Ship safety early warning system based on feature matching

Abstract

The invention discloses a ship safety early warning system based on feature matching, relates to the technical field of ship navigation safety monitoring, and aims to solve the technical problems of recognition lag and insufficient accuracy caused by single feature monitoring of an existing system, the method is realized based on a BBBB device, and the ship safety early warning system comprises an information acquisition module, an information processing module and a processing module, the data acquisition module is used for acquiring each piece of ship state feature information of ship dangerous states from historical ship safety reports as each piece of ship state dangerous feature data, and acquiring static feature data, historical feature data and dynamic feature data of ships at the same time; the data preprocessing module is used for carrying out normalization processing on the danger characteristic data acquired by the information acquisition module to obtain danger characteristic standard data of the state of each ship; and the dynamic identification module comprises a first dynamic monitoring module, a second dynamic monitoring module, a first dynamic data storage module and a second dynamic data storage module. The method has the advantage of improving the accuracy and timeliness of early warning.

Description

Technical Field

[0001] This invention relates to the field of ship navigation safety monitoring technology, and more specifically, to a ship safety early warning system based on feature matching. Background Technology

[0002] In the field of ship navigation safety monitoring, existing ship safety early warning systems mostly rely on monitoring single dynamic features (such as real-time speed) or analyzing static features (such as ship size), lacking comprehensive matching analysis of ship static features, historical navigation features, and real-time dynamic features. This single-dimensional monitoring method is difficult to cope with the dynamic changes in ship status under complex navigation environments, resulting in a lag in the identification of dangerous ship conditions. Furthermore, due to incomplete feature matching, it is prone to misjudgment or omission, lacking accuracy and failing to accurately warn of potential risks such as collisions, overloading, and abnormal navigation attitudes. In view of this, we propose a ship safety early warning system based on feature matching. Summary of the Invention

[0003] The purpose of this invention is to provide a ship safety early warning system based on feature matching to solve the technical problems of recognition lag and insufficient accuracy caused by the monitoring of single features in existing systems.

[0004] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a ship safety early warning system based on feature matching, comprising: The information acquisition module is used to obtain the ship status characteristic information of each ship in dangerous state from historical ship safety reports as the ship status dangerous characteristic data, and at the same time obtain the ship's static characteristic data, historical characteristic data and dynamic characteristic data. The data preprocessing module is used to normalize the hazard characteristic data acquired by the information acquisition module to obtain standard data of each ship state hazard characteristic. The dynamic identification module includes a first dynamic monitoring module, a second dynamic monitoring module, a first dynamic data storage module, and a second dynamic data storage module. It is used to monitor the dynamic characteristic data of ships through training feature models and output the matching result with the standard data of ship dangerous state characteristics as the judgment result. The ship status characteristic determination module is used to match ship status characteristic data with hazard characteristic standard data, determine ship status risk value groups, train to obtain ship status safety characteristic distribution, and perform cluster analysis on the comprehensive navigation safety index dataset to obtain ship status abnormal characteristic data. The comparison and analysis module is used to compare and analyze the judgment results output by the dynamic identification module and the abnormal ship status feature data output by the ship status feature determination module, and transmit the comparison and analysis results after feature matching cross-validation. The ship danger status early warning module is used to determine the ship danger status index through feature matching based on the comparative analysis results and the current comprehensive navigation safety index dataset of the ship, and to provide early warning of dangerous situations.

[0005] Preferably, it also includes a main data storage module for storing static feature data, historical feature data, dynamic feature data, and intermediate data generated during system operation acquired by the information acquisition module; The first dynamic data storage module is used to store the dynamic feature data and judgment results processed by the first dynamic monitoring module. The second dynamic data storage module is used to store the dynamic feature data and judgment results processed by the second dynamic monitoring module.

[0006] The preferred static feature data covers the inherent attribute information of the ship, including the ship number used to uniquely identify the ship, the ship length and ship width reflecting the physical dimensions of the ship, the ship deadweight and ship weight reflecting the ship's carrying capacity, the ship type representing the ship type, and the draft and water level information related to the ship's buoyancy. The historical feature data includes navigation feature data generated during the ship's past voyages and detection feature data recorded during the detection process. The navigation feature data includes the ship's voyage distance, voyage time, voyage speed, and steering state and angle corresponding to the steering operation. The detection feature data includes the ship's draft and water level data recorded. The dynamic characteristic data includes the ship's real-time speed and real-time steering information during real-time navigation.

[0007] Preferably, the first dynamic monitoring module consists of a first data processing unit, a first data output unit, and a first judgment unit; The working process of the first data processing unit is as follows: it constructs an initial data sample based on the historical feature data of the ship, uses at least one of the static feature data and dynamic feature data as a training sample, forms a feature model that can be used to identify the ship's state through training, and applies the trained feature model to the ship's state identification work. The first data output unit is responsible for obtaining dynamic feature data from the data storage module and sending the obtained dynamic feature data to the first judgment unit. At the same time, it sends a read signal to the first data storage module. After receiving the read signal, the first data storage module loads the dynamic feature data from the data storage module and transmits it to the first data output unit.

[0008] The first judgment unit judges the matching of dynamic feature data and dangerous feature standard data according to preset comparison rules to determine the safety of dynamic feature data. If the judgment result is safe, the dynamic feature data is stored in the first data storage module. If the judgment result is unsafe, an alarm signal is issued. The feature model adopts a convolutional neural network model. During the model training process, the difference between the predicted value and the actual value is evaluated by a loss function. When there is a deviation in the prediction result, the relevant parameters in the feature model are adjusted. When there is no deviation in the prediction result, the model parameters are not adjusted.

[0009] The preferred second dynamic monitoring module includes a second data processing unit, a second data output unit, a second judgment unit, and a data generation unit; The second dynamic monitoring module works in conjunction with the first dynamic monitoring module to jointly monitor the dynamic characteristics data of the ship; The second data processing unit uses the feature model trained by the first dynamic monitoring module to carry out ship status identification. The second data output unit can obtain the first dynamic feature data from the first data storage module, and can also obtain dynamic feature data from the data storage module. The second judgment unit judges the matching of the first dynamic feature data and the dynamic feature data with the dangerous feature standard data according to the feature model to determine its safety. If the judgment result is safe, the data generation unit generates new dynamic feature data. If the judgment result is unsafe, an alarm signal with a warning mark is sent to the second data output unit.

