A method and system for maritime safety assessment based on real-time images

Abstract

The present application provides a method and system for marine safety assessment based on real-time images, the method comprising: acquiring real-time image data, real-time operation data and a construction topology map of each engineering equipment; inputting each of the real-time image data and the construction topology map into a preset environmental condition prediction model, so that the environmental condition prediction model predicts the environmental condition of each construction point; inputting the real-time operation data into a preset equipment condition prediction model, so that the equipment condition prediction model predicts the operating condition of each engineering equipment; evaluating the risk level of each construction point according to the environmental condition of each construction point and the operating condition of each engineering equipment, and constructing an initial construction risk topology map in combination with the construction topology map; performing risk transfer on each construction point according to the initial construction risk topology map, obtaining and outputting the construction risk topology map, and improving the accuracy and comprehensiveness of marine safety assessment.

Description

Technical Field

[0001] The present application relates to the field of image processing and machine learning technology, and in particular to a maritime safety assessment method and system based on real-time images. Background Art

[0002] With the development of marine science and technology, the scale of marine engineering projects carried out at sea is expanding. Since the marine environment is more complex and changeable than the land environment, various equipment are faced with more frequent emergencies and risk events in the marine environment. At the same time, due to the limitations of the marine environment, when an emergency occurs in a production area, it is difficult for operation and maintenance personnel to reach the target location in time for repair. Therefore, compared with the land environment, sudden emergencies without warning will have a greater impact on normal marine activities. Therefore, it is necessary to establish a complete marine risk monitoring and assessment system to timely predict possible risks and deal with them accordingly.

[0003] Maritime safety assessment refers to the systematic assessment and analysis of risks that may arise in maritime activities, determining their potential impact on maritime safety, and ensuring the safety and sustainability of maritime activities. Most existing maritime safety assessment methods are based on real-time monitoring of equipment operating data and marine environmental data, and timely detection of abnormal data and alarms through analysis of monitoring data. However, this method only takes into account the current status of the equipment and the environment, and cannot predict possible risk events in the future. Even if the safety assessment system can detect abnormal conditions of offshore equipment, due to the large size of the ocean area and the restrictions on sea operations due to weather conditions, operation and maintenance personnel cannot guarantee that they can quickly arrive at the site for repairs.

[0004] Benefiting from the development of big data and machine learning technology, the existing safety assessment system can predict the future safety status of each device by establishing a machine learning model. However, in the process of safety assessment, the information contained in the device's own operating data is often not comprehensive enough to help the model accurately predict the future risk level of the device. Summary of the invention

[0005] In response to the above technical problems, the present application provides a maritime safety assessment method and system based on real-time images to improve the accuracy and comprehensiveness of maritime safety assessment.

[0006] In a first aspect, the present application provides a method for maritime safety assessment based on real-time images, comprising:

[0007] Obtain real-time image data and real-time operation data of various engineering equipment in marine engineering;

[0008] Obtaining a construction topology map of the marine engineering project, wherein the construction topology map is constructed based on big data technology and includes location data and environmental data of each construction point and topological relationships between each construction point;

[0009] Input each of the real-time image data and the construction topology map into a preset environmental condition prediction model, so that the environmental condition prediction model predicts the environmental condition of each construction point, wherein the environmental condition prediction model is constructed based on a graph convolutional network model, a two-dimensional feature extraction network, and a three-dimensional feature extraction network;

[0010] Inputting the real-time operation data into a preset equipment condition prediction model so that the equipment condition prediction model predicts the operation status of each engineering equipment, wherein the equipment condition prediction model is constructed based on a machine learning model;

[0011] Assess the risk level of each construction site based on the environmental conditions of each construction site and the operating conditions of each engineering equipment;

[0012] According to the risk level of each construction point, combined with the construction topology map, an initial construction risk topology map is constructed;

[0013] According to the inherent risk level of each construction point in the initial construction risk topology map and the topological relationship between each construction point, risk is transferred to each construction point, and a construction risk topology map including comprehensive safety assessment results of each construction point is obtained and output.

