Safety identification and inspection algorithm and system for unmanned ships and boats based on artificial intelligence

Abstract

The present invention relates to the field of safety identification technology, specifically to an artificial intelligence-based unmanned ship and boat safety identification inspection algorithm and system, comprising the following steps: based on the collected sonar, radar and camera data, data integration is performed, and multiple data sources are synchronized through timestamp comparison to generate an integrated environmental data body. In the present invention, through the joint application of spectrum analysis and MaskR‑CNN, obstacle recognition is not limited to basic forms, but can also be classified in real time according to dynamic changes, which greatly enhances the accuracy of recognition and the timeliness of operation. In addition, the introduction of the decision tree algorithm allows the risk assessment strategy to be dynamically adjusted according to real-time data and environmental changes, providing a reliable safety decision support for unmanned boats and ships, optimizing the data processing and analysis process, so that unmanned boats can effectively perform obstacle avoidance tasks under a wider range of conditions, and improve the automation level and safety of navigation.

Description

Technical Field

[0001] The present invention relates to the field of safety identification technology, and in particular to an artificial intelligence-based unmanned ship and boat safety identification inspection algorithm and system. Background Art

[0002] The field of safety identification technology in unmanned ships and boats involves the use of sensors, cameras and other data acquisition equipment to continuously monitor and analyze the surrounding environment. Its core functions include real-time detection of surface and underwater obstacles, other ships and navigation markers. By integrating artificial intelligence and machine learning models, it is possible to predict potential collision risks and automatically adjust the course or speed to avoid accidents, ensuring that unmanned ships can operate safely in complex and changing environments.

[0003] Among them, the unmanned ship and boat safety identification and inspection algorithm involves the development and implementation of specialized algorithms for unmanned ships and boats to automatically perform safety monitoring and inspection tasks, enabling unmanned ships to identify and respond to potential security threats, such as approaching ships, floating objects, and other factors that may affect the safe navigation of ships. It has a wide range of uses, including commercial transportation, marine research, and recreational navigation, aiming to reduce manpower requirements, improve safety, and optimize the operating efficiency of ships. This technology enables unmanned ships to navigate autonomously and perform complex tasks, greatly improving the safety and reliability of maritime operations.

[0004] Traditional methods rely on a single data source and static risk assessment, which cannot provide accurate obstacle identification and risk prediction in a rapidly changing environment, resulting in an inability to effectively respond to emergencies. The lack of data integration capabilities results in poor performance in multi-source information fusion, making it difficult to achieve efficient data synchronization and time correction, affecting the response speed and accuracy of decision-making, resulting in unmanned vessels facing higher safety risks and operational efficiency issues in actual operations. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an artificial intelligence-based unmanned ship and boat safety identification and inspection algorithm and system.

[0006] In order to achieve the above purpose, the present invention adopts the following technical solution: based on artificial intelligence unmanned ship and boat safety identification and inspection algorithm, including the following steps:

[0007] S1: Based on the collected sonar, radar and camera data, data integration is performed, and data synchronization of multiple data sources is performed through timestamp comparison to generate an integrated environmental data body;

[0008] S2: Based on the environmental data integration, the sonar and radar data are first analyzed by spectrum analysis, and then the camera data is processed in combination with the Mask R-CNN algorithm, and static and dynamic obstacles are distinguished according to the features to obtain an obstacle characteristic classification map;

[0009] S3: Based on the obstacle characteristic classification diagram, the size and distance of the obstacle are measured, and the risk level of the obstacle is determined by using a decision tree algorithm according to the comparison between the measurement results and the preset safety parameters, and each obstacle classification is risk-labeled to generate an obstacle risk label;

[0010] S4: Based on the obstacle risk label, analyze multiple potential obstacle avoidance strategies, including heading adjustment and speed control, and execute control instructions through a proximal strategy optimization algorithm to adjust the navigation path of the unmanned ship and boat to obtain a route adjustment plan;

[0011] S5: Based on the route adjustment plan, update the navigation control strategy, adjust the heading and speed, implement the new navigation command, monitor the command execution status, verify whether the navigation of the unmanned ship and boat is updated, and generate a navigation status update record;

[0012] S6: Based on the navigation status update record, navigation data is collected, path changes and speed adjustments before and after navigation are compared, and the matching degree between the navigation status and the expected obstacle avoidance effect is evaluated to generate a navigation safety assessment log.

[0013] As a further solution of the present invention, the environmental data integration body includes sonar sound wave signal data, radar electromagnetic wave echo data, and camera visual image data; the obstacle characteristic classification map includes obstacle type identification, relative position, and moving speed information; the obstacle risk label includes the risk level, distance parameter, and size estimate of each obstacle; the route adjustment plan includes a predetermined heading change, speed adjustment amount, and obstacle avoidance strategy options; the navigation status update record includes updated GPS coordinates, speed change records, and heading angle adjustment data; the navigation safety assessment log includes path change records, speed change analysis, and risk matching level.