[0010] Preferably, after the first dynamic monitoring module obtains the first dynamic feature data from the first data storage module, if it decides to continue storing the first dynamic feature data, it continuously monitors the matching of the first dynamic feature data with the standard data of dangerous features. When an error occurs during the monitoring process, the information acquisition module collects the corresponding alarm information and incorporates the alarm information into the historical feature data. After the second dynamic monitoring module obtains the second dynamic feature data from the second data storage module, if it decides to continue storing the second dynamic feature data, it will continuously monitor the matching of the second dynamic feature data with the standard data of dangerous features. When an error occurs during the monitoring process, the information acquisition module will collect alarm information and store the alarm information as historical feature data.

[0011] Preferably, after receiving an alarm signal, the first judgment unit judges the safety of matching dynamic feature data with dangerous feature standard data by using the warning label carried by the alarm signal. If it is judged to be safe, the alarm signal corresponding to the warning label is obtained from the first data output unit. If it is judged to be unsafe, the safety of dynamic feature data is judged according to the warning label, and the judgment result is transmitted to the comparison analysis module for cross-validation of feature matching. After receiving the alarm signal, the second judgment unit first determines whether the first dynamic feature data is safe. If it is safe, it directly obtains the alarm signal corresponding to the warning icon from the first judgment unit. If it is not safe, it continues to monitor the dynamic feature data. When continuous monitoring finds that the dynamic feature data is incorrect, it directly sends an alarm signal to the second data output unit, and the judgment result is transmitted to the comparison and analysis module for cross-validation of feature matching.

[0012] When determining whether dynamic feature data is safe, both the first and second judgment units first determine whether the ship's turning angle is within a preset range. If it is within the preset range, the first and second data output units extract dynamic feature data that does not include the turning angle from the data storage module, and then the first and second judgment units determine the safety of this dynamic feature data.

[0013] The preferred specific working process of the information acquisition module is as follows: extracting ship status feature data corresponding to each ship status feature of each ship, as well as the ship status value of each ship, from historical ship safety reports; From the extracted ship condition feature data, ship type feature data, ship size feature data, and ship weight feature data are selected; The ship status value of each ship is used as the initial feature value of the corresponding ship status feature data; The extracted ship status feature data of each ship are sorted, and ship status feature data with a proportion lower than a preset threshold are removed. The preset threshold is determined based on the sample size of historical ship safety reports, and valid ship status feature data are retained. Extract the first feature data corresponding to the ship type feature data, ship size feature data, and ship weight feature data from the valid ship status feature data; Compare the ship state values ​​corresponding to each ship state characteristic data with the ship state values ​​corresponding to the first characteristic data; input the first characteristic data into the normalization model of each hazard characteristic data to obtain the hazard characteristic standard data corresponding to each ship state characteristic data; Based on the initial and limit characteristic values ​​of each ship state characteristic, the second characteristic data of each characteristic data in each ship state characteristic data is obtained; combining the first and second characteristic data of each characteristic data, the difference factor of each characteristic data in each ship state characteristic data is obtained. Based on the difference factor, the ship state characteristic data, and the corresponding characteristic data in the standard data of dangerous characteristics, the ship state risk value group of each ship state characteristic is obtained. The feature weights of the risk value group are trained to obtain the ship state characteristic distribution corresponding to each ship state characteristic, and the distribution is transmitted to the data preprocessing module.

[0014] The preferred workflow of the data preprocessing module is as follows: based on the current navigation data of each ship, extract the current ship feature data of each ship; sort the extracted current ship feature data of each ship to obtain the initial feature values ​​of each ship state feature under the current ship feature data of each ship.

[0015] The initial feature values ​​of each ship's current state feature are compared with the ship state values ​​of each ship's state feature under the corresponding ship feature data, and the normalized difference between the two is calculated. This difference is calculated based on the initial feature values ​​and limit feature values ​​provided by the information acquisition module. The normalized initial feature values ​​of each ship's current ship feature data and the calculated normalized difference are input into the ship comprehensive index model. The comprehensive navigation safety index of each ship is calculated by the model and then transmitted to the ship state feature determination module.

[0016] Preferably, the specific working method of the ship dangerous state early warning module is as follows: normalize the comprehensive navigation safety index of each ship to obtain the ship state feature data of each state feature in the comprehensive navigation safety index of each standard ship; compare the ship state feature data of each state feature in the comprehensive navigation safety index of each ship and the ship state feature data of each state feature in the comprehensive navigation safety index of each standard ship with the state features in the corresponding dangerous feature standard data to obtain the ship state deviation value of each state feature in the comprehensive navigation safety index of each standard ship. The ship state deviation value of each feature data in the comprehensive navigation safety index of each standard ship is input into the ship deviation degree calculation model. This model is trained based on the feature weights output by the ship state feature determination module, and the ship state safety feature weights of each standard ship state feature in the comprehensive navigation safety index of each ship are obtained. The weight of the ship's state safety feature is input into the ship safety assessment model, which is trained based on the feature model of the dynamic identification module, to obtain the probability of the ship's state safety feature for each standard ship state in the comprehensive navigation safety index of each ship. The ship state safety characteristic probability and corresponding ship state deviation value of each standard ship state in the current comprehensive navigation safety index of each ship are input into the ship safety characteristic analysis model to obtain the ship state characteristic deviation index of each standard ship state in the current comprehensive navigation safety index of each ship; by comparing these deviation indices, the current danger state index of each ship is obtained. The current danger status index of each ship is compared with the preset danger status threshold. When the danger status index is greater than the preset threshold, the corresponding ship danger warning information is given and the danger warning level of the ship is specified.

[0017] Compared with the prior art, the beneficial effects of the present invention are: 1. This invention integrates static feature data, historical feature data, and dynamic feature data of ships, and utilizes the collaborative work of the dual monitoring units of the dynamic identification module and feature matching cross-validation to achieve multi-dimensional comprehensive identification of dangerous states of ships. This solves the problems of identification lag and insufficient accuracy caused by single feature monitoring in existing systems, and improves the accuracy and timeliness of early warning.

