A method and system for identifying a climbing hazard of a ship's accommodation ladder

Abstract

The application relates to a ship gangway climbing risk identification method and system, which comprises the following steps: when an image is collected, the identity information of a target object is identified, and monitoring data associated with the identity information is acquired in real time; the action trend characteristics of the target object and the state characteristics and environmental characteristics of the target gangway are extracted; the action trend characteristics, the state characteristics, the environmental characteristics and the monitoring data are input into a risk prediction model; when prediction information is received, a corresponding response mechanism is triggered according to the risk level thereof; the application realizes comprehensive and real-time risk identification of the ship gangway climbing process by combining image recognition based on the action trend of the crew, the gangway and the environment, physiological parameter monitoring of the crew and the risk prediction model, improves the accuracy and timeliness of risk identification, can automatically trigger a corresponding response mechanism according to the risk level, and effectively reduces the safety risk in the climbing process, thereby guaranteeing the safety of the crew.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of risk prediction, in particular to a ship gangway climbing risk identification method and system. BACKGROUND

[0002] In marine transportation and port operations, the safety use of ship gangway, as the main access for crew to board and disembark the ship, is of great importance. With the development of intelligent technology, there is an increasing demand for safety monitoring during the process of ship gangway climbing, aiming to identify and prevent potential risks in advance through technical means and protect the personal safety of crew.

[0003] In the prior art, the safety management of ship gangway mainly relies on manual supervision and the execution of safety regulations, but this approach has many limitations. On the one hand, manual monitoring is easily affected by factors such as fatigue and distraction, making it difficult to maintain high alertness continuously. On the other hand, relying solely on experience to judge cannot fully cover the complex and variable operating environment and individual differences in crew state, leading to the difficulty in discovering safety hazards in time. In recent years, although some technologies have attempted to monitor the use of gangways by installing monitoring cameras, most of them can only achieve basic image recording functions and lack the ability to analyze and predict risks in crew behavior, gangway state and surrounding environment.

[0004] Specifically, the defects of the prior art mainly manifest in the following aspects:

[0005] Lack of intelligent identification and analysis: Most monitoring systems can only capture simple images and cannot identify the action trends of crew and the specific state of gangways and environment, making it difficult to predict potential risks;

[0006] Lack of physiological state monitoring: The physical condition of crew, such as heart rate, blood pressure and other physiological parameters, directly affects the safety of climbing the gangway, but the real-time monitoring and integrated analysis of this important factor are often ignored in the prior art. SUMMARY

[0007] To solve the above-mentioned defects, the present application provides a ship gangway climbing risk identification method and system.

[0008] The above-mentioned invention purpose of the present application is achieved by the following technical solutions:

[0009] A ship gangway climbing risk identification method, comprising the steps of:

[0010] When receiving the collected images from the image collection terminal, the identity information of the target object is identified, and the monitoring data of the physiological parameter monitoring device associated with the identity information is obtained in real time;

[0011] extracting action trend features of the target object in the collected image, and state features and environment features of the target ladder;

[0012] inputting the extracted action trend features, state features, environment features, and monitoring data into a pre-trained danger prediction model for danger prediction;

[0013] when receiving the prediction information output by the danger prediction model, triggering a corresponding response mechanism according to the danger level thereof.

[0014] By adopting the technical solutions, the present application combines image recognition technology, physiological parameter monitoring, and a danger prediction model, aiming to provide real-time danger identification and response for the process of climbing a ship ladder: receiving real-time images from an image collection terminal, identifying the identity information of a target object (i.e., a crew member) who is climbing a ladder, associating the identity information of the crew member, obtaining monitoring data transmitted by a physiological parameter monitoring device worn by the crew member in real time, deeply analyzing the collected images, extracting action trend features of the target object, such as climbing speed and body posture change, identifying state features of the ladder and environment features of the surrounding environment, inputting the extracted action trend features, state features of the ladder, environment features, and physiological parameter monitoring data of the crew member into a pre-trained danger prediction model, the model being capable of predicting the possibility and danger level of danger occurring in the current climbing situation based on a large amount of historical data and a machine learning algorithm, when receiving the prediction information output by the danger prediction model, triggering a corresponding response mechanism according to the predicted danger level; the present application combines image recognition based on the action trend of the crew member, the ladder, and the environment, physiological parameter monitoring of the crew member, and a danger prediction model, achieving comprehensive and real-time danger identification for the process of climbing a ship ladder, improving the accuracy and timeliness of danger identification, and automatically triggering a corresponding response mechanism according to the danger level, thereby effectively reducing the safety risk in the climbing process and protecting the life safety of the crew member.

[0015] In a preferred example, the present application can be further configured such that the step of extracting action trend features of the target object in the collected image, and state features and environment features of the target ladder, comprises the steps of:

[0016] identifying body contour features, joint motion features, and speed change features of the target object in the collected image;

[0017] identifying shaking amplitude features and surface state features of the target ladder in the collected image, and identifying obstacle features and weather and light features of the surrounding environment of the target ladder;

[0018] extracting action trend features based on the body contour features, joint motion features, and speed change features of the target object, and the shaking amplitude features of the target ladder;

[0019] extract the state features and environment features of the target gangway based on the surface state features, the shaking amplitude features of the target gangway, and the obstacle features, the weather and light features of the environment around the target gangway.

