Multi-source sensing fusion ship full-scene intelligent safety monitoring and decision early warning system

Abstract

The invention discloses a multi-source sensing fusion ship full-scene intelligent safety monitoring and decision early warning system, and relates to the field of ship safety, and the system comprises an acquisition module which is used for collecting meteorological, hydrological, channel obstacle, ship equipment operation parameters and scene image original data in real time; the adaptation module is used for receiving the data acquired by the acquisition module, dynamically calibrating the deviation of the sensor based on the navigation scene, adapting the data dimension required by risk identification, and outputting the calibrated data; according to the method, deviation is corrected through dynamic calibration, data precision is optimized in combination with scene complexity, transmission bandwidth is dynamically allocated according to priority to guarantee efficient transmission of key information, invalid data is eliminated through cross-dimension verification to improve credibility of fused data, and navigation potential safety hazards and equipment abnormal working conditions are accurately recognized.

Description

Technical Field

[0001] This invention relates to the field of ship safety technology, specifically to a multi-source sensor fusion intelligent safety monitoring and decision-making early warning system for ships in all scenarios. Background Technology

[0002] Ship safety monitoring is a core safety assurance measure for ship navigation and operation. It integrates technologies such as sensing, satellite positioning, and wireless communication to collect real-time data on the operating status of ship power systems, navigation equipment, and hull structures. At the same time, it monitors environmental data such as surrounding hydrology, meteorology, and waterway navigation. Through data analysis, it can predict risks and issue early warnings of anomalies, promptly manage potential navigation safety hazards, avoid accidents such as collisions and malfunctions, ensure the safety of ships, crew members, and cargo, and improve the safety management level of water traffic operations.

[0003] Patent application number 202510231069.3 discloses a ship monitoring and early warning system and method. This application aims to solve the problems of "collecting radar signals, image signals, and Beidou positioning signals from passing ships, which requires the deployment of complex equipment, such as radar and video monitoring equipment. In addition, due to the influence of bridges on signals, it is impossible to perform real-time perception, interaction, and early warning of navigation attitude when ships are passing under bridges. Existing ship heading guidance display systems have limitations in guiding passing ships, and can only guide within a short distance of the bridge. In adverse weather conditions such as low visibility, visibility may limit the operator's operation, and the guidance area is too close to the bridge, so there is no buffer time for early warning and response in case of an accident."

[0004] However, ships face a variety of environmental and safety risks during navigation, such as severe weather, waterway obstacles, and equipment failures. Traditional monitoring systems usually rely on a single sensor or simple data processing, making it difficult to comprehensively and accurately assess the environmental and safety status.

[0005] To address this, we propose a multi-source sensor fusion-based intelligent safety monitoring and decision-making early warning system for ships across all scenarios. Summary of the Invention

[0006] In view of the above-mentioned shortcomings of the existing technology, the present invention provides a multi-source sensor fusion intelligent safety monitoring and decision-making early warning system for ships in all scenarios, which can effectively solve the problems of the existing technology.

[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions; This invention discloses a multi-source sensor fusion intelligent safety monitoring and decision-making early warning system for ships in all scenarios, comprising: The system comprises the following modules: a data acquisition module for real-time acquisition of raw data on meteorology, hydrology, waterway obstacles, ship equipment operating parameters, and scene images; an adaptation module for receiving data from the acquisition module, dynamically calibrating sensor deviations based on the navigation scenario, adapting to the data dimensions required for risk identification, and outputting calibrated data; a fusion module for generating fused data by eliminating invalid information through cross-dimensional data verification based on the calibrated data; a construction module for identifying navigation safety hazards and abnormal equipment conditions in real-time based on the fused data, and constructing a decision model adapted to the ship's navigation scenario; an early warning module for triggering graded audible and visual early warnings based on the confidence level of the joint decision model obtained from the identification results of navigation safety hazards and abnormal equipment conditions; and a management module for classifying, storing, and backing up the raw data, calibration data, fused data, and decision models in different locations. The acquisition module is interconnected with an adaptation module and a fusion module via a local area network. The adaptation module is interconnected with the fusion module via a local area network. The adaptation module and the fusion module are interconnected with a construction module, an early warning module, and a management module via a wireless network. The construction module is interconnected with the early warning module via a local area network. The early warning module is interconnected with the management module via a local area network. Meteorological data includes at least wind speed, wind direction, temperature, humidity, and air pressure, while hydrological data includes at least water temperature, salinity, depth, flow velocity, and flow direction.

[0008] Furthermore, the acquisition module consists of a meteorological sensing unit, a hydrological sensing unit, an obstacle detection unit, an equipment sensing unit, and an image acquisition unit. The obstacle detection unit uses a combination of lidar and millimeter-wave radar for detection, and the image acquisition unit uses a combination of an infrared camera and a visible light camera for image capture. During the operation of the acquisition module, waterway obstacle data and key equipment operating parameters are set as first-level priority, meteorological and hydrological extreme value data and abnormal frame data of scene images are set as second-level priority, and routine meteorological and hydrological data and routine equipment operating parameters are set as third-level priority. The transmission bandwidth is dynamically allocated according to a preset ratio.