[0018] 2. This invention also trains feature data using a convolutional neural network model and dynamically adjusts model parameters using a loss function to continuously optimize feature matching accuracy. This avoids the problem of insufficient adaptability to complex navigation scenarios caused by a fixed model and further improves the robustness of dangerous state identification.

[0019] 3. This invention also utilizes a hierarchical storage mechanism of main data storage module and dynamic data storage module, as well as the historical feature data update function of alarm information, to achieve full lifecycle management of feature data, reduce matching deviations caused by missing or outdated data, and further ensure the reliability of early warning results. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the system framework of the present invention. Detailed Implementation

[0021] To facilitate understanding of the technical solution of the present invention by those skilled in the art, the technical solution of the present invention will now be further described in conjunction with the accompanying drawings.

[0022] Example 1, such as Figure 1 As shown, the present invention provides a ship safety early warning system based on feature matching, comprising: The information acquisition module is used to obtain the ship status characteristic information of each ship in dangerous state from historical ship safety reports as the ship status dangerous characteristic data, and at the same time obtain the ship's static characteristic data, historical characteristic data and dynamic characteristic data. The data preprocessing module is used to normalize the hazard characteristic data acquired by the information acquisition module to obtain standard data of each ship state hazard characteristic. The dynamic identification module includes a first dynamic monitoring module, a second dynamic monitoring module, a first dynamic data storage module, and a second dynamic data storage module. It is used to monitor the dynamic characteristic data of ships through training feature models and output the matching result with the standard data of ship dangerous state characteristics as the judgment result. The ship status characteristic determination module is used to match ship status characteristic data with hazard characteristic standard data, determine ship status risk value groups, train to obtain ship status safety characteristic distribution, and perform cluster analysis on the comprehensive navigation safety index dataset to obtain ship status abnormal characteristic data. The comparison and analysis module is used to compare and analyze the judgment results output by the dynamic identification module and the abnormal ship status feature data output by the ship status feature determination module, and transmit the comparison and analysis results after feature matching cross-validation. The ship danger status early warning module is used to determine the ship danger status index through feature matching based on the comparative analysis results and the current comprehensive navigation safety index dataset of the ship, and to provide early warning of dangerous situations.

[0023] The information acquisition module is connected to the data preprocessing module and the ship state characteristic determination module. The data preprocessing module is connected to the information acquisition module, the dynamic identification module, and the ship state characteristic determination module. The first dynamic monitoring module and the second dynamic monitoring module are both connected to the data preprocessing module, the comparison analysis module, the first dynamic data storage module, and the second dynamic data storage module. The first dynamic data storage module and the second dynamic data storage module are both connected to the main data storage module. The ship state characteristic determination module is connected to the information acquisition module, the data preprocessing module, and the comparison analysis module. The comparison analysis module is connected to the dynamic identification module, the ship state characteristic determination module, and the ship dangerous state early warning module. The ship dangerous state early warning module is connected to the comparison analysis module.

[0024] In embodiments of the present invention, it further includes: The main data storage module is used to store static feature data, historical feature data, dynamic feature data, and intermediate data generated during system operation acquired by the information acquisition module. The first dynamic data storage module is used to store the dynamic feature data and related judgment results processed by the first dynamic monitoring module. The second dynamic data storage module is used to store the dynamic feature data and related judgment results processed by the second dynamic monitoring module.

[0025] In embodiments of the present invention, the static feature data covers the inherent attribute information of the ship, including the ship number used to uniquely identify the ship, the ship length and ship width reflecting the physical dimensions of the ship, the ship deadweight and ship weight reflecting the ship's carrying capacity, the ship type characterizing the ship type, and the draft and water level information related to the ship's buoyancy. The historical feature data includes navigation feature data generated during the ship's past voyages and detection feature data recorded during the detection process. The navigation feature data includes the ship's voyage distance, voyage time, voyage speed, and steering state and angle corresponding to the steering operation. The detection feature data includes the ship's draft and water level data recorded. The dynamic characteristic data includes the ship's real-time speed and real-time steering information during real-time navigation.

[0026] In an embodiment of the present invention, the first dynamic monitoring module comprises a first data processing unit, a first data output unit, and a first judgment unit; The working process of the first data processing unit is as follows: an initial data sample is constructed based on the historical feature data of the ship; at least one of the static feature data and dynamic feature data is used as a training sample; a feature model that can be used to identify the ship's state is formed through training; the training objective is to improve the matching accuracy of dynamic feature and hazard feature standard data; and the trained feature model is applied to the ship state identification work. The first data output unit is responsible for obtaining dynamic feature data from the data storage module and sending the obtained dynamic feature data to the first judgment unit. At the same time, it sends a read signal to the first data storage module. After receiving the read signal, the first data storage module loads the dynamic feature data from the data storage module and transmits it to the first data output unit.

[0027] The first judgment unit judges the matching of dynamic feature data and dangerous feature standard data according to preset comparison rules to determine the safety of dynamic feature data. If the judgment result is safe, the dynamic feature data is stored in the first data storage module. If the judgment result is unsafe, an alarm signal is issued.

[0028] In an embodiment of the present invention, the feature model adopts a convolutional neural network model. During the model training process, the difference between the predicted value and the actual value is evaluated by a loss function. When there is a deviation in the prediction result, the relevant parameters in the feature model are adjusted. When there is no deviation in the prediction result, the model parameters are not adjusted. The formula for calculating the loss function is as follows: ; in, This represents the difference between the predicted value obtained through the feature model and the actual value. For the same type of physical quantity, for example, when monitoring ship speed, the unit can be m / s; when monitoring ship length, the unit can be m, etc., depending on the ship's state characteristics being monitored. The actual value representing the ship's condition characteristics, compared to the predicted value. They belong to the same type of physical quantity and have the same unit, and are used as a benchmark to measure the accuracy of the predicted value; This indicates a deviation, with a value of either 1 or 0. When a deviation exists in the data sample—that is, the difference between the predicted and actual values—it needs to be addressed and processed. A value of 1 indicates that the data sample is unbiased, meaning the predicted value is basically consistent with the actual value, and no parameter adjustment is required. The value is 0; This represents the loss value, and its unit is the same as the predicted value. Actual value The units are consistent and are used to quantify the degree of difference between the predicted value and the actual value. The larger the loss value, the more serious the deviation between the prediction result and the actual situation.