[0014] As a further solution of the present invention, based on the environmental data integration, the sonar and radar data are first analyzed by spectrum analysis, and then the camera data is processed in combination with the MaskR-CNN algorithm, and static and dynamic obstacles are distinguished according to the characteristics. The steps of obtaining the obstacle characteristic classification map are specifically as follows:

[0015] S201: Based on the environmental data integration body, perform spectrum analysis on the sonar and radar data, separate key frequency components, and extract key signal features to generate a spectrum feature set;

[0016] S202: Based on the spectral feature set, using the Mask R-CNN algorithm, identifying the image data captured by the camera, identifying the contour of the object in the image, distinguishing static and dynamic objects through shape analysis, and generating edge recognition data;

[0017] S203: Based on the edge recognition data, combined with the signal characteristics of sonar and radar, the identified objects are classified and marked as static or dynamic obstacles, and a risk assessment is performed on each object to obtain an obstacle characteristic classification map.

[0018] As a further solution of the present invention, based on the obstacle characteristic classification diagram, the size and distance of the obstacle are measured, and the risk level of the obstacle is judged by using a decision tree algorithm according to the comparison between the measurement results and the preset safety parameters, and each obstacle classification is risk labeled. The steps of generating the obstacle risk label are specifically as follows:

[0019] S301: Based on the obstacle characteristic classification map, measure the straight-line distance and size of the obstacles marked in the map, and record the size and distance parameters of each obstacle to generate an obstacle measurement record;

[0020] S302: Based on the obstacle measurement record, using a decision tree algorithm, comparing with the safe operation parameters of the boat, calculating the obstacle risk level that may cause a collision, assessing the potential threat to navigation safety, and generating an obstacle risk assessment record;

[0021] S303: Based on the obstacle risk assessment record, mark the risk level of each obstacle, including low, medium and high, and mark the risk information on the navigation map to generate an obstacle risk label.

[0022] As a further solution of the present invention, based on the obstacle risk label, a variety of potential obstacle avoidance strategies are analyzed, including heading adjustment and speed control, and control instructions are executed through a proximal strategy optimization algorithm to adjust the navigation paths of unmanned ships and boats. The specific steps of obtaining the route adjustment plan are:

[0023] S401: Based on the obstacle risk label, identify the path area that needs to avoid obstacles, formulate a preliminary obstacle avoidance strategy for each risk level of obstacles, including heading deviation and speed adjustment strategy, and generate an obstacle avoidance planning scheme;

[0024] S402: Based on the obstacle avoidance planning scheme, determine control instructions, including heading change angle and speed adjustment parameters, prepare update instructions for navigation control, and generate navigation control parameters;

[0025] S403: Based on the navigation control parameters, a proximal strategy optimization algorithm is used to adjust the navigation paths of the unmanned ships and boats, and the adjusted headings and speeds are applied to ensure that the ships navigate along the optimized paths, thereby generating a route adjustment plan.

[0026] As a further solution of the present invention, the proximal strategy optimization algorithm is according to the formula:

[0027]

[0028] Calculate probability ratios;

[0029] Where s is the current state of the ship, v is the current speed, d is the distance between the target and the current position, θ is the angle between the ship and the target, ω is the weight coefficient dynamically adjusted according to the changes in the navigation environment, and P new (a|s, v, d, θ) is the probability of selecting action a given state s under the new strategy, P old (a|s, v, d, θ) is the probability under the old strategy.

[0030] As a further solution of the present invention, based on the route adjustment plan, the navigation control strategy is updated, the heading and speed are adjusted, the new navigation command is implemented, and the command execution status is monitored to verify whether the navigation of the unmanned ship and the boat is updated. The steps of generating the navigation status update record are specifically as follows:

[0031] S501: Based on the route adjustment plan, adjusting the navigation control strategy parameters, including inputting new heading angle and speed settings, adjusting the automatic navigation settings, matching the predetermined obstacle avoidance route, and generating a heading speed configuration record;

[0032] S502: Based on the heading and speed configuration record, activate the navigation command, track the command execution effect in real time, record each adjustment result of the heading and speed, and generate an execution status monitoring log;

[0033] S503: Based on the execution status monitoring log, evaluate whether the navigation status is updated according to the new parameters, compare and analyze the navigation path with the plan, check the execution status of the new route, and generate a navigation status update record.

[0034] The artificial intelligence-based unmanned ship and boat safety identification and inspection system is used to execute the artificial intelligence-based unmanned ship and boat safety identification and inspection algorithm, and the system includes:

[0035] The data integration module collects data from sonar, radar and cameras, excludes abnormal data, performs time correction on data points, merges data streams from different sources, unifies data formats and creates a single data interface to generate an integrated environmental data body;

[0036] The obstacle recognition module uses the environmental data integration to analyze sonar and radar data through spectrum analysis, uses the MaskR-CNN algorithm to process camera image data, identifies and distinguishes static and dynamic obstacles, and then integrates sonar, radar and image data information to generate an obstacle characteristic classification map;

[0037] The risk assessment module measures the obstacle size and distance based on the obstacle characteristic classification diagram, compares the obstacle size and distance with the preset safety parameters, uses a decision tree algorithm to assess the risk level of each obstacle, performs risk labeling, and outputs an obstacle risk label;

[0038] The heading adjustment module analyzes the obstacle avoidance strategy according to the obstacle risk label, including the adjustment of heading and speed, uses the proximal strategy optimization algorithm to execute the control instruction, adjusts the navigation path, and generates a route adjustment plan;

[0039] The navigation assessment module updates the navigation control strategy based on the route adjustment plan, monitors the command execution status, compares the path and speed adjustment before and after navigation, evaluates the matching degree between the navigation status and the obstacle avoidance effect, and generates a navigation status and navigation safety assessment log.