[0009] Furthermore, the dynamic calibration in the adaptation module is performed based on sensor bias and the complexity of the navigation scenario, and the calibration formula is: ; In the formula: Data after calibration; This refers to the raw data collected by the sensor. This is the deviation correction factor; This represents the scene complexity coefficient. in, , This represents the historical average deviation value of the sensor. This represents the average value of the sensor's standard reference value. This represents the sensor accuracy weighting coefficient.

[0010] Furthermore, the scene complexity coefficient The value of is quantified by the following formula: ; In the formula: As a scene-related influencing factor; For the ship's real-time navigation speed; Real-time effective wave height; This refers to the deviation between the equipment's real-time operating parameters and standard parameters. This refers to the ship's normal sailing speed; The normal significant wave height; These are the standard operating parameters for the equipment; in, All are positive numbers, and their sum is 1.

[0011] Furthermore, the fusion module performs an invalid information removal stage, which performs preliminary removal based on data credibility, and determines the k-th dimension calibration data with credibility lower than the preset credibility threshold as invalid information and directly excludes it from the fusion calculation; The credibility of the data ; Next, perform cross-dimensional consistency checks on the remaining valid dimension data, and calculate the relative deviation between any two valid dimension data. ,like If the deviation exceeds the preset consistency deviation threshold, the one with the larger deviation will be judged as invalid information and removed. In the formula: The standard deviation of the data for the k-th dimension; The average value of the calibration data in the k-th dimension; The fluctuation tolerance coefficient for the k-th dimension data; This is the calibrated data for the k1th dimension; This is the calibrated data for the k2th dimension, where k1 and k2 represent two different dimensions.

[0012] Furthermore, during the data fusion stage of the fusion module, various types of data are normalized to ensure that the values ​​of all types of data are constrained within the range of [0,1] before fusion. ; In the formula: To integrate data; The k-th dimension is the fused input data, which corresponds to the normalization result. During the data generation phase of the fusion module, the calculated fusion data is standardized according to a preset format, which includes data dimension identifiers, fusion timestamps, and overall credibility levels. ; in, This represents the total number of valid data dimensions participating in the fusion process after removing invalid information. This indicates the credibility of the data in the k-th dimension. This represents the dynamic fusion weight of the k-th dimension.

[0013] Furthermore, during the runtime phase of the building module, based on fused data... Extracting multidimensional feature vectors, including: Navigation safety hazard feature vector ; Equipment abnormal operating condition feature vector ; Real-time identification of navigation safety hazards and abnormal equipment conditions is achieved through comprehensive anomaly calculation. Comprehensive anomaly level: ; In the formula; For overall anomaly degree; Weighting of safety hazard characteristics; This is a standard feature vector for safety hazards; Weights for equipment anomaly features; This represents the standard feature vector of the equipment under normal operating conditions. The Euclidean norm of a vector; , All are positive numbers and their sum is 1. When the value exceeds the preset initial anomaly detection threshold, it is determined that there is a potential navigation safety hazard or an abnormal equipment condition. The decision model ; In the formula: This is the decision-making level coefficient; Scene dynamic adaptation threshold ; Scene adaptation factor ; In the formula: To preset the initial anomaly detection threshold; Adjust the coefficients to suit the scene; , , The influence weights are determined by ship tonnage, type of navigation waterway, and priority of navigation mission. This refers to the current actual tonnage of the vessel; The complexity coefficient of the navigation waterway; This refers to the priority coefficient for navigation missions. For standard reference tonnage; The standard water area complexity coefficient; This refers to the standard task priority coefficient. This represents the influence coefficient of scene complexity. This represents the scene complexity coefficient.

[0014] Furthermore, the confidence level of the decision model in the early warning module is: ; In the formula: Confidence level of the decision model; , , To integrate data credibility weight, feature matching credibility weight, and scenario adaptation credibility weight, , , The values ​​of all are in the range of (0,1), and their sum is 1. To integrate the overall credibility level of the data; To assess the credibility of feature matching; For scene adaptation factors; in, .

[0015] Furthermore, the triggering logic for the tiered audible and visual early warning is as follows: The preset warning level thresholds include: high confidence threshold C1, medium confidence threshold C2, and low confidence threshold C3; the preset decision level coefficient thresholds include: high risk threshold K1, medium risk threshold K2, and low risk threshold K3. Level 1 warning trigger conditions: when ≥C1 and When K1 is ≥1, a Level 1 audible and visual warning is triggered: the audible alarm uses a high-frequency pulse tone, which is continuously output without interruption; the visual alarm uses a red strobe light, which is synchronized with the ship's control system to pop up an emergency braking prompt interface and record the fused data and feature vector at the time of the warning trigger. Level 2 warning trigger conditions: when ≥C2 and K2≤ When the threshold is <K1, a Level 2 audible and visual warning is triggered: the audible alarm uses a continuous medium-frequency tone; the visual alarm is a yellow flashing light, and a hazard investigation prompt is simultaneously pushed to the crew terminal, recording abnormal characteristics and data deviation information; Level 3 warning trigger conditions: When ≥C3 and K3≤ When K2 is less than 2, a level 3 audible and visual warning is triggered: the audible alarm uses a low-frequency beep; the visual alarm uses a slow-flashing blue light, and only abnormal data and the warning trigger timestamp are recorded. The early warning and suppression condition is: when When <C3, do not consider None of these events trigger audible or visual warnings; instead, the relevant data is marked as low-confidence anomalies and stored in the management module. If the same abnormal condition continuously triggers the same level of warning for more than a preset duration, and... If the alert level does not decrease, the warning level will be automatically raised by one level.