[0029] In an embodiment of the present invention, the second dynamic monitoring module includes a second data processing unit, a second data output unit, a second judgment unit, and a data generation unit; The second dynamic monitoring module works in conjunction with the first dynamic monitoring module to jointly monitor the dynamic characteristics data of the ship; The second data processing unit uses the feature model trained by the first dynamic monitoring module to carry out ship status identification. The second data output unit can obtain the first dynamic feature data from the first data storage module, and can also obtain dynamic feature data from the data storage module. The second judgment unit judges the matching of the first dynamic feature data and the dynamic feature data with the dangerous feature standard data according to the feature model to determine its safety. If the judgment result is safe, the data generation unit generates new dynamic feature data. If the judgment result is unsafe, an alarm signal with a warning mark is sent to the second data output unit.

[0030] In an embodiment of the present invention, after the first dynamic monitoring module obtains the first dynamic feature data from the first data storage module, if it decides to continue storing the first dynamic feature data, it continuously monitors the matching status of the first dynamic feature data with the standard data of dangerous features. When an error occurs during the monitoring process, the information acquisition module collects the corresponding alarm information and incorporates the alarm information into the historical feature data. After the second dynamic monitoring module obtains the second dynamic feature data from the second data storage module, if it decides to continue storing the second dynamic feature data, it will continuously monitor the matching of the second dynamic feature data with the standard data of dangerous features. When an error occurs during the monitoring process, the information acquisition module will collect alarm information and store the alarm information as historical feature data.

[0031] The criteria for judging monitoring errors are: continuous The deviation values ​​of all feature matching results exceeded the threshold. ,Right now ; in, Indicates the first The loss value of the matching is used to represent the loss value of the matching. The degree of deviation between the predicted value and the actual value during feature matching; the larger the value, the greater the deviation of the matching. This represents the preset deviation threshold, which is the critical value for judging whether the feature matching deviation is acceptable. Its value is determined based on the safety requirements and historical data of different ship state characteristics, and is used to define the range of normal deviation and abnormal deviation. This represents the number of consecutive monitoring cycles, which is a positive integer. Its value is set according to the system's requirements for monitoring accuracy and response speed. For example, it can be set to 3 times, 5 times, etc. The larger the value, the higher the rigor of error judgment, which can reduce false judgments, but may slightly delay the time of error detection. The judgment is based on the calculation logic of comparing the deviation values ​​of multiple consecutive feature matching with a preset threshold to determine whether an error has occurred in the monitoring process. First, for each feature matching, the deviation value is calculated. Then, the deviation value each time With preset threshold Comparison, when consecutive The results of all comparisons were If an error occurs during the monitoring process, it is determined that an error has occurred. This logic avoids misjudgments caused by single, accidental deviations. Through repeated verification, it improves the accuracy and reliability of monitoring error determination, ensuring that a monitoring error is only identified when the matching deviation of the ship's condition characteristics continuously exceeds a reasonable range.

[0032] In an embodiment of the present invention, after receiving an alarm signal, the first judgment unit judges the safety of matching dynamic feature data with dangerous feature standard data by using the warning label carried by the alarm signal. If it is judged to be safe, the alarm signal corresponding to the warning label is obtained from the first data output unit. If it is judged to be unsafe, the safety of dynamic feature data is judged according to the warning label, and the judgment result is transmitted to the comparison analysis module for cross-validation of feature matching. After receiving the alarm signal, the second judgment unit first determines whether the first dynamic feature data is safe. If it is safe, it directly obtains the alarm signal corresponding to the warning icon from the first judgment unit. If it is not safe, it continues to monitor the dynamic feature data. When continuous monitoring finds that the dynamic feature data is incorrect, it directly sends an alarm signal to the second data output unit, and the judgment result is transmitted to the comparison and analysis module for cross-validation of feature matching.

[0033] When determining whether dynamic feature data is safe, both the first and second judgment units first determine whether the ship's turning angle is within a preset range. If it is within the preset range, the first and second data output units extract dynamic feature data that does not include the turning angle from the data storage module, and then the first and second judgment units determine the safety of this dynamic feature data.

[0034] The formula for determining the safe range of steering angle is: ; in, It represents the real-time turning angle of a ship, which is the turning angle value monitored in real time during the ship's navigation, and directly reflects the ship's current turning status; This indicates the preset minimum safe steering angle, which is a lower limit of the steering angle determined based on factors such as ship type and navigation environment. Steering angles below this value may pose safety hazards. This indicates the preset maximum safe steering angle, which is the upper limit of the steering angle determined based on factors such as ship performance and navigation rules. Steering angles exceeding this value may cause instability or danger to the ship. This formula is used to determine whether a ship's real-time turning angle is within a safe range. Its calculation logic is as follows: [Calculate the ship's real-time turning angle...] With the preset minimum safe steering angle and maximum safe steering angle The comparison is performed when the real-time steering angle is greater than or equal to the minimum safe steering angle and less than or equal to the maximum safe steering angle, thus satisfying the condition. If the steering angle is within a safe range, it is determined that the steering angle is outside the safe range; otherwise, it is determined that the steering angle is outside the safe range. This logic can quickly and intuitively determine whether there is a safety risk in the ship's steering operation, providing a prerequisite for subsequent safety assessments of other dynamic characteristic data, and ensuring that the system prioritizes the safety of steering as a key dynamic characteristic. Determining the safe range of the steering angle provides an important basis for assessing the safety of ship dynamic characteristic data, offering the following technical benefits: First, by prioritizing the safety assessment of the steering angle, a key feature affecting ship navigation stability, potential steering risks can be quickly identified, allowing the system time to issue timely warnings or take countermeasures. Second, when the steering angle is within the safe range, the system can further analyze other dynamic characteristic data, improving the accuracy and comprehensiveness of the overall ship status assessment. Third, pre-setting a clear safe angle range ensures a unified and objective judgment standard, avoiding the subjectivity and uncertainty of human judgment, ensuring consistency between the first and second judgment units during collaborative work, enhancing the reliability of system monitoring, and providing strong protection for safe ship navigation.