[0014] The embodiment of the present application provides a method for marine safety assessment based on real-time images, which predicts the environmental conditions of each construction point based on the real-time image data of each device. At the same time, considering that in the marine environment, there is a large linkage between the environmental conditions of each construction point, and the change of environmental conditions will not occur independently at a certain point, but will radiate in the form of a surface. Therefore, the embodiment of the present application combines the construction topology of the marine engineering with the real-time image data of each device for prediction, thereby improving the accuracy of the environmental condition prediction. Further, the environmental conditions of each construction point are not the only factor affecting the safety of marine engineering, and the operating conditions of each engineering equipment in the marine engineering are also important for the safety of marine engineering. Therefore, while predicting the environmental conditions, the embodiment of the present application predicts the operating conditions of each engineering equipment according to the real-time operating data of each engineering equipment, and comprehensively evaluates the risk level of each construction point through the operating conditions and environmental conditions, thereby improving the accuracy and comprehensiveness of the marine safety assessment. Finally, the embodiment of the present application introduces the concept of risk transfer, that is, the risk level of each construction point will affect and transfer each other, and the risk is transferred based on the topological relationship between each construction point. After risk transfer based on the initial construction risk topology map, a construction risk topology map containing the comprehensive safety assessment results of each construction point can be obtained, completing the safety assessment of each offshore construction point and improving the accuracy and comprehensiveness of offshore safety assessment.

[0015] Furthermore, the construction topology map is obtained based on big data technology, including:

[0016] Based on GIS technology, we can obtain the location data of each construction point and the topological relationship between them;

[0017] According to the location data of each construction point, environmental data of each construction point is obtained based on big data technology, wherein the environmental data includes temperature, wind direction, wind force level, light intensity and tidal data of the location of the construction point;

[0018] The construction topology map is constructed according to the location data of each construction point, the topological relationship between each construction point and the environmental data of each construction point.

[0019] The embodiment of the present application provides a method for constructing a construction topology map, firstly, the location data and topological relationship of each construction point are obtained based on GIS technology. Since the amount of data involved in the environmental data of the ocean is very large, the embodiment of the present application considers using big data technology to collect environmental data, and in the data collection process, the location data of each construction point can assist big data technology in data positioning and data screening, improve the efficiency and accuracy of environmental data collection, and thus improve the accuracy of marine safety assessment.

[0020] In a possible implementation, the environmental condition prediction model predicts the environmental condition of each construction site, including:

[0021] Using a preset encoder-decoder structure network based on a self-attention mechanism to extract two-dimensional features of the construction topology map;

[0022] Using a preset long short-term memory convolutional network to respectively extract the three-dimensional features of each of the real-time image data;

[0023] Constructing a feature matrix and an adjacency matrix according to the two-dimensional features, the three-dimensional features and the construction topology map, wherein the feature matrix is ​​used to store each of the two-dimensional features and each of the three-dimensional features, and the adjacency matrix is ​​used to store the topological relationship between each of the construction points;

[0024] The feature matrix and the adjacency matrix are jointly input into a preset graph convolutional network model, so that the graph convolutional network model performs feature fusion on the two-dimensional features and the three-dimensional features, and then predicts the environmental conditions of each construction point based on the topological relationship between the construction points.

[0025] The embodiment of the present application provides a method for predicting the environmental conditions of each construction site. Since the embodiment of the present application predicts by combining the construction topology map with the real-time image data of each device, and the construction topology map belongs to two-dimensional data, and the real-time image data belongs to three-dimensional data, it is necessary to perform different feature extraction on the construction topology map and the real-time image data respectively. In the embodiment of the present application, an encoder-decoder structure network is used to extract two-dimensional features, and a long short-term memory convolutional network is used to extract three-dimensional features, wherein the self-attention mechanism in the encoder-decoder structure network can effectively capture the dependencies between different parts of the input data, so that the network can accurately extract the two-dimensional features in the construction topology map. The long short-term memory convolutional network can effectively capture and understand the complex dependencies in long sequences, avoid the influence of a large number of parameters, and use the long short-term memory convolutional network to extract the three-dimensional features of real-time image data, which can effectively improve the efficiency of three-dimensional feature extraction. Finally, the embodiment of the present application uses a graph convolutional network model to perform feature fusion processing on each three-dimensional feature and two-dimensional feature. The graph convolutional network model is good at processing graph or network type data, and the real-time image data and construction topology map obtained in the embodiment of the present application are graph type data or grid type data. Therefore, the embodiment of the present application uses a graph convolutional network model to perform the final prediction to improve the accuracy of environmental condition prediction.