[0016] Compared with the known prior art, the technical solution provided by this invention has the following beneficial effects: This invention integrates various types of data, including meteorological, hydrological, waterway obstacle, equipment operation, and scene images. It dynamically calibrates and corrects deviations, optimizes data accuracy based on scene complexity, dynamically allocates transmission bandwidth according to priority to ensure efficient transmission of critical information, and enhances the credibility of fused data by eliminating invalid data through cross-dimensional verification. It accurately identifies navigation safety hazards and abnormal equipment conditions. The decision-making model adapts to scene factors such as ship tonnage, water type, and task priority. Furthermore, it uses graded audible and visual warnings to accurately match risk levels, taking into account both emergency response prompts and routine anomaly records. Various types of data are classified, stored, and backed up off-site to ensure data security. This effectively improves the timeliness and accuracy of ship navigation safety monitoring, dynamically adapts to the needs of different navigation scenarios, reduces invalid warning interference, and effectively reduces navigation risks. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are merely some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative effort.

[0018] Figure 1 This is a schematic diagram of a multi-source sensor fusion intelligent safety monitoring and decision-making early warning system for ships in all scenarios. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0020] The present invention will be further described below with reference to embodiments.

[0021] Example 1: This embodiment presents a multi-source sensor fusion-based intelligent safety monitoring and decision-making early warning system for ships across all scenarios, such as... Figure 1 As shown, it includes: The data acquisition module is used to collect raw data in real time, including meteorological, hydrological, waterway obstacle, ship equipment operating parameters, and scene images. The data acquisition module consists of a meteorological sensing unit, a hydrological sensing unit, an obstacle detection unit, an equipment sensing unit, and an image acquisition unit. The obstacle detection unit uses a combination of lidar and millimeter-wave radar for detection, and the image acquisition unit uses a combination of an infrared camera and a visible light camera for image capture. During the data acquisition module's operation phase, channel obstacle data and critical equipment operating parameters are assigned a first-level priority; meteorological and hydrological extreme value data and abnormal frame data from scene images are assigned a second-level priority; and routine meteorological and hydrological data and routine equipment operating parameters are assigned a third-level priority. Transmission bandwidth is dynamically allocated according to a preset ratio. ; In the formula: The actual bandwidth allocated for the k-th priority data; This represents the total available transmission bandwidth in real time. This is the preset base bandwidth ratio for the k-th priority level; The real-time data volume requirement coefficient for the k-th priority level is the ratio of the real-time data transmission rate to the regular data transmission rate for the current level, with the regular data transmission rate set as the baseline value; 3 indicates the preset three-level priority. The above formula combines the three-level priority division of ship navigation data, refers to the preset basic bandwidth ratio of each level of data, and associates the real-time data volume demand reflected by the ratio of real-time data transmission rate to conventional transmission rate. It dynamically allocates the corresponding bandwidth from the total real-time available transmission bandwidth in proportion, which not only ensures the transmission resources of first-priority data such as waterway obstacle data and key equipment operating parameters, but also flexibly adjusts according to the actual transmission needs of different priority data, adapting to the transmission adaptability of data of different importance in ship navigation. The adaptation module is used to receive data collected by the acquisition module, dynamically calibrate sensor deviations based on the navigation scenario, adapt to the data dimensions required for risk identification, and output calibrated data. The dynamic calibration in the adaptation module is performed based on sensor bias and the complexity of the navigation scenario. Since deviations in equipment operating parameters can indirectly increase scenario complexity by affecting the ship's handling stability, the calibration formula is as follows: ; In the formula: Data after calibration; This refers to the raw data collected by the sensor. This is the deviation correction factor; This represents the scene complexity coefficient. The above formula is based on the original data collected by the sensor, takes into account the correlation between the historical average deviation of the sensor and the average value of the standard reference value, and determines the deviation correction coefficient by combining the accuracy weight coefficient corresponding to the actual measurement accuracy of the sensor. At the same time, it incorporates the scenario complexity coefficient, which is jointly affected by ship operation, marine environment and equipment status. By adjusting the original data in conjunction with the two, the calibration result can fit the dynamic changes of the navigation scenario, effectively offsetting the impact of the sensor's own deviation and scenario complexity, thereby improving the accuracy of the data and scenario adaptability. in, , This represents the historical average deviation value of the sensor. This represents the average value of the sensor's standard reference value. This represents the sensor accuracy weighting coefficient. ∈[0,1], the closer the actual measurement accuracy of the sensor is to the preset standard accuracy and the smaller the measurement error, the larger the value will be, and vice versa; The above formula calculates the deviation correction coefficient by combining the ratio of the sensor's historical average deviation value to the average value of the sensor's standard reference value with the sensor's accuracy weighting coefficient. This ensures that the deviation correction can accurately match the actual performance of the sensor, making the corrected result more consistent with the real situation of data acquisition. Scene complexity coefficient The value of is quantified by the following formula: ; In the formula: As a scene-related influencing factor; For the ship's real-time navigation speed; Real-time effective wave height; This refers to the deviation between the equipment's real-time operating parameters and standard parameters. This refers to the ship's normal sailing speed; The normal significant wave height; These are the standard operating parameters for the equipment; The above formula comprehensively considers the deviation of the ship's real-time navigation speed from the conventional navigation speed, the difference between the real-time significant wave height and the conventional significant wave height, and the deviation of the equipment's real-time operating parameters from the standard parameters. It is then weighted by a preset scenario influence factor that sums to 1, and quantified by taking the square root of the sum of squares. This comprehensively integrates the influence of three core factors—ship