[0035] In an embodiment of the present invention, the specific working process of the information acquisition module is as follows: extracting ship status feature data corresponding to each ship status feature of each ship and ship status value of each ship from historical ship safety reports; From the extracted ship condition feature data, ship type feature data, ship size feature data, and ship weight feature data are selected; The ship status value of each ship is used as the initial feature value of the corresponding ship status feature data; The extracted ship status feature data of each ship are sorted, and ship status feature data with a proportion lower than a preset threshold are removed. The preset threshold is determined based on the sample size of historical ship safety reports, and valid ship status feature data are retained. Extract the first feature data corresponding to the ship type feature data, ship size feature data, and ship weight feature data from the valid ship status feature data; Compare the ship state values ​​corresponding to each ship state characteristic data with the ship state values ​​corresponding to the first characteristic data; input the first characteristic data into the normalization model of each hazard characteristic data to obtain the hazard characteristic standard data corresponding to each ship state characteristic data; Based on the initial and limit characteristic values ​​of each ship state characteristic, the second characteristic data of each characteristic data in each ship state characteristic data is obtained; combining the first and second characteristic data of each characteristic data, the difference factor of each characteristic data in each ship state characteristic data is obtained. The formula for calculating the difference factor is: ; in, Indicates the first The difference factor of each feature data is used to represent the magnitude of the difference between the first feature data and the second feature data. The larger the value, the more significant the difference between the two feature data. Indicates the first The first feature data is a feature value extracted from valid ship condition feature data that is related to ship type, size, or weight, and the unit is determined according to the feature type; Indicates the first The second feature data of a feature data is a feature value calculated based on the initial feature value and the limit feature value of the feature, reflecting the theoretical reference value of the feature; This formula is used to calculate the degree of difference of the same feature data across different dimensions. First, determine the... The first feature data of each feature data Second feature data The first feature data is extracted from the effective ship state feature data, and the second feature data is obtained based on the initial feature value and the limiting feature value. Then, the difference between the two is calculated. By taking the absolute value to eliminate the positive and negative effects of the difference, only the magnitude of the difference is retained, ultimately yielding the difference factor. This logic can quantify the degree of deviation of the same feature under different data sources, providing basic data for subsequent calculation of ship condition risk value and ensuring the objectivity of risk assessment; Based on the difference factor, the ship state characteristic data and the corresponding characteristic data in the standard data of dangerous characteristics, the ship state risk value group of each ship state characteristic is obtained. The feature weight of the risk value group is trained to obtain the ship state characteristic distribution corresponding to each ship state characteristic, and the distribution is transmitted to the data preprocessing module. The formula for calculating the ship condition risk value is: ; in, Indicates the first The risk value of each feature is used to measure the magnitude of the risk that the feature poses to ship safety; the higher the value, the higher the risk of the feature. Indicates the first The difference factor of each feature data reflects the difference between the first feature data and the second feature data; Indicates the first The standard data for the hazard characteristics of each feature are baseline values ​​obtained through a normalization model, used to define the safety range of that feature. Indicates the first The feature weights of each feature are set according to the degree of influence of the feature on ship safety. The larger the weight, the more important the feature is in risk assessment. This formula is used to evaluate the first The risk level of each feature data. First, the difference factor. Standard data for hazardous characteristics Divide by the normalized difference ratio to eliminate the influence of differences in different feature dimensions; then divide this ratio by the feature weights. Multiplication and the introduction of weights reflect the differences in importance of different features in ship safety assessment, ultimately yielding a ship condition risk value. This logic, through normalization and weight allocation, makes the risk values ​​of different types of features comparable and can comprehensively reflect the degree of impact of each feature on ship safety. The difference factor formula quantifies the differences of the same feature across different data dimensions, providing raw difference data for risk assessment and ensuring the reliability of the data source. The ship condition risk value formula eliminates the influence of dimensions through normalization and, combined with feature weights, enables a comprehensive comparison of risks of different features, making the risk values ​​more meaningful. The ship condition risk value set formed by the combination of these two formulas lays the foundation for feature weight training and the generation of ship condition feature distributions. This allows subsequent data preprocessing and ship condition feature determination modules to operate based on accurate risk assessment results, improving the accuracy of the entire system in identifying dangerous ship conditions and providing a scientific basis for ship safety early warning.

[0036] In an embodiment of the present invention, the specific workflow of the data preprocessing module is as follows: based on the current navigation data of each ship, extract the current ship feature data of each ship; sort the extracted current ship feature data of each ship to obtain the initial feature value of each ship state feature under the current ship feature data of each ship.