[0040] An artificial intelligence-based unmanned ship or boat comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the artificial intelligence-based unmanned ship or boat safety identification and inspection system as described above is implemented.

[0041] Compared with the prior art, the advantages and positive effects of the present invention are:

[0042] In the present invention, through the joint application of spectrum analysis and MaskR-CNN, obstacle recognition is not limited to basic forms, but can also be classified in real time according to dynamic changes, which greatly enhances the accuracy of recognition and the timeliness of operation. In addition, the introduction of the decision tree algorithm allows the risk assessment strategy to be dynamically adjusted according to real-time data and environmental changes, providing a reliable safety decision support for unmanned boats, optimizing the data processing and analysis process, enabling unmanned boats to effectively perform obstacle avoidance tasks under a wider range of conditions, and improving the automation level and safety of navigation. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a schematic diagram of the workflow of the present invention;

[0044] Figure 2 This is a detailed flow chart of S1 of the present invention;

[0045] Figure 3 This is a detailed flow chart of S2 of the present invention;

[0046] Figure 4This is a detailed flow chart of S3 of the present invention;

[0047] Figure 5 This is a detailed flow chart of S4 of the present invention;

[0048] Figure 6 This is a detailed flow chart of S5 of the present invention;

[0049] Figure 7 This is a detailed flow chart of S6 of the present invention;

[0050] Figure 8 It is a system flow chart of the present invention. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0052] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0053] Embodiment 1

[0054] See also Figure 1 The present invention provides a technical solution: an artificial intelligence-based unmanned ship and boat safety identification and inspection algorithm, comprising the following steps:

[0055] S1: Based on the collected sonar, radar and camera data, data integration is performed, and data synchronization of multiple data sources is performed through timestamp comparison to generate an integrated environmental data body;

[0056] S2: Based on the environmental data integration, the sonar and radar data are first analyzed through spectrum analysis, and then the camera data is processed in combination with the Mask R-CNN algorithm. Static and dynamic obstacles are distinguished according to the features to obtain an obstacle characteristic classification map;

[0057] S3: Based on the obstacle characteristic classification diagram, the size and distance of the obstacle are measured. According to the comparison between the measurement results and the preset safety parameters, the decision tree algorithm is used to determine the risk level of the obstacle, and the risk of each obstacle classification is marked to generate an obstacle risk label;

[0058] S4: Based on the obstacle risk labels, analyze multiple potential obstacle avoidance strategies, including heading adjustment and speed control, and execute control instructions through the proximal strategy optimization algorithm to adjust the navigation path of the unmanned ship and boat to obtain the route adjustment plan;

[0059] S5: Based on the route adjustment plan, update the navigation control strategy, adjust the heading and speed, implement the new navigation command, monitor the command execution status, verify whether the navigation of the unmanned ship and boat is updated, and generate the navigation status update record;

[0060] S6: Based on the navigation status update record, collect navigation data, compare the path changes and speed adjustments before and after navigation, evaluate the matching degree between the navigation status and the expected obstacle avoidance effect, and generate a navigation safety assessment log.

[0061] The environmental data integration includes sonar sound wave signal data, radar electromagnetic wave echo data, and camera visual image data. The obstacle characteristic classification map includes the type identification, relative position, and moving speed information of the obstacle. The obstacle risk label includes the risk level, distance parameter, and size estimate of each obstacle. The route adjustment plan includes the predetermined heading change, speed adjustment amount, and obstacle avoidance strategy options. The navigation status update record includes the updated GPS coordinates, speed change record, and heading angle adjustment data. The navigation safety assessment log includes path change records, speed change analysis, and risk matching level.

[0062] See also Figure 2 , based on the collected sonar, radar and camera data, data integration is performed, and data synchronization of multiple data sources is performed through timestamp comparison. The specific steps to generate the environmental data integration are as follows:

[0063] S101: Based on the collected data, perform an initial data quality check, identify and delete data that does not meet the standards, including signals exceeding preset thresholds and obvious data corruption, and generate preliminary screening data;

[0064] S101: First, set the data quality threshold to the range of upper and lower limits, such as the signal strength threshold not exceeding 100 units. Then, traverse each data point in the data set and perform a comparison operation on each point. If the data point exceeds the above range, the point is removed from the data set. At the same time, check whether the data point has any form of damage. For example, if the data format is incorrect or the value is missing, the data will also be removed. Through this operation, ensure that all remaining data meets the preset quality standards and generate preliminary screening data.