operating status, marine environmental conditions, and equipment operating parameters—on the navigation scenario, thereby accurately presenting the complexity of the navigation scenario under different working conditions. in, All are positive numbers, and their sum is 1. The above formula uses conventional and standard values, which are all preset fixed values. The fusion module is used to generate fused data by eliminating invalid information through cross-dimensional data verification based on calibrated data. The fusion module performs an invalid information removal phase, which initially removes data based on data credibility. Calibration data of the k-th dimension with a credibility lower than the preset credibility threshold is judged as invalid information and is directly excluded from the fusion calculation. Data credibility ; The above formula is based on the ratio of the standard deviation to the mean of the calibration data in the k-th dimension. Combined with the inherent fluctuation characteristics of the data itself and the fluctuation tolerance coefficient set for the core correlation of navigation safety, the data credibility is constrained to a reasonable range of 0 to 1 through a specific calculation method. This can not only eliminate invalid information caused by excessive data fluctuation, but also flexibly adjust the screening criteria according to the importance of the data to navigation safety, avoiding the accidental deletion of valid data with significant inherent fluctuations but critical data. Next, perform cross-dimensional consistency checks on the remaining valid dimension data, and calculate the relative deviation between any two valid dimension data. ,like If the deviation exceeds the preset consistency deviation threshold, the one with the larger deviation will be judged as invalid information and removed. The above formula calculates the relative deviation for any two calibrated data of different dimensions by comparing the difference between the two with the average value. This avoids the drawback of absolute deviation being affected by differences in data magnitude and can objectively reflect the degree of consistency between data of different dimensions. When the deviation exceeds the preset threshold, invalid data with large deviations are promptly removed to ensure that the data participating in subsequent fusion calculations have collaborative effectiveness and lay a reliable foundation for multi-dimensional data fusion. In the formula: The standard deviation of the data for the k-th dimension; The average value of the calibration data in the k-th dimension; The fluctuation tolerance coefficient for the k-th dimension data; This is the calibrated data for the k1th dimension; This is the calibrated data for the k2th dimension, where k1 and k2 represent two different dimensions; in, The value is larger when the inherent fluctuation characteristics of the data are more significant and the correlation with the core of navigation safety is lower; the value is smaller when the inherent stability of the data is stronger and the impact on navigation safety decision is more critical. In the data fusion stage of the fusion module, various types of data are normalized to ensure that the values ​​of all types of data are constrained to the range of [0,1] before fusion. ; In the formula: To integrate data; The k-th dimension is the fused input data, which corresponds to the normalization result. The above formula normalizes the effective calibration data of each dimension to the range of 0 to 1, eliminates the fusion interference caused by the difference in the units of different types of data, and then assigns dynamic weights according to the importance of each dimension of data and performs weighted summation to obtain fused data, so that the fusion result can comprehensively reflect the key information of multiple dimensions to adapt to the difference in the importance of different dimensions of data. During the data generation phase of the fusion module, the calculated fusion data is standardized according to a preset format, which includes data dimension identifiers, fusion timestamps, and overall credibility level. ; in, This represents the total number of valid data dimensions participating in the fusion process after removing invalid information. This indicates the credibility of the data in the k-th dimension. Let represent the dynamic fusion weight of the k-th dimension, and >0, and ; The module is used to identify navigation safety hazards and abnormal equipment conditions in real time based on fused data, and to build a decision model adapted to ship navigation scenarios. During the module runtime phase, based on fused data Extracting multidimensional feature vectors, including: Navigation safety hazard feature vector Each dimension corresponds to a characteristic value related to navigation safety, such as the deviation of the distance to the waterway obstacle, the deviation of the extreme meteorological value, the mutation rate of hydrological parameters, and the navigation density of the waterway. Equipment abnormal operating condition feature vector Each dimension corresponds to characteristic values ​​related to equipment operating conditions, such as equipment operating parameter fluctuation rate, key component temperature deviation, abnormal energy consumption growth, and equipment operating vibration value. Real-time identification of navigation safety hazards and abnormal equipment conditions is achieved through comprehensive anomaly calculation. Comprehensive anomaly level: ; In the formula; For overall anomaly degree; Weighting of safety hazard characteristics; This is a standard feature vector for safety hazards; Weights for equipment anomaly features; This represents the standard feature vector of the equipment under normal operating conditions. The Euclidean norm of a vector; The above formula extracts the feature vectors corresponding to navigation safety hazards and abnormal equipment conditions, calculates the Euclidean norm ratio of each vector to the corresponding standard feature vector, and then weights them according to the preset safety hazard feature weights and equipment abnormal feature weights. The overall degree of abnormality is comprehensively quantified by taking the square root of the sum of squares, which fully covers the two core risk dimensions of navigation safety and equipment operation, and can accurately identify abnormal conditions caused by single risks or compound risks. , All are positive numbers and their sum is 1. When the value exceeds the preset initial anomaly detection threshold, it is determined that there is a potential navigation safety hazard or an abnormal equipment condition. Decision Model ; In the formula: This is the decision-making level coefficient; The above formula takes the deviation between the comprehensive anomaly degree and the dynamic adaptation threshold as the core, and combines the scenario adaptation factor to obtain the decision level coefficient. The dynamic adaptation threshold changes dynamically with the characteristics of the ship, the water environment and the type of mission. The scenario adaptation factor integrates the scenario complexity and the fixed coefficient, so that the decision level can accurately adapt to different navigation conditions, reflect the actual risk level of the abnormal situation, and provide quantitative support for graded early warning. Scene dynamic adaptation threshold ; The above formula takes the preset initial anomaly judgment threshold as the benchmark, introduces the scenario adaptation adjustment coefficient, and combines the ratio of the actual tonnage of the ship, the complexity of the