[0037] The initial feature values ​​of each ship's current state feature are compared with the ship state values ​​of each ship's state feature under the corresponding ship feature data, and the normalized difference between the two is calculated. This difference is calculated based on the initial feature values ​​and limit feature values ​​provided by the information acquisition module. The formula for calculating the normalized eigenvalue difference is: ; in, Indicates the first The difference between the features is used to represent the relative deviation between the current initial feature value and the ship's state value. The larger the value, the more significant the deviation. Indicates the current ship's number The initial feature value of each state feature is the feature value extracted from the current ship feature data, and the unit is determined according to the feature type; Indicates the first item under the corresponding ship characteristic data The ship state value of a state feature, that is, the reference state value of that feature; Indicates the first The standard data of the hazard characteristics of each feature are used as the benchmark for normalization to eliminate the influence of dimensions. This formula is used to calculate the current ship's... The normalized difference between the initial eigenvalues ​​of each state feature and the ship's state values ​​is calculated. First, the initial eigenvalues ​​are calculated. Ship condition values The difference is then calculated by taking its absolute value to eliminate positive and negative influences while retaining the magnitude of the difference; finally, this absolute difference is divided by the standard data for the hazard characteristics. This process normalizes the difference, eliminating the influence of different feature units, and ultimately yields the normalized feature difference. This logic can objectively quantify the degree of deviation between the current characteristics and the baseline state, making the differences between different types of characteristics comparable, and providing standardized data for the subsequent calculation of comprehensive navigation safety indicators; The normalized initial feature value of each ship's status feature under the current ship feature data and the calculated normalized difference are input into the ship comprehensive index model. The comprehensive navigation safety index of each ship is calculated by the model and then transmitted to the ship status feature determination module. The formula for calculating the comprehensive navigation safety index is as follows: ; in, This represents the comprehensive navigation safety index, used to measure the overall navigation safety level of a ship; the higher the value, the better the safety condition. This represents the total number of features, a positive integer, reflecting the number of ship condition features involved in the assessment; Indicates the first The initial eigenvalues ​​of each feature are normalized to eliminate the influence of dimensions. Indicates the first The weighting coefficients for the initial values ​​of each feature are set based on the importance of the positive performance of that feature to the security assessment. Indicates the first The weighting coefficients for each feature difference are set according to the impact of the degree of deviation of that feature on the security assessment. This formula is used to comprehensively assess the overall navigational safety of a vessel. First, initial characteristic values ​​are set for each feature. Perform normalization, then multiply by the corresponding weighting coefficient. This yields the positive contribution value of the feature; simultaneously, the normalized difference is... Multiply by weighting factor This yields the negative deduction value for that feature; then, the negative deduction value is subtracted from the positive contribution value to obtain the net contribution of a single feature to the comprehensive index; finally, for all... The comprehensive navigation safety index is obtained by summing the net contributions of each feature. This logic, by integrating the positive performance and deviation of various characteristics, comprehensively reflects the safety status of a ship; the higher the index value, the safer the ship's navigation. The normalized feature difference formula eliminates dimensional differences, allowing direct comparison of the deviations of different features and providing a unified standard for comprehensive assessment. The comprehensive navigation safety index formula, through weight allocation and integration of positive and negative contributions, transforms multiple dispersed feature data into a single comprehensive index, intuitively reflecting the overall safety status of the vessel. The combination of these two approaches represents a leap from single-feature analysis to overall safety assessment, providing precise input for cluster analysis and anomaly identification in the vessel status feature determination module. This enhances the comprehensiveness and scientific rigor of the system's vessel safety status assessment, laying a reliable foundation for subsequent hazard warnings.

[0038] In an embodiment of the present invention, the specific working method of the ship dangerous state early warning module is as follows: normalize the comprehensive navigation safety index of each ship to obtain the ship state feature data of each state feature in the comprehensive navigation safety index of each standard ship; compare the ship state feature data of each state feature in the comprehensive navigation safety index of each ship and the ship state feature data of each state feature in the comprehensive navigation safety index of each standard ship with the state features in the corresponding dangerous feature standard data to obtain the ship state deviation value of each state feature in the comprehensive navigation safety index of each standard ship. The formula for calculating the deviation of a ship's condition is: ; in, Indicates the first The deviation value of each feature reflects the degree of relative deviation between the current ship characteristics and the standard ship characteristics; the larger the value, the more obvious the deviation. Indicates the current ship's number The feature data of each feature are real-time collected ship status parameters, and the unit is determined according to the feature type; Indicates the standard ship number The feature data of each feature, that is, the benchmark reference value of that feature; Indicates the first The standard data of the hazard characteristics of each feature are used as a normalization benchmark to eliminate the influence of dimensions; This formula is used to calculate the current ship's... The degree of deviation between each state characteristic and the corresponding characteristics of a standard ship. First, calculate the current ship characteristic data. Compared with standard ship characteristic data The absolute value of the difference is taken to eliminate positive and negative effects, while retaining the magnitude of the deviation; then the absolute difference is divided by the standard data of the hazard characteristic. This process achieves normalization, eliminates differences in the dimensions of different features, and yields the ship's state deviation value. This logic can accurately quantify the difference between the current ship and the standard state, making the degree of deviation of different features comparable and providing a standardized basis for subsequent safety assessments. The ship state deviation value of each feature data in the comprehensive navigation safety index of each standard ship is input into the ship deviation degree calculation model. This model is trained based on the feature weights output by the ship state feature determination module, and the ship state safety feature weights of each standard ship state feature in the comprehensive navigation safety index of each ship are obtained. The formula for calculating the weights of ship condition safety characteristics is as follows: ; in, Indicates the first The security feature weight of each feature reflects the dynamic importance of that feature in the security assessment. The larger the value, the more significant the impact of that feature on the security assessment. The output of the ship state characteristic determination module represents the first... The basic weights of each feature are obtained by training on historical data and reflect the inherent importance of that feature. This formula is used to calculate the first... The safety feature weights of each feature reflect the dynamic change in the importance of that feature in the safety assessment as the degree of deviation increases. First, the deviation value is added to 1. The denominator is obtained, and its value increases as the deviation increases. Then, 1 is divided by this denominator to obtain a coefficient that decreases as the deviation increases; this coefficient is used to dynamically adjust the base weights. Finally, this coefficient is compared with the base weights output by the ship state characteristic determination module. Multiply to obtain the security feature weights. This logic assigns lower weights to features with greater deviations, preventing abnormally deviating features from excessively affecting the overall evaluation and ensuring the rationality of weight allocation. The weight of the ship's state safety feature is input into the ship safety assessment model, which is trained based on the feature model of the dynamic identification module, to obtain the probability of the ship's state safety feature for each standard ship state in the comprehensive navigation safety index of each ship. The formula for calculating the probability of a ship's state safety characteristics is: ; in, Indicates the first The safety feature probability of each feature ranges from 0 to 1. The closer the value is to 1, the safer the feature is, and the closer it is to 0, the more dangerous it is. This represents the product of the deviation value and the safety feature weight, comprehensively reflecting the risk level of the feature; This represents the natural exponential function, used to convert risk levels into probability values; This formula is used to calculate the first... The safety probability of a feature quantifies the likelihood that the feature is in a safe state. First, calculate the deviation value. With security feature weights The product of these factors comprehensively reflects the degree of deviation and importance of the features; then, taking the negative value of this product and substituting it into the exponential function... Due to the properties of the exponential function, when the product is 0, the probability is 1 (completely safe); as the product increases, the probability gradually approaches 0 (highly dangerous). This logic transforms the deviation and weight of features into an intuitive probability of safety, facilitating comprehensive evaluation. The ship state safety characteristic probability and corresponding ship state deviation value of each standard ship state in the current comprehensive navigation safety index of each ship are input into the ship safety characteristic analysis model to obtain the ship state characteristic deviation index of each standard ship state in the current comprehensive navigation safety index of each ship; by comparing these deviation indices, the current danger state index of each ship is obtained. The formula for calculating the ship's hazardous condition index is: ; in, The Hazard State Index measures the overall risk level of a ship; a higher value indicates a higher risk to the ship. This represents the total number of features, a positive integer, reflecting the number of ship status features involved in the assessment; Indicates the first The deviation coefficient of each feature reflects the degree of danger of that feature; the larger the value, the greater the danger. This formula is used to comprehensively calculate the overall hazardous state index of a ship, quantifying the overall risk level of the ship. First, subtract the probability of the safety feature from 1. Obtain the deviation coefficient of this feature (the larger the value, the more dangerous the feature); then compare this deviation coefficient with the deviation value. Multiplying these values ​​yields the contribution of each individual feature to the overall risk index; finally, multiplying all... The risk index is obtained by summing the contribution values ​​of each feature. This logic, by integrating the degree of danger and the magnitude of deviation of various characteristics, comprehensively reflects the overall risk of a ship; the larger the index value, the more dangerous the ship. The current danger status index of each vessel is compared with a preset danger status threshold. When the danger status index is greater than the preset threshold, a corresponding warning message for the vessel's danger situation is given, and the warning level for that vessel is specified. The specific level classification is as follows: Level 1 warning: When If the situation is deemed to be in minor danger, a warning will be issued to remind the crew to pay attention to changes in the ship's condition. Level 2 warning: When At that time, the situation was determined to be moderately dangerous, a warning was issued, and the crew was required to conduct targeted inspections and adjustments; Level 3 Warning: When When the situation is deemed to be in serious danger, an emergency warning is issued, and the crew is instructed to immediately take emergency measures and prepare to dock for maintenance. in, To preset the threshold for dangerous conditions, , The threshold values ​​for the progressively increasing warning levels are as follows: .