[0065] S102: Based on the preliminary screening data, time correction is performed on each data point to correct the time difference in the data, and the time stamp of the data source is synchronized to verify whether the timeline is consistent, and generate time synchronization data;

[0066] S102: Calculate the difference between the time stamp of each point in the data and the reference time stamp, set the time adjustment function T(h)=m1+h*δ, where m1 is the mean of all time differences, h is the index of the specific time point, and δ is the correction coefficient. Apply this function to adjust the time stamp of each data point one by one to minimize the overall time error. After the adjustment is completed, check the continuity and consistency of the timeline again to ensure that there is no time deviation and generate time synchronization data.

[0067] S103: Based on the time synchronization data, merge the data streams of sonar, radar and camera, unify the difference data format and create a single data interface, then verify the consistency and operability of the data set to generate an integrated environmental data body;

[0068] S103: First, define a unified data format specification, including data type, precision and size, and then convert the format of each data source to the unified specification one by one. For the data interface, set the integration rule Ω={ω1, ω2, ω3}, where ω1 represents the sonar data weight, ω2 represents the radar data weight, and ω3 represents the camera data weight. Use the weight set and data type to merge the data to ensure that the merged data stream is not distorted. After completing the data format unification and data merging, verify the consistency and operability of the data set to ensure that the data integration can provide support for subsequent analysis and application, and generate an environmental data integration.

[0069] See also Figure 3 Based on the environmental data integration, the sonar and radar data are first analyzed through spectrum analysis, and then the camera data is processed in combination with the MaskR-CNN algorithm. Static and dynamic obstacles are distinguished according to the characteristics. The specific steps to obtain the obstacle characteristic classification map are as follows:

[0070] S201: Based on the environmental data integration, perform spectrum analysis on sonar and radar data, separate key frequency components, extract key signal features, and generate spectrum feature sets;

[0071] S201: First, apply Fourier transform to sonar and radar data to convert time domain signals into frequency domain signals, including setting the Fourier transform formula Among them, f(t) is the signal function in the time domain, ω is the angular frequency. Through transformation, the main frequency components in the data are extracted. Then, according to the set frequency screening criteria, such as retaining the frequency components with the top 10% of energy, the key signals are determined, and the frequency components are further analyzed to extract their amplitude and phase characteristics to generate a spectrum feature set.

[0072] S202: Based on the spectral feature set, the Mask R-CNN algorithm is used to identify the image data captured by the camera, identify the outline of the object in the image, distinguish static and dynamic objects through shape analysis, and generate edge recognition data;

[0073] S202: Based on the spectral feature set, the image data captured by the camera is recognized, and the shape matching technology is used to identify the contours of objects in the image. Static and dynamic objects are distinguished through shape analysis. The image data is converted into a grayscale image, and then the Canny edge detection algorithm is applied to extract the edge information of the object, including setting a double threshold to detect strong edges and weak edges, and linking edges to form a complete object contour. Static and dynamic objects are distinguished by comparing the position changes of the object contours in consecutive frames. Finally, edge recognition data is generated based on shape characteristics such as area, perimeter and geometric center.

[0074] S203: Based on the edge recognition data, combined with the signal characteristics of sonar and radar, the identified objects are classified and marked as static or dynamic obstacles, and risk assessment is performed on each object to obtain an obstacle characteristic classification map;

[0075] S203: Using an integrated analysis method, the features of sonar and radar data, such as signal strength and frequency response, are combined with the shape features of image data to create classification rules to determine whether an object is a static or dynamic obstacle. During the classification process, the comprehensive score of each object feature is calculated using the formula S=α1P1+α2P2+α3P3, where α1, α2, and α3 are weight coefficients corresponding to the importance of sonar signal strength, radar frequency response, and image shape features, respectively. P1, P2, and P3 are quantitative values ​​of each feature. Classification is performed based on the score, and a risk assessment is performed on each object, including its potential threat level to navigation safety, to obtain an obstacle characteristic classification diagram.

[0076] See also Figure 4 Based on the obstacle characteristic classification diagram, the size and distance of the obstacle are measured. According to the comparison between the measurement results and the preset safety parameters, the decision tree algorithm is used to determine the risk level of the obstacle, and each obstacle classification is labeled with risk. The specific steps for generating obstacle risk labels are as follows:

[0077] S301: Based on the obstacle characteristic classification diagram, measure the straight-line distance and size of the obstacles marked in the diagram, record the size and distance parameters of each obstacle, and generate an obstacle measurement record;

[0078] S301: First, use image processing technology to determine the boundaries of each obstacle, apply the measurement tools in digital image processing to calculate the area and perimeter of each obstacle, and convert the image size into actual distance through image resolution and actual scale. For example, the actual distance represented by each pixel is 0.5 meters. In addition, use triangulation to calculate the straight-line distance between the obstacle and the observation point. The specific operation is to determine the position of the obstacle in the image and the position of the reference point, calculate the pixel distance between the two points, and then convert it into actual distance. The data is recorded to generate an obstacle measurement record.