navigation water area, the priority of the navigation task and the corresponding standard reference value. By allocating the influence weights of ship tonnage, water area type and task priority, the anomaly judgment threshold is dynamically adjusted, breaking the limitation of the fixed threshold and allowing the threshold to be flexibly adapted to the changes of the core factors of the navigation scenario, thereby improving the fit between anomaly judgment and actual navigation conditions. Scene adaptation factor ; The above formula uses the scenario complexity influence coefficient to weight and fuse the scenario complexity coefficient and the fixed coefficient. The scenario complexity influence coefficient is flexibly adjusted according to the diversity of ship navigation scenarios and the frequency of changes in operating conditions. Higher values ​​are assigned to ships with diverse scenario types and cross-water navigation, while lower values ​​are assigned to ships with fixed routes and simple scenarios. This allows the scenario adaptation factor to reflect the complexity of the current navigation scenario while maintaining appropriate stability to adapt to the needs of ships with different navigation modes. In the formula: To preset the initial anomaly detection threshold; Adjust the coefficients to suit the scene; , , The influence weights are determined by ship tonnage, type of navigation waterway, and priority of navigation mission. This refers to the current actual tonnage of the vessel; The complexity coefficient of the navigation waterway; This refers to the priority coefficient for navigation missions. For standard reference tonnage; The standard water area complexity coefficient; This refers to the standard task priority coefficient. This represents the influence coefficient of scene complexity. This represents the scene complexity coefficient. in, The value of ∈(0,1] is larger for navigation scenarios with higher safety sensitivity requirements to changes in navigation scenarios and more stringent ship safety control levels, and smaller for navigation scenarios with lower safety sensitivity requirements to changes in navigation scenarios and routine safety control measures. , , All are positive numbers, and their sum is 1; The values ​​are preset intervals, dimensionless, with 1 for open sea areas, 1.5~2 for nearshore waters, and 2.0~3.0 for narrow channels; These are preset values, dimensionless. 1 is used for routine transportation, 1.8 for emergency rescue, and 2.2 for hazardous materials transportation. ∈[0,1], the larger the value is for ships with diverse navigation scenarios and varied cross-water navigation conditions, the smaller the value is for ships with simple navigation scenarios and fixed routes. The early warning module is used to trigger graded audible and visual early warnings based on the confidence level of the joint decision-making model that obtains the identification results of navigation safety hazards and abnormal equipment operating conditions. The confidence level of the decision model in the early warning module is: ; In the formula: Confidence level of the decision model; , , To integrate data credibility weight, feature matching credibility weight, and scenario adaptation credibility weight, , , The values ​​of all are in the range of (0,1), and their sum is 1. To integrate the overall credibility level of the data; To assess the credibility of feature matching; For scene adaptation factors; The above formula integrates three core dimensions: the overall credibility level of the data, the credibility of feature matching, and the scenario adaptation factor. It calculates the confidence level of the decision model through the pre-set weight allocation of each dimension, and comprehensively covers the reliability of the data itself, the accuracy of feature matching with the standard model, and the rationality of adaptation to the navigation scenario. in, ; The above formula calculates the Euclidean norm ratio of the feature vector of navigation safety hazards, the feature vector of abnormal equipment operation, and the corresponding standard feature vector. It takes the minimum value of each of them within the range of 0 to 1 and calculates the average value. Then, it subtracts the average value from 1 to obtain the feature matching confidence. The smaller the deviation between the feature vector and the standard vector, the higher the matching confidence. It transforms the abstract feature matching situation into an intuitive and quantifiable confidence index, which is necessary support for the calculation of the confidence of the decision model. The triggering logic for tiered audible and visual warnings is as follows: The preset warning level thresholds include: high confidence threshold C1, medium confidence threshold C2, and low confidence threshold C3; the preset decision level coefficient thresholds include: high risk threshold K1, medium risk threshold K2, and low risk threshold K3. Level 1 warning trigger conditions: when ≥C1 and When K1 is ≥1, a Level 1 audible and visual warning is triggered: the audible alarm uses a high-frequency pulse tone (frequency 2000Hz~3000Hz) and is continuously output without interruption; the visual alarm uses a red strobe light (flashing frequency 5Hz~8Hz), and the ship's control system simultaneously displays an emergency braking prompt interface, recording the fused data and feature vector at the time of the warning trigger. Level 2 warning trigger conditions: when ≥C2 and K2≤ When <K1, a level 2 audible and visual warning is triggered: the audible alarm uses a medium-frequency continuous tone (frequency 1000Hz~2000Hz) with an interval of "3 seconds output + 2 seconds pause"; the visual alarm is a yellow flashing light (flashing frequency 2Hz~5Hz), and a hidden danger investigation prompt is pushed to the crew terminal at the same time, and abnormal characteristics and data deviation information are recorded. Level 3 warning trigger conditions: When ≥C3 and K3≤ When the threshold is <K2, a three-level audible and visual alarm is triggered: the audible alarm uses a low-frequency prompt tone (frequency 500Hz~1000Hz) with an interval of "1 second output + 5 seconds pause"; the visual alarm uses a slow blue flashing light (flashing frequency 1Hz~2Hz), and only records abnormal data and the alarm trigger timestamp. The early warning and suppression condition is: when When <C3, do not consider None of these events trigger audible or visual warnings; instead, the relevant data is marked as low-confidence anomalies and stored in the management module. If the same abnormal condition continuously triggers the same level of warning for more than a preset duration, and... If the alert level does not decrease, automatically raise it by one level. The management module is used to classify, store, and back up raw data, calibration data, fused data, and decision models in different locations. The data acquisition module is interconnected with the adaptation module and the fusion module via a local area network. The adaptation module is interconnected with the fusion module via a local area network. The adaptation module and the fusion module are interconnected with the construction module, the early warning module and the management module via a wireless network. The construction module is interconnected with the early warning module via a local area network. The early warning module is interconnected with the management module via a local area network. Meteorological data includes at least wind speed, wind direction, temperature, humidity, and air pressure, while hydrological data includes at least water temperature, salinity, depth, flow velocity, and flow direction.