[0026] In a possible implementation, the step of obtaining the equipment condition prediction model based on a machine learning model includes:

[0027] Construct an initial equipment condition prediction model based on the support vector machine model;

[0028] Obtain the historical operation data and operation and maintenance records of each device;

[0029] Manually annotate the historical operation data of each device based on the operation and maintenance records of each device to construct a training data set;

[0030] The initial equipment condition prediction model is supervisedly learned using the training data set, and the model parameters in the initial equipment condition prediction model are updated through a preset interval loss function to obtain the equipment condition prediction model.

[0031] In a possible implementation, when risk transfer is performed on any first construction point among the construction points according to the risk level of each construction point in the initial construction risk topology map and the topological relationship between the construction points, risk transfer is performed on each construction point according to the risk level of each construction point in the initial construction risk topology map and the topological relationship between the construction points, including:

[0032] Determining a number of second construction points for risk transfer to the first construction point according to the initial construction risk topology map and a preset number of risk transfer layers;

[0033] Calculate the transmission risk level of each second construction point according to the risk level of each second construction point, the number of layers between each second construction point and the first construction point, and the relative distance between each second construction point and the first construction point;

[0034] By combining the transferred risk level of each second construction point and the inherent risk level of the first construction point, a comprehensive safety assessment result of the first construction point is calculated to complete the risk transfer to the first construction point.

[0035] The embodiment of the present application provides a method for risk transfer based on topological relationships. Taking the first construction point as an example, before risk transfer, several second construction points around the first construction point for risk transfer to the first construction point are determined according to the number of risk transfer layers, wherein the risk transfer range of each construction point can be controlled by the number of risk transfer layers to avoid unlimited risk transfer and ensure the rationality of marine safety assessment. This parameter can be flexibly set according to the complexity of the connection between each construction point in the actual production process and the actual scale of the marine project. Then, in the process of risk transfer, the embodiment of the present application further considers the influence of the relative distance between each second construction point and the first construction point on risk transfer, and calculates the transfer risk level of each second construction point. Finally, according to the risk level of the first construction point itself and the transfer risk level of each second construction point, the comprehensive safety assessment result of the first construction point is calculated, which realizes the multi-dimensional safety assessment of the marine project and improves the accuracy and comprehensiveness of the marine safety assessment.

[0036] Furthermore, the transfer risk level of each second construction point is calculated based on the risk level of each second construction point, the number of layers between each second construction point and the first construction point, and the relative distance between each second construction point and the first construction point. The specific formula is:

[0037]

[0038] in, represents the transfer risk level of the k-th second construction point, represents the risk level of the k-th second construction point, Indicates the preset number of risk transfer layers, represents the number of interval construction points on the shortest path between the kth second construction point and the first construction point, D represents the Dijkstra algorithm, represents the shortest distance from the kth second construction point to the first construction point j, N represents the number of second construction points, represents the shortest distance from the i-th second construction point to the first construction point j;

[0039] The comprehensive safety assessment result of the first construction point is calculated by combining the transmission risk level of each second construction point and the risk level of the first construction point itself. The specific formula is:

[0040]

[0041] in, is the comprehensive safety assessment result of the first construction point j, is the transfer risk level of the i-th second construction point, is the risk level of the first construction site j, is the preset risk transfer weight coefficient, It is the preset risk weight factor.