[0079] S302: Based on the obstacle measurement record, a decision tree algorithm is used to compare with the safe operation parameters of the vessel, calculate the obstacle risk level that may cause a collision, evaluate the potential threat to navigation safety, and generate an obstacle risk assessment record;

[0080] S302: First, define the risk assessment criteria, including the size and distance of obstacles and their relative position to the navigation path. Use decision analysis to calculate the potential threat to navigation safety based on the size and distance parameters of the obstacles. Set the risk calculation formula to R=β1D+β2S, where D represents the distance from the obstacle to the vessel, S represents the size of the obstacle, β1 and β2 are weight factors, and adjust the factors to reflect the impact of distance and size on risk assessment. Apply this formula to assign a risk level to each obstacle and generate an obstacle risk assessment record.

[0081] S303: Based on the obstacle risk assessment record, mark the risk level of each obstacle, including low, medium and high, and mark the risk information on the navigation map to generate an obstacle risk label;

[0082] S303: Mark the risk levels with different colors or symbols on the navigation map to visually identify potential danger areas. Use GIS technology and image marking tools to mark the corresponding risk level at the location of each obstacle. For example, use red marks to indicate high-risk obstacles, yellow marks to indicate medium risks, and green marks to indicate low risks. In addition, include additional annotation information, such as the specific type and size of the obstacle, to provide more contextual information. Through this process, an obstacle risk label is generated.

[0083] See also Figure 5 , based on the obstacle risk label, analyze a variety of potential obstacle avoidance strategies, including heading adjustment and speed control, and execute control instructions through the proximal strategy optimization algorithm to adjust the navigation path of the unmanned ship and boat. The specific steps for obtaining the route adjustment plan are as follows:

[0084] S401: Based on the obstacle risk label, identify the path area that needs to avoid obstacles, formulate a preliminary obstacle avoidance strategy for each risk level of obstacles, including heading deviation and speed adjustment strategy, and generate an obstacle avoidance planning scheme;

[0085] S401: First, the navigation map is analyzed to determine the areas marked as medium and high risks, and paths to avoid the areas are planned. For each high-risk area, the heading deviation angle is set, specifically the minimum angle change between the obstacle position and the current heading of the ship, to ensure sufficient obstacle avoidance distance. For speed adjustment, the speed reduction ratio is set according to the distance and size of the obstacle. For example, large obstacles that are close are required to be reduced to 50% of the normal speed. The strategy is integrated into the obstacle avoidance planning scheme to ensure that the ship can safely bypass the obstacle.

[0086] S402: Based on the obstacle avoidance planning scheme, determine the control instructions, including the heading change angle and speed adjustment parameters, prepare the navigation control update instructions, and generate the navigation control parameters;

[0087] S402: Convert each obstacle avoidance strategy into specific heading and speed control instructions. For example, if the heading needs to be shifted 30 degrees to the right, the control instruction will be encoded as increasing the heading angle by 30 degrees. The speed adjustment is based on the ratio set in the obstacle avoidance strategy. The engine output is adjusted through the control strategy to achieve the required speed. The control instructions are integrated and updated instructions for navigation control are prepared, including specific adjustment values ​​for heading and speed, to generate navigation control parameters.

[0088] S403: Based on the navigation control parameters, the proximal strategy optimization algorithm is used to adjust the navigation paths of the unmanned ships and boats, and the adjusted headings and speeds are applied to ensure that the ships navigate along the optimized paths, thereby generating a route adjustment plan;

[0089] S403: Based on the navigation control parameters, the proximal strategy optimization algorithm is used to adjust the navigation path of the unmanned ships and boats. By real-time analysis of the current navigation environment and the expected navigation path, the heading and speed adjustment are optimized to adapt to sudden environmental changes and unmarked obstacles. The heading and speed parameters optimized by the algorithm are applied to ship control to ensure that the ship navigates along the optimized path. In this way, the ship can effectively avoid identified obstacles while maintaining navigation efficiency and safety, and generate a route adjustment plan.

[0090] The proximal strategy optimization algorithm is based on the formula:

[0091]

[0092] Calculate probability ratios;

[0093] Where s is the current state of the ship, v is the current speed, d is the distance between the target and the current position, θ is the angle between the ship and the target, ω is the weight coefficient dynamically adjusted according to the changes in the navigation environment, and P new (a|s, v, d, θ) is the probability of selecting action a given state s under the new strategy, P old (a|s, v, d, θ) is the probability under the old strategy.

[0094] Execution process:

[0095] Firstly, the probability ratio of selecting action a under the given state s, current speed v, distance to target d and relative angle θ under the new and old strategies is calculated. Then, the strategy update step is adjusted according to the real-time data of the navigation environment through the weight coefficient ω. The determination of the weight ω involves environmental factors such as wind speed, flow speed and close-range dynamics of other ships. The weight is optimized using a machine learning model by analyzing historical data and real-time sensor feedback to ensure that each adjustment is made under the premise of minimizing risks. In this way, the heading and speed adjustments are optimized while maintaining navigation efficiency and safety, the optimal heading and speed parameters are obtained, and the final action probability ratio is calculated to generate a more accurate heading control solution.