[0022] In this embodiment, the acquisition module collects raw data in real time, including meteorological, hydrological, waterway obstacle, ship equipment operating parameters, and scene images. The adaptation module, running after the acquisition module, receives the data, dynamically calibrates sensor deviations based on the navigation scenario, adapts to the data dimensions required for risk identification, and outputs calibrated data. The fusion module further uses the calibrated data to eliminate invalid information through cross-dimensional data verification, generating fused data. The construction module then uses the fused data to identify navigation safety hazards and abnormal equipment conditions in real time, constructs a decision model adapted to the ship navigation scenario, and triggers graded audible and visual warnings by obtaining the confidence level of the joint decision model based on the identification results of navigation safety hazards and abnormal equipment conditions through the early warning module. Finally, the management module classifies, stores, and backs up the raw data, calibration data, fused data, and decision model in different locations.

[0023] In the above embodiments, the system can capture various key navigation-related data in real time and accurately. Through dynamic calibration and fusion processing, it can quickly identify safety hazards and equipment anomalies. The system also provides more targeted responses through tiered early warnings. The secure storage and backup of the data facilitates traceability and effectively improves the safety and operational efficiency of ship navigation.

[0024] Application example: The XX3000-ton conventional cargo ship (actual tonnage is consistent with standard reference tonnage) sails in near-shore waters (water complexity coefficient 1.8, standard water complexity coefficient 1) and performs conventional transportation tasks (task priority coefficient 1, standard task priority coefficient 1). The system's preset parameters are as follows: high confidence threshold 0.9, medium confidence threshold 0.7, low confidence threshold 0.5; high risk threshold 0.8, medium risk threshold 0.4, low risk threshold 0.2; preset initial anomaly judgment threshold 0.3; scenario adaptation adjustment coefficient 0.6; influence weights of ship tonnage, navigation water type, and navigation task priority are 0.3, 0.4, and 0.3, respectively; scenario complexity influence coefficient 0.5; safety hazard feature weight 0.6, equipment anomaly feature weight 0.4; fusion data credibility, feature matching credibility, and scenario adaptation credibility weights are 0.4, 0.3, and 0.3, respectively; data fluctuation tolerance coefficient is 0.3 for all; sensor accuracy weight coefficient is 0.85; preset basic bandwidth ratios for the three priority levels are 0.5, 0.3, and 0.2, respectively; consistency deviation threshold is 0.2.

[0025] During navigation, the data acquisition module collects data in real time through various units: meteorological data includes wind speed of 8 m / s, wind direction of northeast-east, temperature of 25℃, humidity of 65%, and air pressure of 1012 hPa; hydrological data includes water temperature of 22℃, salinity of 32‰, depth of 15 m, current velocity of 1.2 m / s, and current direction of southwest; lidar and millimeter-wave radar work together to detect floating objects 500 m ahead; equipment operating parameters include main unit speed of 1200 r / min, key component temperature of 75℃, energy consumption of 80 kW, and vibration value of 0.3 mm / s; infrared and visible light cameras work together to capture uncaptured abnormal frames. The system dynamically allocates a total transmission bandwidth of 100 Mbps according to priority, with first-priority data receiving 55 Mbps, second-priority data receiving 25 Mbps, and third-priority data receiving 20 Mbps.

[0026] After receiving the data, the adaptation module calculates a deviation correction coefficient of 0.02. Combined with the ship's real-time sailing speed of 12 knots (normally 10 knots), real-time effective wave height of 1.5m (normally 1m), and equipment operating parameter deviation of 5%, the scenario complexity coefficient is calculated to be 0.68. Finally, the data calibration is completed, the main engine speed is calibrated to 1224 r / min, the temperature of key components is calibrated to 76.5℃, and the remaining parameters are all within a reasonable range after calibration.