[0039] The embodiments disclosed in this invention are preferred embodiments, but are not limited thereto. Those skilled in the art can easily understand the spirit of this invention based on the above embodiments and make different extensions and variations, but as long as they do not depart from the spirit of this invention, they are all within the protection scope of this invention.

Claims

1. A ship safety early warning system based on feature matching, characterized in that, include: The information acquisition module is used to obtain the ship status characteristic information of each ship in dangerous state from historical ship safety reports as the ship status dangerous characteristic data, and at the same time obtain the ship's static characteristic data, historical characteristic data and dynamic characteristic data. The data preprocessing module is used to normalize the hazard characteristic data acquired by the information acquisition module to obtain standard data of each ship state hazard characteristic. The dynamic identification module includes a first dynamic monitoring module, a second dynamic monitoring module, a first dynamic data storage module, and a second dynamic data storage module. It is used to monitor the dynamic characteristic data of ships through training feature models and output the matching result with the standard data of ship dangerous state characteristics as the judgment result. The ship status characteristic determination module is used to match ship status characteristic data with hazard characteristic standard data, determine ship status risk value groups, train to obtain ship status safety characteristic distribution, and perform cluster analysis on the comprehensive navigation safety index dataset to obtain ship status abnormal characteristic data. The comparison and analysis module is used to compare and analyze the judgment results output by the dynamic identification module and the abnormal ship status feature data output by the ship status feature determination module, and transmit the comparison and analysis results after feature matching cross-validation. The ship danger status early warning module is used to determine the ship danger status index through feature matching based on the comparative analysis results and the current comprehensive navigation safety index dataset of the ship, and to provide early warning of dangerous situations.

2. The ship safety early warning system based on feature matching according to claim 1, characterized in that, Also includes: The main data storage module is used to store static feature data, historical feature data, dynamic feature data, and intermediate data generated during system operation acquired by the information acquisition module. The first dynamic data storage module is used to store the dynamic feature data and judgment results processed by the first dynamic monitoring module. The second dynamic data storage module is used to store the dynamic feature data and judgment results processed by the second dynamic monitoring module.

3. The ship safety early warning system based on feature matching according to claim 2, characterized in that, The static feature data covers the inherent attribute information of the ship, including the ship number used to uniquely identify the ship, the ship length and ship width reflecting the physical dimensions of the ship, the ship deadweight and ship weight reflecting the ship's carrying capacity, the ship type representing the ship type, and the draft and water level information related to the ship's buoyancy. The historical feature data includes navigation feature data generated during the ship's past voyages and detection feature data recorded during the detection process. The navigation feature data includes the ship's voyage distance, voyage time, voyage speed, and steering state and angle corresponding to the steering operation. The detection feature data includes the ship's draft and water level data recorded. The dynamic characteristic data includes the ship's real-time speed and real-time steering information during real-time navigation.

4. The ship safety early warning system based on feature matching according to claim 3, characterized in that, The first dynamic monitoring module consists of a first data processing unit, a first data output unit, and a first judgment unit; The working process of the first data processing unit is as follows: it constructs an initial data sample based on the historical feature data of the ship, uses at least one of the static feature data and dynamic feature data as a training sample, forms a feature model that can be used to identify the ship's state through training, and applies the trained feature model to the ship's state identification work. The first data output unit is responsible for obtaining dynamic feature data from the data storage module and sending the obtained dynamic feature data to the first judgment unit. At the same time, it sends a read signal to the first data storage module. After receiving the read signal, the first data storage module loads the dynamic feature data from the data storage module and transmits it to the first data output unit. The first judgment unit judges the matching of dynamic feature data and dangerous feature standard data according to preset comparison rules to determine the safety of dynamic feature data. If the judgment result is safe, the dynamic feature data is stored in the first data storage module. If the judgment result is unsafe, an alarm signal is issued. The feature model adopts a convolutional neural network model. During the model training process, the difference between the predicted value and the actual value is evaluated by a loss function. When there is a deviation in the prediction result, the relevant parameters in the feature model are adjusted. When there is no deviation in the prediction result, the model parameters are not adjusted.

5. A ship safety early warning system based on feature matching according to claim 3, characterized in that, The second dynamic monitoring module includes a second data processing unit, a second data output unit, a second judgment unit, and a data generation unit; The second dynamic monitoring module works in conjunction with the first dynamic monitoring module to jointly monitor the dynamic characteristics data of the ship; The second data processing unit uses the feature model trained by the first dynamic monitoring module to carry out ship status identification. The second data output unit can obtain the first dynamic feature data from the first data storage module, and can also obtain dynamic feature data from the data storage module. The second judgment unit judges the matching of the first dynamic feature data and the dynamic feature data with the dangerous feature standard data according to the feature model to determine its safety. If the judgment result is safe, the data generation unit generates new dynamic feature data. If the judgment result is unsafe, an alarm signal with a warning mark is sent to the second data output unit.