[0096] See also Figure 6 , based on the route adjustment plan, update the navigation control strategy, adjust the heading and speed, implement the new navigation command, and monitor the command execution status to verify whether the navigation of the unmanned ship and boat is updated. The specific steps for generating the navigation status update record are:

[0097] S501: Based on the route adjustment plan, adjust the navigation control strategy parameters, including inputting new heading angle and speed settings, adjusting the automatic navigation settings, matching the predetermined obstacle avoidance route, and generating a heading speed configuration record;

[0098] S501: First, the optimal heading and speed parameters determined in the route adjustment plan are input into the ship's automatic navigation system, the heading angle is adjusted to ensure that the ship can bypass obstacles and navigate according to the predetermined obstacle avoidance route, and the speed setting is fine-tuned to adapt to the safety requirements of different navigation areas, such as reducing the speed in high-risk areas to increase reaction time and safety margin. The adjustment is made through the ship's navigation control interface to ensure that the input angle and speed accurately match the obstacle avoidance path requirements. After completing the steps, a heading speed configuration record is generated.

[0099] S502: Based on the heading and speed configuration records, activate the navigation command, track the command execution effect in real time, record each adjustment result of the heading and speed, and generate an execution status monitoring log;

[0100] S502: Based on the heading and speed configuration records, the navigation command is activated and the command execution effect is tracked in real time. After the navigation command is activated, the actual performance of the heading and speed is continuously monitored through the ship's navigation system, and compared with the preset parameters. Each heading adjustment and speed change is automatically recorded to ensure that all instructions are executed according to the predetermined configuration. During the execution process, any deviation from the predetermined heading or speed will be detected and adjusted immediately to ensure the accurate execution of the navigation command and correct any deviation in a timely manner, thereby generating an execution status monitoring log.

[0101] S503: Based on the execution status monitoring log, evaluate whether the navigation status is updated according to the new parameters, compare and analyze the navigation path with the plan, check the execution status of the new route, and generate a navigation status update record;

[0102] S503: Utilize the detailed navigation data recorded in the log to analyze the consistency between the actual execution of the navigation and the plan, check whether the adjustment of the heading and speed has achieved the predetermined obstacle avoidance effect, and ensure navigation safety and efficiency. In addition, for any non-compliant parts, analyze the reasons, such as environmental factors, equipment performance or operating errors, so as to carry out subsequent strategy optimization. After the analysis is completed, generate a navigation status update record, which describes the compliance assessment results of the navigation status and necessary adjustment suggestions.

[0103] See also Figure 7 , based on the navigation status update record, collect navigation data, compare the path changes and speed adjustments before and after navigation, and evaluate the matching degree between the navigation status and the expected obstacle avoidance effect. The specific steps for generating the navigation safety assessment log are as follows:

[0104] S601: Based on the navigation status update record, collect navigation related data, including current heading, speed and position change, and record key parameters to generate a navigation data set;

[0105] S601: First, extract the real-time data of heading, speed and position from the ship's navigation system. The data is automatically recorded by GPS and ship navigation instruments to ensure the accuracy and timeliness of the data. Analyze each navigation cycle and record any important changes, such as heading adjustment or speed change. Key parameters will be sorted and stored to form a navigation data set, which includes detailed records of each voyage, such as each modification of the heading, each adjustment of the speed and its corresponding time and position information.

[0106] S602: Based on the navigation data set, perform data analysis, compare the deviation between the planned route and the current route, analyze the speed adjustment effect, determine whether it meets the obstacle avoidance target, and generate navigation deviation analysis information;

[0107] S602: By comparing the actual navigation path in the navigation data set with the planned route, the path deviation is calculated, including the angle difference and distance difference between the actual heading and the planned heading. In addition, the effect of the speed adjustment is analyzed. By evaluating the impact of the speed change on the obstacle avoidance effect, it is ensured that the speed adjustment meets the obstacle avoidance needs. The linear regression analysis is used to determine the statistical significance of the navigation deviation and speed adjustment, thereby generating navigation deviation analysis information, which records the deviation during navigation and the effect of the speed adjustment.

[0108] S603: Based on the navigation deviation analysis information, evaluate the matching degree between the navigation status and the predetermined obstacle avoidance strategy, verify whether the navigation adjustment meets the expected safety standard, and generate a navigation safety assessment log;

[0109] S603: Use the data in the navigation deviation analysis information to conduct a detailed evaluation to compare the consistency between the actual navigation behavior and the predetermined obstacle avoidance strategy, including checking whether the navigation path effectively avoids obstacles and whether the speed adjustment ensures the safety of the ship at critical moments. Based on the evaluation results, a navigation safety assessment log is generated, which records the safety performance of each navigation, including the obstacle avoidance success rate, navigation safety margin and any safety hazards, providing an important basis for subsequent navigation safety strategies.