[0020] By adopting the above technical solutions, the key features in the collected image are extracted: the body contour features of the target object are recognized, the joint movement features of the target object, such as the angle change of the joint, the movement trajectory, etc., and the speed change features of the target object, i.e. the dynamic information such as acceleration and deceleration in the climbing process, are captured and analyzed, and the above features jointly constitute the action trend features of the target object, which provides an important basis for evaluating the stability and safety in the climbing process; the state of the target gangway is analyzed in detail: the shaking amplitude features of the gangway, i.e. the swinging degree of the gangway under the action of external factors such as wind and sea waves, are detected, and at the same time, the surface state features of the gangway, such as whether there are potential safety hazards such as wear, rust, cracks, etc., are paid attention to, and the environment around the gangway is scanned comprehensively to identify possible obstacle features, such as protruding hull structures, floating objects, etc., and weather and light features, such as light intensity, whether it is raining, whether there is haze, etc., which may affect the climbing safety of the crew and the stability of the gangway; based on the above-identified features, the action trend features, the state features of the gangway and the environment features are further extracted, the action trend features integrate the body contour, joint movement and speed change of the crew, which can reflect the action stability and coordination of the crew in the climbing process, while the state features of the gangway and the environment features cover the self-condition of the gangway and the external environmental factors, which provide comprehensive information for evaluating the potential risks in the climbing process; the present application provides a comprehensive and accurate information basis for the dangerous identification in the climbing process of the ship gangway by identifying and analyzing the action trend features of the crew, the state features of the gangway and the environment features, which not only improves the accuracy and efficiency of the dangerous identification, but also can evaluate the safety risk in the climbing process in real time according to the identified features, effectively improving the safety of the crew and the safety operation of the ship.

[0021] In a preferred example, the present application can be further configured as: the dangerous prediction model includes an integration layer, a coupling analysis layer, an adjustment layer and an evaluation layer, and the step of inputting the extracted action trend features, state features, environment features and monitoring data into the pre-trained dangerous prediction model for dangerous prediction includes the steps of:

[0022] The integration layer dynamically integrates the input environment features to obtain a comprehensive environment feature vector;

[0023] The coupling analysis layer performs coupling effect analysis based on the comprehensive environment feature vector, the action trend features, the state features and the monitoring data, and outputs analysis result information;

[0024] The adjustment layer adjusts parameters and structure of the risk prediction model based on the analysis result information;

[0025] The evaluation layer performs risk level evaluation based on the analysis result information, the adjustment result of the adjustment layer, and the pre-set risk level classification standard, and outputs the prediction information.

[0026] By adopting the above technical solutions, the risk prediction model comprises an integration layer, a coupling analysis layer, an adjustment layer, and an evaluation layer. The integration layer receives input environmental features and dynamically integrates them to fuse them into a comprehensive environmental feature vector. The coupling analysis layer receives the comprehensive environmental feature vector, action trend features, state features, and monitoring data, and performs deep coupling effect analysis on the received features to output analysis result information containing correlation information among the features. The adjustment layer dynamically adjusts parameters and structure of the risk prediction model based on the analysis result information output by the coupling analysis layer. The evaluation layer performs risk level evaluation on the climbing situation of the target object based on the analysis result information received from the coupling analysis layer, the adjustment result of the adjustment layer, and the pre-set risk level classification standard, and outputs prediction information. The application realizes comprehensive analysis and dynamic prediction of multiple features in the ship gangway climbing process by constructing a risk prediction model comprising an integration layer, a coupling analysis layer, an adjustment layer, and an evaluation layer. Not only does this improve the accuracy and timeliness of risk prediction, but it also dynamically adjusts the model according to the current environment and crew state, enhancing the adaptability and robustness of the model.

[0027] In a preferred example, the application can be further configured such that the step of adjusting parameters and structure of the risk prediction model based on the analysis result information by the adjustment layer comprises the steps of:

[0028] The adjustment layer evaluates the contribution degrees of the comprehensive environmental feature vector, the action trend features, the state features, and the monitoring data based on the analysis result;

[0029] The adjustment layer assigns attention weights to the comprehensive environmental feature vector, the action trend features, the state features, and the monitoring data based on the evaluation result of the contribution degrees;

[0030] The adjustment layer adjusts parameters and structure of the risk prediction model based on the attention weights.

[0031] By adopting the technical solution, the adjustment layer evaluates the contribution degrees of the comprehensive environmental feature vector, the action trend feature, the state feature, and the monitoring data according to the analysis result information output by the coupling analysis layer, so as to measure the importance and influence of each feature in predicting the danger. Based on the evaluation result of the contribution degrees, the adjustment layer allocates an attention weight to each feature, and the weight reflects the relative importance of the feature in the model decision-making process. The adjustment layer dynamically adjusts the parameters (such as the weights and biases of the neural network) and the structure (such as the number of network layers and the number of neurons in each layer) of the danger prediction model according to the allocated attention weight. By introducing the dynamic adjustment strategy based on the contribution degree evaluation and the attention weight, the danger prediction model can more accurately capture the features that are crucial to the danger prediction, and dynamically adjust the model parameters and structure to adapt to the changes of the features. This not only improves the prediction accuracy of the model, but also enhances the robustness and adaptability of the model, so that the model can better cope with the complex and variable ship ladder climbing environment.