6. A ship safety early warning system based on feature matching according to claim 5, characterized in that, After the first dynamic monitoring module obtains the first dynamic feature data from the first data storage module, if it decides to continue storing the first dynamic feature data, it will continuously monitor the matching of the first dynamic feature data with the standard data of dangerous features. When an error occurs during the monitoring process, the information acquisition module will collect the corresponding alarm information and include the alarm information in the historical feature data. After the second dynamic monitoring module obtains the second dynamic feature data from the second data storage module, if it decides to continue storing the second dynamic feature data, it will continuously monitor the matching of the second dynamic feature data with the standard data of dangerous features. When an error occurs during the monitoring process, the information acquisition module will collect alarm information and store the alarm information as historical feature data.

7. A ship safety early warning system based on feature matching according to claim 6, characterized in that, After receiving the alarm signal, the first judgment unit judges the safety of matching dynamic feature data with dangerous feature standard data by the warning label carried by the alarm signal. If it is judged to be safe, the alarm signal corresponding to the warning label is obtained from the first data output unit. If it is judged to be unsafe, the safety of dynamic feature data is judged according to the warning label, and the judgment result is transmitted to the comparison and analysis module for cross-validation of feature matching. After receiving the alarm signal, the second judgment unit first judges whether the first dynamic feature data is safe. If it is safe, it directly obtains the alarm signal corresponding to the warning label from the first judgment unit. If it is not safe, it continues to monitor the dynamic feature data. When continuous monitoring finds that the dynamic feature data is wrong, it directly sends an alarm signal to the second data output unit, and the judgment result is transmitted to the comparison analysis module for cross-validation of feature matching. When determining whether dynamic feature data is safe, both the first and second judgment units first determine whether the ship's turning angle is within a preset range. If it is within the preset range, the first and second data output units extract dynamic feature data that does not include the turning angle from the data storage module, and then the first and second judgment units determine the safety of this dynamic feature data.

8. A ship safety early warning system based on feature matching according to claim 1, characterized in that, The specific working process of the information acquisition module is as follows: extract the ship status feature data corresponding to each ship status feature of each ship and the ship status value of each ship from historical ship safety reports; From the extracted ship condition feature data, ship type feature data, ship size feature data, and ship weight feature data are selected; The ship status value of each ship is used as the initial feature value of the corresponding ship status feature data; The extracted ship status feature data of each ship are sorted, and ship status feature data with a proportion lower than a preset threshold are removed. The preset threshold is determined based on the sample size of historical ship safety reports, and valid ship status feature data are retained. Extract the first feature data corresponding to the ship type feature data, ship size feature data, and ship weight feature data from the valid ship status feature data; Compare the ship state values ​​corresponding to each ship state characteristic data with the ship state values ​​corresponding to the first characteristic data; input the first characteristic data into the normalization model of each hazard characteristic data to obtain the hazard characteristic standard data corresponding to each ship state characteristic data; Based on the initial and limit characteristic values ​​of each ship state characteristic, the second characteristic data of each characteristic data in each ship state characteristic data is obtained; combining the first and second characteristic data of each characteristic data, the difference factor of each characteristic data in each ship state characteristic data is obtained. Based on the difference factor, the ship state characteristic data, and the corresponding characteristic data in the standard data of dangerous characteristics, the ship state risk value group of each ship state characteristic is obtained. The feature weights of the risk value group are trained to obtain the ship state characteristic distribution corresponding to each ship state characteristic, and the distribution is transmitted to the data preprocessing module.

9. A ship safety early warning system based on feature matching according to claim 1, characterized in that, The specific workflow of the data preprocessing module is as follows: Based on the current navigation data of each ship, extract the current ship feature data of each ship; sort the extracted current ship feature data of each ship to obtain the initial feature values ​​of each ship state feature under the current ship feature data of each ship. The initial feature values ​​of each ship's current state feature are compared with the ship state values ​​of each ship's state feature under the corresponding ship feature data, and the normalized difference between the two is calculated. This difference is calculated based on the initial feature values ​​and limit feature values ​​provided by the information acquisition module. The normalized initial feature values ​​of each ship's current ship feature data and the calculated normalized difference are input into the ship comprehensive index model. The comprehensive navigation safety index of each ship is calculated by the model and then transmitted to the ship state feature determination module.

10. A ship safety early warning system based on feature matching according to claim 1, characterized in that, The specific working method of the ship dangerous state early warning module is as follows: normalize the comprehensive navigation safety index of each ship to obtain the ship state feature data of each state feature in the comprehensive navigation safety index of each standard ship; compare the ship state feature data of each state feature in the comprehensive navigation safety index of each ship and the ship state feature data of each state feature in the comprehensive navigation safety index of each standard ship with the state features in the corresponding dangerous feature standard data to obtain the ship state deviation value of each state feature in the comprehensive navigation safety index of each standard ship. The ship state deviation value of each feature data in the comprehensive navigation safety index of each standard ship is input into the ship deviation degree calculation model. This model is trained based on the feature weights output by the ship state feature determination module, and the ship state safety feature weights of each standard ship state feature in the comprehensive navigation safety index of each ship are obtained. The weight of the ship's state safety feature is input into the ship safety assessment model, which is trained based on the feature model of the dynamic identification module, to obtain the probability of the ship's state safety feature for each standard ship state in the comprehensive navigation safety index of each ship. By inputting the ship state safety characteristic probability and the corresponding ship state deviation value of each standard ship state in the current comprehensive navigation safety index of each ship into the ship safety characteristic analysis model, the ship state characteristic deviation index of each standard ship state in the current comprehensive navigation safety index of each ship is obtained. By comparing these deviation indices, the current danger status index of each ship can be obtained; The current danger status index of each ship is compared with the preset danger status threshold. When the danger status index is greater than the preset threshold, the corresponding ship danger warning information is given and the danger warning level of the ship is specified.

Images