[0027] The fusion module first determined that the credibility of the calibration data in each dimension was higher than 0.6, and no data was initially rejected. In the cross-dimensional consistency check, the relative deviation between hydrological velocity and flow direction data was 0.12, which is less than the consistency deviation threshold, so all data were valid. After normalization and combined with dynamic fusion weights (channel obstacle data 0.4, equipment parameters 0.4, meteorological data 0.2, hydrological data 0.2, total weight 1.2), fused data with an overall credibility level of 0.82 was generated.

[0028] After extracting feature vectors, the construction module calculates a comprehensive anomaly score of 0.35, a scene dynamic adaptation threshold of 0.32, a decision level coefficient of 0.09, and a scene adaptation factor of 0.84. The early warning module calculates a decision model confidence score of 0.84, which is higher than the medium confidence threshold, but the decision level coefficient is lower than the low-risk threshold. Therefore, it does not meet any level of early warning triggering conditions and only marks the relevant data as normal operating conditions. The management module classifies and stores the raw data, calibration data, fused data, and decision model, performs off-site backups, and synchronously records the timestamps and processing results of all data.

[0029] In summary, the above embodiments integrate multiple types of data, including meteorological, hydrological, waterway obstacle, equipment operation, and scene images. It dynamically calibrates and corrects deviations, optimizes data accuracy based on scene complexity, dynamically allocates transmission bandwidth according to priority to ensure efficient transmission of critical information, and enhances the credibility of fused data by eliminating invalid data through cross-dimensional verification. It accurately identifies navigation safety hazards and abnormal equipment conditions, adapts the decision-making model to scene factors such as ship tonnage, water type, and task priority, and accurately matches risk levels with tiered audible and visual warnings. It also considers both emergency response prompts and routine anomaly records. Various types of data are categorized, stored, and backed up off-site to ensure data security. This effectively improves the timeliness and accuracy of ship navigation safety monitoring, dynamically adapts to different navigation scenario needs, reduces invalid warning interference, and effectively lowers navigation risks.

[0030] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions will not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A multi-source sensor fusion intelligent safety monitoring and decision-making early warning system for ships in all scenarios, characterized in that, include: The data acquisition module is used to collect raw data in real time, including meteorological, hydrological, waterway obstacle, ship equipment operating parameters, and scene images. The adaptation module is used to receive data collected by the acquisition module, dynamically calibrate sensor deviations based on the navigation scenario, adapt to the data dimensions required for risk identification, and output calibrated data. The fusion module is used to generate fused data by eliminating invalid information through cross-dimensional data verification based on calibrated data. The module is used to identify navigation safety hazards and abnormal equipment conditions in real time based on fused data, and to build a decision model adapted to ship navigation scenarios. The early warning module is used to trigger graded audible and visual early warnings based on the confidence level of the joint decision-making model that obtains the identification results of navigation safety hazards and abnormal equipment operating conditions. The management module is used to classify, store, and back up raw data, calibration data, fused data, and decision models in different locations. Meteorological data includes at least wind speed, wind direction, temperature, humidity, and air pressure, while hydrological data includes at least water temperature, salinity, depth, flow velocity, and flow direction.

2. The intelligent safety monitoring and decision-making early warning system for ships in all scenarios based on multi-source sensor fusion as described in claim 1, characterized in that, The acquisition module consists of a meteorological sensing unit, a hydrological sensing unit, an obstacle detection unit, an equipment sensing unit, and an image acquisition unit. The obstacle detection unit uses a combination of lidar and millimeter-wave radar for detection, and the image acquisition unit uses a combination of an infrared camera and a visible light camera for image capture. During the operation of the acquisition module, waterway obstacle data and key equipment operating parameters are set as first-level priority, meteorological and hydrological extreme value data and abnormal frame data of scene images are set as second-level priority, and routine meteorological and hydrological data and routine equipment operating parameters are set as third-level priority. The transmission bandwidth is dynamically allocated according to a preset ratio.

3. The intelligent safety monitoring and decision-making early warning system for ships in all scenarios based on multi-source sensor fusion as described in claim 1, characterized in that, The dynamic calibration in the adaptation module is performed based on sensor bias and navigation scenario complexity, and the calibration formula is: ； In the formula: Data after calibration; This refers to the raw data collected by the sensor. This is the deviation correction factor; This represents the scene complexity coefficient. in, , This represents the historical average deviation value of the sensor. This represents the average value of the sensor's standard reference value. This represents the sensor accuracy weighting coefficient.

4. The intelligent safety monitoring and decision-making early warning system for ships in all scenarios based on multi-source sensor fusion as described in claim 3, characterized in that, The scene complexity coefficient The value of is quantified by the following formula: ； In the formula: As a scene-related influencing factor; For the ship's real-time navigation speed; Real-time effective wave height; This refers to the deviation between the equipment's real-time operating parameters and standard parameters. This refers to the ship's normal sailing speed; The normal significant wave height; These are the standard operating parameters for the equipment; in, All are positive numbers, and their sum is 1.

5. The intelligent safety monitoring and decision-making early warning system for ships in all scenarios based on multi-source sensor fusion as described in claim 1, characterized in that, The fusion module performs an invalid information removal phase, which performs preliminary removal based on data credibility. Calibration data of the k-th dimension with a credibility lower than a preset credibility threshold is judged as invalid information and directly excluded from the fusion calculation. The credibility of the data ; Next, perform cross-dimensional consistency checks on the remaining valid dimension data, and calculate the relative deviation between any two valid dimension data. ,like If the deviation exceeds the preset consistency deviation threshold, the one with the larger deviation will be judged as invalid information and removed. In the formula: The standard deviation of the data for the k-th dimension; The average value of the calibration data in the k-th dimension; The fluctuation tolerance coefficient for the k-th dimension data; This is the calibrated data for the k1th dimension; This is the calibrated data for the k2th dimension, where k1 and k2 represent two different dimensions.