[0032] In a preferred example, the application can be further configured to: the evaluation layer performs danger level evaluation based on the analysis result information, the adjustment result of the adjustment layer, and the pre-set danger level classification standard, and outputs the prediction information, and the step includes the steps of:

[0033] The evaluation layer performs initial danger level evaluation based on the analysis result information, the adjustment result of the adjustment layer, and the pre-set danger level classification standard.

[0034] The evaluation layer obtains the stored historical danger event data and the crew behavior pattern data, and performs similarity calculation on the initial danger level evaluation result.

[0035] The evaluation layer performs final danger level evaluation based on the similarity calculation result, and outputs the prediction information.

[0036] By adopting the technical solution, the evaluation layer comprehensively considers the analysis result information, the adjustment result of the model parameters and structure by the adjustment layer according to the real-time situation, and the pre-set danger level classification standard to perform initial danger level evaluation. The evaluation layer obtains the stored historical danger event data and the crew behavior pattern data from the database. These data contain past danger events and their related features, as well as the behavior reactions of the crew in similar situations. The similarity calculation technology is used to compare the initial danger level evaluation result with these historical data to find the most similar historical data or historical cases. Based on the similarity calculation result, the evaluation layer corrects or confirms the initial danger level evaluation result to obtain the final danger level evaluation result.

[0037] The second application purpose is achieved by the following technical solution:

[0038] A ship gangway climbing danger identification system comprises:

[0039] A data acquisition module is configured to identify identity information of a target object and acquire monitoring data of a physiological parameter monitoring device associated with the identity information in real time when receiving an image collection terminal.

[0040] A feature extraction module is configured to extract action trend features of the target object in the collected image and state features and environmental features of the target gangway.

[0041] An input module is configured to input the extracted action trend features, state features, environmental features and monitoring data into a pre-trained danger prediction model for danger prediction.

[0042] A response module is configured to trigger a corresponding response mechanism according to a danger level of the prediction information output by the danger prediction model when receiving the prediction information.

[0043] According to the above technical scheme, the data acquisition module is configured to identify identity information of a target object and acquire monitoring data of a physiological parameter monitoring device associated with the identity information in real time when receiving an image collection terminal. The feature extraction module is configured to extract action trend features of the target object in the collected image and state features and environmental features of the target gangway. The input module is configured to input the extracted action trend features, state features, environmental features and monitoring data into a pre-trained danger prediction model for danger prediction. The response module is configured to trigger a corresponding response mechanism according to a danger level of the prediction information output by the danger prediction model when receiving the prediction information.

[0044] In a preferred example, the feature extraction module can further comprise:

[0045] A first identification sub-module is configured to identify body contour features, joint movement features and speed change features of the target object in the collected image.

[0046] A second identification sub-module is configured to identify swing amplitude features and surface state features of the target gangway in the collected image and identify obstacle features and weather and light features of the environment around the target gangway.

[0047] A first extraction sub-module is configured to extract action trend features based on the body contour features, joint movement features, speed change features of the target object and the swing amplitude features of the target gangway.

[0048] A second extraction sub-module is configured to extract state features and environmental features of the gangway based on the surface state features and swing amplitude features of the target gangway and the obstacle features and weather and light features of the environment around the target gangway.

[0049] By adopting the technical scheme, the first identification sub-module is configured to identify the body contour feature, joint motion feature and speed change feature of the target object in the collected image; the second identification sub-module is configured to identify the shaking amplitude feature and surface state feature of the target suspension ladder in the collected image, and identify the obstacle feature and weather and light feature of the environment around the target suspension ladder;

[0050] The first extraction sub-module is configured to extract the action trend feature based on the body contour feature, joint motion feature, speed change feature of the target object and the shaking amplitude feature of the target suspension ladder; and the second extraction sub-module is configured to extract the state feature and environmental feature of the suspension ladder based on the surface state feature, shaking amplitude feature of the target suspension ladder and the obstacle feature and weather and light feature of the environment around the target suspension ladder.

[0051] In summary, the present application has at least one of the following beneficial technical effects:

[0052] 1. The present application realizes comprehensive and real-time danger identification for the process of climbing the suspension ladder of the ship by combining the image recognition based on the action trend of the crew, the suspension ladder and the environment, the monitoring of the physiological parameters of the crew and the danger prediction model, which not only improves the accuracy and timeliness of danger identification, but also automatically triggers the corresponding response mechanism according to the danger level, thereby effectively reducing the safety risk in the climbing process and protecting the life safety of the crew;

[0053] 2. The present application provides a comprehensive and accurate information basis for the danger identification in the process of climbing the suspension ladder of the ship by identifying and analyzing the action trend feature of the crew, the state feature of the suspension ladder and the environmental feature, which not only improves the precision and efficiency of danger identification, but also can evaluate the safety risk in the climbing process in real time according to the identified features, effectively improving the protection of the life safety of the crew and the safe operation of the ship;