[0110] See also Figure 8 The AI-based unmanned ship and boat safety identification and inspection system includes:

[0111] The data integration module collects data from sonar, radar and cameras, excludes abnormal data, performs time correction on data points, merges data streams from different sources, unifies data formats and creates a single data interface to generate an integrated environmental data body;

[0112] The obstacle recognition module uses the environmental data integration to analyze sonar and radar data through spectrum analysis, uses the MaskR-CNN algorithm to process camera image data, identifies and distinguishes static and dynamic obstacles, and then integrates sonar, radar and image data information to generate an obstacle characteristic classification map;

[0113] The risk assessment module measures the obstacle size and distance based on the obstacle characteristic classification diagram, compares it with the preset safety parameters, uses the decision tree algorithm to evaluate the risk level of each obstacle, performs risk annotation, and outputs the obstacle risk label;

[0114] The heading adjustment module analyzes the obstacle avoidance strategy based on the obstacle risk label, including the adjustment of heading and speed, uses the proximal strategy optimization algorithm to execute control instructions, adjusts the navigation path, and generates a route adjustment plan;

[0115] The navigation assessment module updates the navigation control strategy based on the route adjustment plan, monitors the command execution status, compares the path and speed adjustment before and after navigation, evaluates the matching degree between the navigation status and the obstacle avoidance effect, and generates a navigation status and navigation safety assessment log.

[0116] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. Based on artificial intelligence unmanned ship and boat safety identification and inspection algorithm, it is characterized by: The following steps are involved: Based on the collected sonar, radar and camera data, data integration is performed, and data synchronization of multiple data sources is performed through timestamp comparison to generate an integrated environmental data body; Based on the environmental data integration, the sonar and radar data are first analyzed by spectrum analysis, and then the camera data is processed in combination with the MaskR-CNN algorithm, and static and dynamic obstacles are distinguished according to the features to obtain an obstacle characteristic classification map; Based on the obstacle characteristic classification diagram, the size and distance of the obstacle are measured, and the risk level of the obstacle is judged by using a decision tree algorithm according to the comparison between the measurement results and the preset safety parameters, and each obstacle classification is risk-labeled to generate an obstacle risk label; Based on the obstacle risk labels, multiple potential obstacle avoidance strategies are analyzed, including heading adjustment and speed control, and control instructions are executed through a proximal strategy optimization algorithm to adjust the navigation paths of the unmanned ships and boats to obtain a route adjustment plan; Based on the route adjustment plan, update the navigation control strategy, adjust the heading and speed, implement the new navigation command, monitor the command execution status, verify whether the navigation of the unmanned ship and boat is updated, and generate a navigation status update record; Based on the navigation status update record, navigation data is collected, path changes and speed adjustments before and after navigation are compared, and the matching degree between the navigation status and the expected obstacle avoidance effect is evaluated to generate a navigation safety assessment log; Based on the environmental data integration, the sonar and radar data are first analyzed by spectrum analysis, and then the camera data is processed in combination with the MaskR-CNN algorithm, and static and dynamic obstacles are distinguished according to the characteristics. The steps to obtain the obstacle characteristic classification map are as follows: Based on the environmental data integration, spectrum analysis is performed on the sonar and radar data to separate key frequency components, and key signal feature extraction is performed to generate a spectrum feature set; Apply Fourier transforms to sonar and radar data to convert time domain signals to frequency domain signals, including setting up the Fourier transform formula Among them, f(t) is the signal function in the time domain, ω is the angular frequency, and the main frequency components in the data are extracted through transformation. Then, according to the set frequency screening criteria, such as retaining the frequency components with the top 10% of energy, the key signals are determined, and the frequency components are further analyzed to extract their amplitude and phase characteristics to generate a spectrum feature set; Based on the spectral feature set, the Mask R-CNN algorithm is used to identify the image data captured by the camera, identify the contours of objects in the image, distinguish static and dynamic objects through shape analysis, and generate edge recognition data; Based on the edge recognition data, combined with the signal characteristics of sonar and radar, the identified objects are classified and marked as static or dynamic obstacles, and a risk assessment is performed on each object to obtain an obstacle characteristic classification map; Using an integrated analysis method, the characteristics of sonar and radar data, such as signal strength and frequency response, are combined with the shape characteristics of image data to create classification rules to determine whether an object is a static or dynamic obstacle. During the classification process, the comprehensive score of each object feature is calculated. The formula S=α1P1+α2P2+α3P3, where α1, α2 and α3 are weight coefficients corresponding to the importance of sonar signal strength, radar frequency response and image shape features, respectively. P1, P2 and P3 are the quantitative values ​​of each feature. Objects are classified according to their scores, and a risk assessment is performed on each object, including its potential threat level to navigation safety, to obtain an obstacle characteristic classification map.

2. The artificial intelligence-based unmanned ship and boat safety identification and inspection algorithm according to claim 1 is characterized in that: The environmental data integration body includes the sonar sound wave signal data, the radar electromagnetic wave echo data, and the camera visual image data; the obstacle characteristic classification map includes the type identification, relative position, and moving speed information of the obstacle; the obstacle risk label includes the risk level, distance parameter, and size estimate of each obstacle; the route adjustment plan includes the predetermined heading change, speed adjustment amount, and obstacle avoidance strategy options; the navigation status update record includes the updated GPS coordinates, speed change record, and heading angle adjustment data; the navigation safety assessment log includes path change records, speed change analysis, and risk matching level.