[0042] In the second aspect, accordingly, the present application provides a marine safety assessment system based on real-time images, including a first acquisition module, a second acquisition module, an environmental condition prediction module, an equipment condition prediction module, a self-risk assessment module, a risk topology map construction module and a safety assessment module;

[0043] Wherein, the first acquisition module is used to acquire real-time image data and real-time operation data of each engineering equipment in the marine engineering;

[0044] The second acquisition module is used to acquire a construction topology map of the marine engineering project, wherein the construction topology map is constructed based on big data technology and includes location data and environmental data of each construction point and a topological relationship between each construction point;

[0045] The environmental condition prediction module is used to input each of the real-time image data and the construction topology map into a preset environmental condition prediction model, so that the environmental condition prediction model predicts the environmental condition of each construction point, wherein the environmental condition prediction model is constructed based on a graph convolutional network model, a two-dimensional feature extraction network, and a three-dimensional feature extraction network;

[0046] The equipment condition prediction module is used to input the real-time operation data into a preset equipment condition prediction model so that the equipment condition prediction model predicts the operation status of each engineering equipment, wherein the equipment condition prediction model is constructed based on a machine learning model;

[0047] The self-risk assessment module is used to assess the self-risk level of each construction site according to the environmental conditions of each construction site and the operating conditions of each engineering equipment;

[0048] The risk topology map construction module is used to construct an initial construction risk topology map based on the risk level of each construction point and in combination with the construction topology map;

[0049] The safety assessment module is used to transfer risks to each construction point according to the risk level of each construction point in the initial construction risk topology map and the topological relationship between each construction point, and obtain and output a construction risk topology map containing comprehensive safety assessment results of each construction point.

[0050] Furthermore, the construction topology map is obtained based on big data technology, including:

[0051] Based on GIS technology, we can obtain the location data of each construction point and the topological relationship between them;

[0052] According to the location data of each construction point, environmental data of each construction point is obtained based on big data technology, wherein the environmental data includes temperature, wind direction, wind force level, light intensity and tidal data of the location of the construction point;

[0053] The construction topology map is constructed according to the location data of each construction point, the topological relationship between each construction point and the environmental data of each construction point.

[0054] In a possible implementation, the environmental condition prediction model predicts the environmental condition of each construction site, including:

[0055] Using a preset encoder-decoder structure network based on a self-attention mechanism to extract two-dimensional features of the construction topology map;

[0056] Using a preset long short-term memory convolutional network to respectively extract the three-dimensional features of each of the real-time image data;

[0057] Constructing a feature matrix and an adjacency matrix according to the two-dimensional features, the three-dimensional features and the construction topology map, wherein the feature matrix is ​​used to store each of the two-dimensional features and each of the three-dimensional features, and the adjacency matrix is ​​used to store the topological relationship between each of the construction points;

[0058] The feature matrix and the adjacency matrix are jointly input into a preset graph convolutional network model, so that the graph convolutional network model performs feature fusion on the two-dimensional features and the three-dimensional features, and then predicts the environmental conditions of each construction point based on the topological relationship between the construction points.

[0059] Further, when risk transfer is performed on any first construction point among the construction points according to the risk level of each construction point in the initial construction risk topology map and the topological relationship between the construction points, the safety assessment module performs risk transfer on each construction point according to the risk level of each construction point in the initial construction risk topology map and the topological relationship between the construction points, including:

[0060] Determining a number of second construction points for risk transfer to the first construction point according to the initial construction risk topology map and a preset number of risk transfer layers;

[0061] Calculate the transmission risk level of each second construction point according to the risk level of each second construction point, the number of layers between each second construction point and the first construction point, and the relative distance between each second construction point and the first construction point;

[0062] By combining the transferred risk level of each second construction point and the inherent risk level of the first construction point, a comprehensive safety assessment result of the first construction point is calculated to complete the risk transfer to the first construction point. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] Figure 1 : A flow chart of a maritime safety assessment method based on real-time images provided in an embodiment of the present application.

[0064] Figure 2 : A schematic diagram of a flow chart for predicting the environmental conditions of each construction site in a marine safety assessment method based on real-time images provided in an embodiment of the present application.

[0065] Figure 3 : A partial initial construction risk topology map in a marine safety assessment method based on real-time imaging provided in an embodiment of the present application.