[0054] 3. The present application realizes comprehensive analysis and dynamic prediction of multiple features in the process of climbing the suspension ladder of the ship by constructing a danger prediction model including an integration layer, a coupling analysis layer, an adjustment layer and an evaluation layer, which not only improves the accuracy and timeliness of danger prediction, but also dynamically adjusts the model according to the current environment and crew state, enhancing the adaptability and robustness of the model;

[0055] 4. The present application introduces a dynamic adjustment strategy based on contribution degree evaluation and attention weight, so that the danger prediction model can more accurately capture the features that are crucial to danger prediction, and dynamically adjust the model parameters and structure to adapt to the changes of these features, which not only improves the prediction accuracy of the model, but also enhances the robustness and adaptability of the model, making it better cope with the complex and variable environment of the suspension ladder climbing of the ship. BRIEF DESCRIPTION OF DRAWINGS

[0056] Figure 1is a flowchart of an embodiment of a ship suspended ladder climbing risk identification method of the present application;

[0057] Figure 2 is an implementation flowchart of step S20 in an embodiment of a ship suspended ladder climbing risk identification method of the present application;

[0058] Figure 3 is an implementation flowchart of step S30 in an embodiment of a ship suspended ladder climbing risk identification method of the present application;

[0059] Figure 4 is an implementation flowchart of step S33 in an embodiment of a ship suspended ladder climbing risk identification method of the present application;

[0060] Figure 5 is an implementation flowchart of step S34 in an embodiment of a ship suspended ladder climbing risk identification method of the present application. DETAILED DESCRIPTION

[0061] The following will be described in detail below with reference to the accompanying drawings Figures 1-5 The present application will be further described in detail.

[0062] In an embodiment, as shown in the figure, the present application discloses a ship suspended ladder climbing risk identification method, specifically comprising the following steps: Figure 1

[0063] S10: When receiving the collected image from the image collection terminal, identify the identity information of the target object, and real-time acquire the monitoring data of the physiological parameter monitoring device associated with the identity information;

[0064] In this embodiment, the image collection terminal is a device for capturing and transmitting real-time video images, including a camera, an infrared thermal imager, etc.; the target object is a person climbing the ladder, specifically a crew member on the ship; the identity information is the pre-stored personal identity basic information of each crew member; the physiological parameter monitoring device is a device for monitoring human physiological indicators (such as heart rate, blood pressure, etc.); the monitoring data is the human physiological indicator data monitored by the physiological parameter monitoring device;

[0065] Specifically, real-time images from the image collection terminal are received, and the identity information of the target object (i.e. the crew member) who is climbing the suspended ladder is identified, the identity information of the crew member is associated, and the monitoring data transmitted by the physiological parameter monitoring device worn by the crew member is real-time acquired;

[0066] S20: Extract the action trend features of the target object in the collected image, and the state features and environmental features of the target suspended ladder;

[0067] ​In the embodiment, the action trend feature is a motion pattern, direction, speed, acceleration and other dynamic attributes exhibited by the target object in the continuous captured images; the target ladder is a specific ladder structure in the captured images, and the state feature thereof includes the inclination, stability, whether there is damage or wear, whether someone is using, and the like; the environment feature is other environmental factors in the captured images except the target object and the target ladder, including weather conditions (sunny, rainy, snowy, and the like), light conditions (bright, dim), layout and types of surrounding objects, and the like;

[0068] Specifically, the captured images are deeply analyzed to extract the action trend feature of the target object, such as the climbing speed, body posture change, and the like, while the state feature of the ladder and the surrounding environment feature are recognized;

[0069] S30: inputting the extracted action trend feature, state feature, environment feature and monitoring data into a pre-trained danger prediction model for danger prediction;

[0070] In the embodiment, the danger prediction model is a model constructed based on historical data and machine learning algorithm, which is used to predict potential danger conditions and their levels;

[0071] Specifically, the extracted action trend feature, ladder state feature, environment feature and physiological parameter monitoring data of the crew are input into the pre-trained danger prediction model, which can predict the possibility and danger level of the danger occurring under the current climbing condition based on a large amount of historical data and machine learning algorithm;

[0072] S40: when receiving the prediction information output by the danger prediction model, triggering a corresponding response mechanism according to the danger level thereof;

[0073] In the embodiment, the prediction information is the electrical signal information containing the danger condition and its level output by the danger prediction model after danger prediction; the danger level is a pre-set level mechanism for quantitatively grading the danger condition; the response mechanism is a coping measure or action plan taken based on the danger level when the prediction information containing the danger condition is recognized;

[0074] Specifically, when receiving the prediction information output by the danger prediction model, the corresponding response mechanism is immediately triggered according to the predicted danger level.