6. The intelligent safety monitoring and decision-making early warning system for ships in all scenarios based on multi-source sensor fusion as described in claim 5, characterized in that, In the data fusion stage of the fusion module, various types of data are normalized to ensure that the values ​​of all types of data are constrained within the range of [0,1] before fusion. ； In the formula: To integrate data; The k-th dimension is the fused input data, which corresponds to the normalization result. During the data generation phase of the fusion module, the calculated fusion data is standardized according to a preset format, which includes data dimension identifiers, fusion timestamps, and overall credibility levels. ; in, This represents the total number of valid data dimensions participating in the fusion process after removing invalid information. This indicates the credibility of the data in the k-th dimension. This represents the dynamic fusion weight of the k-th dimension.

7. The intelligent safety monitoring and decision-making early warning system for ships in all scenarios based on multi-source sensor fusion as described in claim 1, characterized in that, During the runtime phase of the construction module, based on fused data Extracting multidimensional feature vectors, including: Navigation safety hazard feature vector ; Equipment abnormal operating condition feature vector ; Real-time identification of navigation safety hazards and abnormal equipment conditions is achieved through comprehensive anomaly calculation. Comprehensive anomaly level: ； In the formula; For overall anomaly degree; Weights for safety hazard characteristics; This is a standard feature vector for safety hazards; Weights for equipment anomaly features; This represents the standard feature vector of the equipment under normal operating conditions. The Euclidean norm of a vector; , All are positive numbers and their sum is 1. When the value exceeds the preset initial anomaly detection threshold, it is determined that there is a potential navigation safety hazard or an abnormal equipment condition. The decision model ; In the formula: This is the decision-making level coefficient; Scene dynamic adaptation threshold ; Scene adaptation factor ; In the formula: To preset the initial anomaly detection threshold; Adjust the coefficients to suit the scene; , , The influence weights are determined by ship tonnage, type of navigation waterway, and priority of navigation mission. This refers to the current actual tonnage of the vessel; The complexity coefficient of the navigation waterway; This refers to the priority coefficient for navigation missions. For standard reference tonnage; The standard water area complexity coefficient; This refers to the standard task priority coefficient. This represents the influence coefficient of scene complexity. This represents the scene complexity coefficient.

8. A multi-source sensor fusion-based intelligent safety monitoring and decision-making early warning system for ships in all scenarios, as described in claim 1 or 7, characterized in that, The confidence level of the decision model in the early warning module is: ； In the formula: Confidence level of the decision model; , , To integrate data credibility weight, feature matching credibility weight, and scenario adaptation credibility weight, , , The values ​​of all are in the range of (0,1), and their sum is 1. To integrate the overall credibility level of the data; To assess the credibility of feature matching; For scene adaptation factors; in, .

9. A multi-source sensor fusion-based intelligent safety monitoring and decision-making early warning system for ships in all scenarios, as described in claim 8, is characterized in that, The triggering logic for the tiered audible and visual early warning is as follows: The preset warning level thresholds include: high confidence threshold C1, medium confidence threshold C2, and low confidence threshold C3; the preset decision level coefficient thresholds include: high risk threshold K1, medium risk threshold K2, and low risk threshold K3. Level 1 warning trigger conditions: when ≥C1 and When K1 is ≥1, a Level 1 audible and visual warning is triggered: the audible alarm uses a high-frequency pulse tone, which is continuously output without interruption; the visual alarm uses a red strobe light, which is synchronized with the ship's control system to pop up an emergency braking prompt interface and record the fused data and feature vector at the time of the warning trigger. Level 2 warning trigger conditions: when ≥C2 and K2≤ When the threshold is <K1, a Level 2 audible and visual warning is triggered: the audible alarm uses a continuous medium-frequency tone; the visual alarm is a yellow flashing light, and a hazard investigation prompt is simultaneously pushed to the crew terminal, recording abnormal characteristics and data deviation information; Level 3 warning trigger conditions: When ≥C3 and K3≤ When K2 is less than 2, a level 3 audible and visual warning is triggered: the audible alarm uses a low-frequency beep; the visual alarm uses a slow-flashing blue light, and only abnormal data and the warning trigger timestamp are recorded. The early warning and suppression condition is: when When <C3, do not consider None of these events trigger audible or visual warnings; instead, the relevant data is marked as low-confidence anomalies and stored in the management module. If the same abnormal condition continuously triggers the same level of warning for more than a preset duration, and... If the alert level does not decrease, the warning level will be automatically raised by one level.

10. A multi-source sensor fusion intelligent safety monitoring and decision-making early warning system for ships in all scenarios, as described in claim 1, is characterized in that, The data acquisition module is interconnected with an adaptation module and a fusion module via a local area network. The adaptation module is interconnected with the fusion module via a local area network. The adaptation module and the fusion module are interconnected with a construction module, an early warning module, and a management module via a wireless network. The construction module is interconnected with the early warning module via a local area network. The early warning module is interconnected with the management module via a local area network.

Images