3. The artificial intelligence-based unmanned ship and boat safety identification and inspection algorithm according to claim 1 is characterized in that: Based on the obstacle characteristic classification diagram, the size and distance of the obstacle are measured. According to the comparison between the measurement results and the preset safety parameters, the risk level of the obstacle is judged using the decision tree algorithm, and each obstacle classification is risk labeled. The steps of generating the obstacle risk label are as follows: Based on the obstacle characteristic classification diagram, measure the straight-line distance and size of the obstacles marked in the diagram, record the size and distance parameters of each obstacle, and generate an obstacle measurement record; Based on the obstacle measurement record, using a decision tree algorithm, comparing with the safe operating parameters of the vessel, calculating the obstacle risk level that may cause a collision, assessing the potential threat to navigation safety, and generating an obstacle risk assessment record; Based on the obstacle risk assessment record, the risk level of each obstacle is marked, including low, medium and high, and the risk information is marked on the navigation map to generate an obstacle risk label.

4. The artificial intelligence-based unmanned ship and boat safety identification and inspection algorithm according to claim 1 is characterized in that: Based on the obstacle risk label, a variety of potential obstacle avoidance strategies are analyzed, including heading adjustment and speed control, and control instructions are executed through the proximal strategy optimization algorithm to adjust the navigation path of the unmanned ship and boat. The specific steps of obtaining the route adjustment plan are as follows: Based on the obstacle risk labels, the path areas that need to be avoided are identified, a preliminary obstacle avoidance strategy is formulated for each risk level of obstacle, including a heading deviation and a speed adjustment strategy, and an obstacle avoidance planning scheme is generated; Based on the obstacle avoidance planning scheme, control instructions are determined, including heading change angles and speed adjustment parameters, and update instructions for navigation control are prepared to generate navigation control parameters; Based on the navigation control parameters, a proximal strategy optimization algorithm is used to adjust the navigation path of the unmanned ship and boat, and the adjusted heading and speed are applied to ensure that the ship navigates along the optimized path and generate a route adjustment plan.

5. The artificial intelligence-based unmanned ship and boat safety identification and inspection algorithm according to claim 1 is characterized in that: The proximal strategy optimization algorithm is based on the formula: Calculate probability ratios; Where s is the current state of the ship, v is the current speed, d is the distance between the target and the current position, θ is the angle between the ship and the target, ω is the weight coefficient dynamically adjusted according to the changes in the navigation environment, and P new (a|s, v, d, θ) is the probability of selecting action a given state s under the new strategy, P old (a|s, v, d, θ) is the probability under the old strategy.

6. The artificial intelligence-based unmanned ship and boat safety identification and inspection algorithm according to claim 1 is characterized in that: Based on the route adjustment plan, the navigation control strategy is updated, the heading and speed are adjusted, the new navigation command is implemented, and the command execution status is monitored to verify whether the navigation of the unmanned ship and boat is updated. The steps of generating the navigation status update record are as follows: Based on the route adjustment plan, adjusting the navigation control strategy parameters, including inputting new heading angle and speed settings, adjusting the automatic navigation settings, matching the predetermined obstacle avoidance route, and generating a heading speed configuration record; Based on the heading and speed configuration records, the navigation command is activated, the command execution effect is tracked in real time, and each adjustment result of the heading and speed is recorded to generate an execution status monitoring log; Based on the execution status monitoring log, it is evaluated whether the navigation status is updated according to the new parameters, the navigation path is compared and analyzed with the plan, and the execution of the new route is checked to generate a navigation status update record.

7. Based on artificial intelligence, the safety identification and inspection system for unmanned ships and boats is characterized by: The artificial intelligence-based unmanned ship and boat safety identification and inspection system is used to execute the artificial intelligence-based unmanned ship and boat safety identification and inspection algorithm according to any one of claims 1 to 6, and the system includes: The data integration module collects data from sonar, radar and cameras, excludes abnormal data, performs time correction on data points, merges data streams from different sources, unifies data formats and creates a single data interface to generate an integrated environmental data body; The obstacle recognition module uses the environmental data integration to analyze sonar and radar data through spectrum analysis, uses the MaskR-CNN algorithm to process camera image data, identifies and distinguishes static and dynamic obstacles, and then integrates sonar, radar and image data information to generate an obstacle characteristic classification map; The risk assessment module measures the obstacle size and distance based on the obstacle characteristic classification diagram, compares the obstacle size and distance with the preset safety parameters, uses a decision tree algorithm to assess the risk level of each obstacle, performs risk labeling, and outputs an obstacle risk label; The heading adjustment module analyzes the obstacle avoidance strategy according to the obstacle risk label, including the adjustment of heading and speed, uses the proximal strategy optimization algorithm to execute the control instruction, adjusts the navigation path, and generates a route adjustment plan; The navigation assessment module updates the navigation control strategy based on the route adjustment plan, monitors the command execution status, compares the path and speed adjustment before and after navigation, evaluates the matching degree between the navigation status and the obstacle avoidance effect, and generates a navigation status and navigation safety assessment log.

8. An artificial intelligence-based unmanned ship or boat, comprising a memory and a processor, characterized in that: The memory stores a computer program, and when the processor executes the computer program, the artificial intelligence-based unmanned ship and boat safety identification and inspection system described in claim 7 is implemented.

Images