[0075] In an embodiment, as shown in FIG. 2, step S20 includes the following steps: Figure 2

[0076] S21: recognizing the body contour feature, joint motion feature and speed change feature of the target object in the captured images;

[0077] ​S22: identify the shaking amplitude feature and the surface state feature of the target catwalk in the collected image, and identify the obstacle feature and the weather and light feature of the environment around the target catwalk;

[0078] S23: extract the action trend feature based on the body contour feature, the joint motion feature, the speed change feature of the target object, and the shaking amplitude feature of the target catwalk;

[0079] S24: extract the state feature and the environment feature of the catwalk based on the surface state feature and the shaking amplitude feature of the target catwalk, and the obstacle feature and the weather and light feature of the environment around the target catwalk;

[0080] In this embodiment, the body contour feature is the feature of the external shape and contour of the body of the target object in the collected image, including the overall size, shape, proportion of the body, and any significant body curve or protruding part; the joint motion feature is the motion condition of the joints of the target object, including the angle change, motion speed, motion range, etc. of the joints; the speed change feature is the change condition of the speed of the target object when moving in the image, including the increase and decrease of the speed, whether uniform speed, etc.; the shaking amplitude feature is the shaking degree of the target catwalk under the action of wind, user activity, etc., which is quantified by the swing amplitude or frequency of the catwalk; the surface state feature is the condition of the surface of the target catwalk, including whether there is damage, wear, corrosion, stain, etc.; the obstacle feature is the obstacle in the environment around the target catwalk, which may affect the safety of use of the catwalk, such as blocking the line of sight, limiting the space, etc.; the weather and light feature is the weather condition and light condition when the image is collected, such as sunny, cloudy, rainy, snowy, and the strength, direction, etc. of the light; the action trend feature is specifically extracted based on the body contour feature, the joint motion feature, the speed change feature of the target object, and the shaking amplitude feature of the target catwalk, and is used to describe the possible action or behavior trend of the target object in the future; the state feature of the catwalk is specifically extracted based on the surface state feature and the shaking amplitude feature of the target catwalk, and is used to describe the current state (such as safety, stability, etc.) of the catwalk; the environment feature is specifically extracted based on the obstacle feature and the weather and light feature of the environment around the target catwalk, and is used to describe the environmental conditions (such as space limitation, line of sight obstruction, weather influence, etc.) of the catwalk;

[0081] Further, the infrared thermal imager can capture the temperature distribution image of the surface of the catwalk, and by analyzing the change of the temperature distribution, identify whether there is a safety hazard such as wet and slippery, damage, etc. on the surface of the catwalk; for example, a wet and slippery surface may cause uneven temperature distribution, and a damaged part may show an abnormal cold and hot spot;

[0082] Specifically, key features in the collected images are extracted: the body contour features of the target object are identified, the joint movement features such as the angle change of the joints, the movement trajectory, etc. of the target object are captured and analyzed, and the speed change features of the target object, i.e. the dynamic information such as acceleration and deceleration in the climbing process, are analyzed, and the above features jointly constitute the action trend features of the target object, which provides an important basis for evaluating the stability and safety of the target object in the climbing process; the state of the target suspended ladder is analyzed in detail: the swing amplitude features of the suspended ladder, i.e. the swing degree of the suspended ladder under the action of external factors such as wind and sea waves, are detected, and the surface state features of the suspended ladder, such as whether there are potential safety hazards such as wear, rust, cracks, etc. are also concerned, and the surrounding environment of the suspended ladder is scanned comprehensively to identify possible obstacle features such as protruding ship body structures, floating objects, etc. and weather and light features such as light intensity, whether it is raining, whether there is haze, etc. These factors can all affect the climbing safety of the crew and the stability of the suspended ladder; based on the above identified features, the action trend features and the state features of the suspended ladder and the environmental features are further extracted, the action trend features integrate the body contour, joint movement and speed change of the crew, and can reflect the action stability and coordination of the crew in the climbing process, while the state features of the suspended ladder and the environmental features cover the self-condition of the suspended ladder and the external environmental factors, and provide comprehensive information for evaluating the potential risks in the climbing process.

[0083] In an embodiment, the risk prediction model includes an integration layer, a coupling analysis layer, an adjustment layer, and an evaluation layer, as shown in Figure 3 As shown in step S30, it includes steps of:

[0084] S31: The integration layer dynamically integrates the input environmental features to obtain a comprehensive environmental feature vector;

[0085] S32: The coupling analysis layer performs coupling effect analysis based on the comprehensive environmental feature vector, the action trend features, the state features, and the monitoring data, and outputs analysis result information;

[0086] S33: The adjustment layer adjusts the parameters and structure of the risk prediction model based on the analysis result information;

[0087] S34: The evaluation layer performs risk level evaluation based on the analysis result information, the adjustment result of the adjustment layer, and the pre-set risk level division standard, and outputs prediction information;

