediction results.

[0074] The gradient descent method is used to optimize the parameters in the intent prediction model and the recommended processing solution to improve the accuracy and real-time performance of the system:

[0075] ;

[0076] in, are model parameters, is the learning rate, is the gradient of the loss function with respect to the parameters.

[0077] Step 5-3: According to the optimized intention prediction model and processing plan, the processing plan is adaptively adjusted, including: processing plan evaluation, processing plan update, multi-processing plan fusion, real-time adjustment, etc., to adapt to different navigation environments and ship conditions.

[0078] The marine navigation incident handling solution recommendation system based on prediction intention of the present invention comprises:

[0079] An intention prediction unit, used to establish an intention prediction model based on a machine learning method, and obtain the predicted intention of the ship using the intention prediction model according to the historical maritime navigation data of the ship;

[0080] An event recognition unit, configured to recognize an event corresponding to the predicted intention according to a matching relationship between the intention and the event, wherein the event is an abnormal marine navigation event;

[0081] The processing solution recommendation unit is used to calculate the similarity between the event and the historical event, and select the processing solution corresponding to the historical event with the highest similarity for recommendation.

[0082] The electronic device described in the present invention includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the computer program is loaded into the processor, the method for recommending a solution for handling marine navigation events based on predicted intentions is implemented.

[0083] The computer-readable storage medium of the present invention stores a computer program, and when the computer program is executed by a processor, the method for recommending a marine navigation event handling solution based on predicted intention is implemented.

[0084] The computer-readable storage media may include RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage or other magnetic storage devices, flash memory, or any other medium that can be used to store program code in the form of instructions or data structures and that can be accessed by a computer.

[0085] The processor is used to execute the computer program stored in the memory to implement each step of the method involved in the above embodiment.

Claims

1. A method for recommending a solution for handling maritime navigation events based on predicted intention, characterized in that: The steps include: (1) Establishing an intention prediction model based on a machine learning method, and using the intention prediction model to obtain the predicted intention of the ship based on the historical maritime navigation data of the ship; (2) identifying an event corresponding to the predicted intention based on the matching relationship between the intention and the event, where the event is an abnormal maritime navigation event; (3) Calculate the similarity between the event and historical events, and select the processing solution corresponding to the historical event with the highest similarity for recommendation.

2. The method for recommending a marine navigation incident handling solution based on predicted intention according to claim 1, characterized in that: In step (1), historical maritime navigation data of the ship is collected from a number of data sources, including automatic identification system data, satellite remote sensing data, radar data, ship communication data, weather and ocean data, historical behavior pattern data, and sensor data in the ship.

3. The method for recommending a marine navigation incident handling solution based on predicted intention according to claim 1, characterized in that: In step (1), the historical maritime navigation data of the ship is preprocessed, and the preprocessing includes data cleaning, time synchronization, and data normalization.

4. The method for recommending a marine navigation incident handling solution based on predicted intention according to claim 1, characterized in that: In step (1), feature engineering methods are used to extract key features from historical maritime navigation data and train the intention prediction model.

5. The method for recommending a marine navigation incident handling solution based on predicted intention according to claim 1, characterized in that: In step (2), after identifying the event corresponding to the predicted intention based on the matching relationship between the intention and the event, the method further includes: verifying the predicted intention based on the current position and speed of the ship, and using a spatiotemporal filtering algorithm to eliminate events that are not related to the current position and time of the ship.

6. The method for recommending a marine navigation incident handling solution based on predicted intention according to claim 1, characterized in that: In step (2), for the identified events, the order of recommended processing solutions is determined according to the event priority. The event priority calculation method is: ； Among them, P is the event priority, S is the severity of the event, and U is the urgency of the event. and is the weight coefficient.

7. The method for recommending a marine navigation incident handling solution based on predicted intention according to claim 1, characterized in that: Event in step (3) and historical events The similarity between them is: ； in, For events The most similar historical event, is the similarity function, is the maximum value function.

8. A marine navigation incident handling solution recommendation system based on predicted intention, characterized in that: include: An intention prediction unit, used to establish an intention prediction model based on a machine learning method, and obtain the predicted intention of the ship using the intention prediction model according to the historical maritime navigation data of the ship; An event recognition unit, configured to recognize an event corresponding to the predicted intention according to a matching relationship between the intention and the event, wherein the event is an abnormal marine navigation event; The processing solution recommendation unit is used to calculate the similarity between the event and the historical event, and select the processing solution corresponding to the historical event with the highest similarity for recommendation.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is loaded into a processor, the method for recommending a marine navigation event handling solution based on predicted intention according to any one of claims 1 to 7 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method for recommending a marine navigation event handling solution based on predicted intention according to any one of claims 1 to 7 is implemented.

Images