[0088] In this embodiment, the integration layer is the first layer in the hazard prediction model, used to dynamically integrate the input environmental features (such as weather, light, obstacles, etc.); the comprehensive environmental feature vector is the result output by the integration layer, which is a vector containing all the input environmental feature information, which is integrated by a specific method (such as weighted average, feature fusion, etc.) to further analyze the subsequent layer; the coupling analysis layer is the second layer in the hazard prediction model, which analyzes the coupling effect based on the comprehensive environmental feature vector, action trend feature, state feature and monitoring data, to analyze the interaction and correlation between these features, and their influence on hazard prediction; the analysis result information is the result output by the coupling analysis layer, which contains the analysis results of the coupling effect between the features, including the correlation between the features, the influence degree, the trend, etc., which provides the basis for the adjustment and evaluation of the subsequent layer; the adjustment layer is the third layer in the hazard prediction model, which adjusts the parameters and structure of the model based on the analysis result information output by the coupling analysis layer; the evaluation layer is the last layer in the hazard prediction model, which evaluates the hazard level based on the analysis result information of the coupling analysis layer, the adjustment result of the adjustment layer and the pre-set hazard level classification standard, to output a clear hazard level prediction information, so as to take corresponding response measures; the hazard level classification standard is the standard for the evaluation layer to evaluate the hazard level, which is set according to the specific application scene and demand, usually including multiple hazard levels (such as low, medium, high, etc.), and the corresponding hazard degree and possible consequences of each level; the prediction information is the final result output by the evaluation layer, which contains the prediction of the current situation or the hazard situation in the future period of time and the corresponding hazard level;

[0089] Among them, the coupling analysis layer analyzes the multi-dimensional coupling effect based on the comprehensive environmental feature vector, the action trend feature, the state feature and the monitoring data (such as wind speed, wave height, crew heart rate, etc.), and the analysis process includes but is not limited to:

[0090] Trend risk assessment: combining the physical condition of the crew and the climbing action, i.e. combining the monitoring data and the action trend feature, to assess the risk level of the crew in the current environment, for example, uncoordinated joint movement or abnormal speed change may indicate the risk of the crew being exhausted or losing balance;

[0091] Environmental adaptability analysis: considering the influence of weather, light and catwalk state on climbing safety, i.e. combining the comprehensive environmental feature vector and the state feature, for example, strong wind, heavy rain or slippery catwalk may significantly increase the risk of climbing;

[0092] Obstacle impact assessment: analyze the potential interference of surrounding obstacles on the climbing path and how they interact with the climbing actions of the crew, i.e., combine the comprehensive environmental feature vector, action trend features, and state features to analyze whether there is an increased risk of collision or stumbling;

[0093] Comprehensive analysis combining the above risk assessment, environmental adaptability analysis, and obstacle impact assessment;

[0094] Specifically, the danger prediction model includes an integration layer, a coupling analysis layer, an adjustment layer, and an evaluation layer. The integration layer receives input environmental features and dynamically integrates them to fuse them into a comprehensive environmental feature vector. The coupling analysis layer receives the comprehensive environmental feature vector, as well as action trend features, state features, and monitoring data, and performs deep coupling effect analysis on the received features, outputting analysis result information containing the correlation information between these features. The adjustment layer dynamically adjusts the parameters and structure of the danger prediction model based on the analysis result information output by the coupling analysis layer. The evaluation layer performs danger level evaluation on the current target object's climbing situation based on the analysis result information received from the coupling analysis layer, the adjustment results of the adjustment layer, and the pre-set danger level classification standard, and outputs prediction information.

[0095] In an embodiment, as shown in FIG. 33, step S33 includes the following steps: Figure 4

[0096] S331: The adjustment layer evaluates the contribution of the comprehensive environmental feature vector, action trend features, state features, and monitoring data based on the analysis result;

[0097] S332: The adjustment layer assigns attention weights to the comprehensive environmental feature vector, action trend features, state features, and monitoring data based on the evaluation result of the contribution;

[0098] S333: The adjustment layer adjusts the parameters and structure of the danger prediction model based on the attention weights;

[0099] In this embodiment, the contribution is the importance or impact of each feature (comprehensive environmental feature vector, action trend features, state features, and monitoring data) in the danger prediction model on the prediction result of the model. The attention weight is a weight value assigned according to the contribution of the feature, used for weighted processing of different features in the model. Adjusting the parameters and structure of the danger prediction model specifically means that the adjustment layer adjusts the parameters and structure of the danger prediction model according to the assigned attention weights, including modifying the parameters such as weights and biases of the model, and possibly changing the network structure, number of layers, number of neurons, etc. of the model;

[0100] ​Specifically, the adjustment layer evaluates the contribution degrees of the comprehensive environmental feature vector, the action trend feature, the state feature, and the monitoring data according to the analysis result information output by the coupling analysis layer, so as to measure the importance and influence of each feature in predicting the danger. Based on the evaluation result of the contribution degrees, the adjustment layer assigns an attention weight to each feature, and the weight reflects the relative importance of the feature in the decision-making process of the model. The adjustment layer dynamically adjusts the parameters and structure of the danger prediction model according to the assigned attention weight.

[0101] In an embodiment, as shown in FIG. 4, step S34 includes the following steps: Figure 5

[0102] S341: The evaluation layer performs initial danger level evaluation based on the analysis result information, the adjustment result of the adjustment layer, and the pre-set danger level classification standard;

[0103] S342: The evaluation layer acquires the stored historical danger event data and the crew behavior pattern data, and performs similarity calculation on the initial danger level evaluation result;

[0104] S343: The evaluation layer performs final danger level evaluation based on the similarity calculation result, and outputs the prediction information;

[0105] In the embodiment, the historical danger event data is a data set recording past danger events and related features thereof; the crew behavior pattern data is data reflecting the behavior reaction of the crew in a specific situation; and the similarity calculation is a method for measuring the degree of similarity between two data objects.

[0106] Further, the crew behavior pattern data includes the skill level and experience of the crew for evaluating the risk in the climbing process;

[0107] Specifically, the evaluation layer comprehensively considers the analysis result information (including multi-dimensional data such as environmental features, action trends, and ship states), the adjustment result of the model parameters and structure by the adjustment layer according to real-time conditions, and the pre-set danger level classification standard (such as low risk, medium risk, and high risk), performs initial danger level evaluation, acquires the stored historical danger event data and the crew behavior pattern data from the database, these data include past danger events and related features thereof, and the behavior reaction of the crew in a similar situation, and uses similarity calculation technology (such as cosine similarity and Euclidean distance) to compare the initial danger level evaluation result with these historical data, find out the most similar historical data or historical case to the current situation, and based on the similarity calculation result, the evaluation layer corrects or confirms the initial danger level evaluation result, and obtains the final danger level evaluation result.

[0108] ​It should be understood that the size of the serial number of each step in the above embodiments does not mean the order of execution, and the execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0109] In an embodiment, a ship suspended ladder climbing danger identification system is provided, which corresponds to the ship suspended ladder climbing danger identification method described above. The ship suspended ladder climbing danger identification system comprises:

[0110] The data acquisition module is configured to identify the identity information of the target object when receiving the image collection terminal issuing the collected image, and acquire the monitoring data of the physiological parameter monitoring device associated with the identity information in real time;

[0111] The feature extraction module is configured to extract the action trend feature of the target object in the collected image, and the state feature and the environmental feature of the target suspended ladder;

[0112] The input module is configured to input the extracted action trend feature, state feature, environmental feature and monitoring data into the pre-trained danger prediction model for danger prediction;

[0113] The response module is configured to trigger a corresponding response mechanism according to the danger level when receiving the prediction information output by the danger prediction model;

[0114] Optionally, the feature extraction module comprises:

[0115] The first identification sub-module is configured to identify the body contour feature, joint motion feature and speed change feature of the target object in the collected image;

[0116] The second identification sub-module is configured to identify the shaking amplitude feature and surface state feature of the target suspended ladder in the collected image, and identify the obstacle feature and weather and light feature of the environment around the target suspended ladder;

[0117] The first extraction sub-module is configured to extract the action trend feature based on the body contour feature, joint motion feature, speed change feature of the target object and the shaking amplitude feature of the target suspended ladder;

[0118] The second extraction sub-module is configured to extract the state feature and environmental feature of the suspended ladder based on the surface state feature and shaking amplitude feature of the target suspended ladder, and the obstacle feature and weather and light feature of the environment around the target suspended ladder;

[0119] Optionally, it further comprises:

[0120] The integration layer module is configured to dynamically integrate the input environmental feature to obtain a comprehensive environmental feature vector;

[0121] The coupling analysis layer module is configured to perform coupling effect analysis based on the comprehensive environment feature vector, the action trend feature, the state feature, and the monitoring data, and output analysis result information;

[0122] The adjustment layer module is configured to adjust parameters and structures of the risk prediction model based on the analysis result information;

[0123] The evaluation layer module is configured to perform risk level evaluation based on the analysis result information, adjustment results of the adjustment layer, and pre-set risk level classification standards, and output prediction information;

[0124] Optionally, the adjustment layer module comprises:

[0125] The contribution degree evaluation sub-module is configured to evaluate contribution degrees of the comprehensive environment feature vector, the action trend feature, the state feature, and the monitoring data based on the analysis result;

[0126] The weight distribution sub-module is configured to distribute attention weights for the comprehensive environment feature vector, the action trend feature, the state feature, and the monitoring data based on the evaluation result of the contribution degrees;

[0127] The adjustment sub-module is configured to adjust parameters and structures of the risk prediction model based on the attention weights;

[0128] Optionally, the evaluation layer module comprises:

[0129] The initial evaluation sub-module is configured to perform initial risk level evaluation based on the analysis result information, adjustment results of the adjustment layer, and pre-set risk level classification standards;

[0130] The similarity calculation sub-module is configured to obtain stored historical risk event data and crew behavior pattern data, and perform similarity calculation on the initial risk level evaluation result;

[0131] The final evaluation sub-module is configured to perform final risk level evaluation based on the similarity calculation result, and output prediction information.

[0132] Specific limitations of the ship suspended ladder climbing risk identification system can be seen from the limitations of the ship suspended ladder climbing risk identification method in the foregoing, which will not be repeated here. Each module in the ship suspended ladder climbing risk identification system can be realized by software, hardware, or a combination thereof. Each module can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory in the computer device in software form, so as to be called and executed by the processor to perform the operations corresponding to each module.

[0133] The above-described embodiments are only used to illustrate the technical solutions of the present application, but not limit them; although the present application is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application, and should be included in the protection scope of the present application.

Claims

1. A method for identifying the danger of climbing a ship's suspended ladder, characterized in that: Including the following steps: When an image is received from the image acquisition terminal, the identity information of the target object is identified, and the monitoring data of the physiological parameter monitoring device associated with the identity information is obtained in real time. Extract the motion trend features of the target object, as well as the state and environmental features of the target ladder from the acquired images; The extracted action trend features, state features, environmental features, and monitoring data are input into a pre-trained hazard prediction model to predict the hazard. When the predicted information output by the hazard prediction model is received, the corresponding response mechanism is triggered according to its hazard level. The hazard prediction model includes an integration layer, a coupling analysis layer, an adjustment layer, and an evaluation layer. The step of inputting the extracted action trend features, state features, environmental features, and monitoring data into the pre-trained hazard prediction model to predict the hazard includes the following steps: The integration layer dynamically integrates the input environmental features to obtain a comprehensive environmental feature vector. The coupling analysis layer performs coupling effect analysis based on comprehensive environmental feature vectors, action trend features, state features, and monitoring data, and outputs analysis results. The adjustment layer adjusts the parameters and structure of the hazard prediction model based on the analysis results. The assessment layer assesses the hazard level based on the analysis results, the adjustment results of the adjustment layer, and the pre-set hazard level classification standards, and outputs predictive information. The coupling analysis layer performs multi-dimensional coupling effect analysis based on comprehensive environmental feature vectors, action trend features, state features, and monitoring data, including: Trend risk assessment: By combining monitoring data and behavioral trend characteristics, assess the risk level of crew members in the current environment; Environmental adaptability analysis: Combining comprehensive environmental feature vectors and state characteristics to assess the impact of weather lighting and ladder condition on climbing safety; Obstacle impact assessment: Analyze the potential interference of surrounding obstacles on the climbing path, as well as the interaction between surrounding obstacles and the crew's climbing actions. That is, combine comprehensive environmental feature vectors, action trend characteristics, and state characteristics to analyze whether there is an increased risk of collision or tripping. The step of the assessment layer assessing the hazard level based on the analysis results, the adjustment results of the adjustment layer, and the pre-set hazard level classification standards, and outputting prediction information, includes the following steps: The assessment layer conducts an initial hazard level assessment based on the analysis results, the adjustment results from the adjustment layer, and the pre-set hazard level classification standards. The assessment layer acquires stored historical hazardous event data and crew behavior pattern data, and performs similarity calculations on the initial hazard level assessment results; The assessment layer performs a final hazard level assessment based on the similarity calculation results and outputs prediction information; The data on crew behavior patterns includes crew members' skill levels and experience.

2. The method for identifying the danger of climbing a ship's suspended ladder according to claim 1, characterized in that: The steps for extracting the motion trend features of the target object, the state features of the target ladder, and the environmental features from the acquired images include the following steps: Identify the body contour features, joint motion features, and velocity change features of the target object in the acquired image; Identify the sway amplitude and surface condition features of the target suspension ladder in the acquired images, and identify the obstacle features and weather and lighting features of the surrounding environment of the target suspension ladder; Action trend features are extracted based on the target object's body contour features, joint motion features, velocity change features, and the swaying amplitude features of the target suspension ladder. Based on the surface state characteristics and sway amplitude characteristics of the target suspension ladder, as well as the obstacle characteristics and weather illumination characteristics of the surrounding environment, the state characteristics and environmental characteristics of the suspension ladder are extracted.

3. The method for identifying the danger of ship ladder climbing according to claim 1, characterized in that: The adjustment layer adjusts the parameters and structure of the hazard prediction model based on the analysis results, including the following steps: The adjustment layer evaluates the contribution of comprehensive environmental feature vectors, action trend features, state features, and monitoring data based on the analysis results; The adjustment layer assigns attention weights based on the contribution-based evaluation results, integrating environmental feature vectors, action trend features, state features, and monitoring data. The adjustment layer adjusts the parameters and structure of the hazard prediction model based on attention weights.

4. A ship ladder climbing hazard identification system, used in the steps of the ship ladder climbing hazard identification method as described in any one of claims 1-3, characterized in that: include: The data acquisition module is used to identify the identity information of the target object when it receives the acquired image from the image acquisition terminal, and to acquire the monitoring data of the physiological parameter monitoring device associated with the identity information in real time. The feature extraction module is used to extract the motion trend features of the target object in the acquired image, as well as the state features and environmental features of the target ladder. The input module is used to input the extracted action trend features, state features, environmental features, and monitoring data into the pre-trained hazard prediction model for hazard prediction; The response module is used to trigger the corresponding response mechanism based on the hazard level when it receives the prediction information output by the hazard prediction model.

5. A ship ladder climbing hazard identification system according to claim 4, characterized in that: The feature extraction module includes: The first recognition submodule is used to recognize the body contour features, joint motion features, and velocity change features of the target object in the acquired image; The second recognition submodule is used to identify the sway amplitude and surface condition features of the target suspension ladder in the acquired image, and to identify the obstacle features and weather lighting features of the environment around the target suspension ladder. The first extraction submodule is used to extract motion trend features based on the target object's body contour features, joint motion features, velocity change features, and the swaying amplitude features of the target suspension ladder. The second extraction submodule is used to extract the state features and environmental features of the target suspended ladder based on the surface state features, sway amplitude features, obstacle features, and weather and lighting features of the surrounding environment.

Images