[0066] Figure 4 : A schematic diagram of the risk transfer process in a maritime safety assessment method based on real-time images provided in an embodiment of the present application.

[0067] Figure 5 : A structural diagram of a maritime safety assessment system based on real-time imaging provided in an embodiment of the present application. DETAILED DESCRIPTION

[0068] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0069] It should be noted that the step numbers in the text are only for the convenience of explanation of the specific embodiments and do not serve to limit the order in which the steps are executed. In the description of this application, the terms "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features.

[0070] Embodiment 1:

[0071] like Figure 1 As shown, embodiment 1 provides a marine safety assessment method based on real-time images, including steps S1-S7:

[0072] Step S1, obtaining real-time image data and real-time operation data of each engineering equipment in the marine engineering;

[0073] Step S2: obtaining a construction topology map of the marine engineering project, wherein the construction topology map is constructed based on big data technology and includes location data and environmental data of each construction point and topological relationships between the construction points;

[0074] Step S3, inputting each of the real-time image data and the construction topology map into a preset environmental condition prediction model, so that the environmental condition prediction model predicts the environmental condition of each construction point, wherein the environmental condition prediction model is constructed based on a graph convolutional network model, a two-dimensional feature extraction network, and a three-dimensional feature extraction network;

[0075] Step S4, inputting the real-time operation data into a preset equipment condition prediction model, so that the equipment condition prediction model predicts the operation condition of each engineering equipment, wherein the equipment condition prediction model is constructed based on a machine learning model;

[0076] Step S5, evaluating the risk level of each construction site according to the environmental conditions of each construction site and the operating conditions of each engineering equipment;

[0077] Step S6: construct an initial construction risk topology map based on the risk level of each construction point and the construction topology map;

[0078] Step S7: Based on the risk level of each construction point in the initial construction risk topology map and the topological relationship between each construction point, risk is transferred to each construction point to obtain and output a construction risk topology map containing comprehensive safety assessment results of each construction point.

[0079] The embodiment of the present application provides a method for marine safety assessment based on real-time images, which predicts the environmental conditions of each construction point based on the real-time image data of each device. At the same time, considering that in the marine environment, there is a large linkage between the environmental conditions of each construction point, and the change of environmental conditions will not occur independently at a certain point, but will radiate in the form of a surface. Therefore, the embodiment of the present application combines the construction topology of the marine engineering with the real-time image data of each device for prediction, thereby improving the accuracy of the environmental condition prediction. Further, the environmental conditions of each construction point are not the only factor affecting the safety of marine engineering, and the operating conditions of each engineering equipment in the marine engineering are also important for the safety of marine engineering. Therefore, while predicting the environmental conditions, the embodiment of the present application predicts the operating conditions of each engineering equipment according to the real-time operating data of each engineering equipment, and comprehensively evaluates the risk level of each construction point through the operating conditions and environmental conditions, thereby improving the accuracy and comprehensiveness of the marine safety assessment. Finally, the embodiment of the present application introduces the concept of risk transfer, that is, the risk level of each construction point will affect and transfer each other, and the risk is transferred based on the topological relationship between each construction point. After risk transfer based on the initial construction risk topology map, a construction risk topology map containing the comprehensive safety assessment results of each construction point can be obtained, completing the safety assessment of each offshore construction point and improving the accuracy and comprehensiveness of offshore safety assessment.

[0080] In a preferred embodiment, in step S1, a camera device is pre-installed in each engineering equipment in the marine engineering, and the real-time image data is obtained through each camera device. The real-time operation data of each engineering equipment can be collected and obtained using the original data monitoring system in the marine engineering, without the need for additional monitoring equipment, which can effectively reduce the deployment cost of the embodiment of the present application.

[0081] Furthermore, in step S2, the construction topology map is obtained based on big data technology, including:

[0082] Based on GIS technology, we can obtain the location data of each construction point and the topological relationship between them;

[0083] According to the location data of each construction point, environmental data of each construction point is obtained based on big data technology, wherein the environmental data includes temperature, wind direction, wind force level, light intensity and tidal data of the location of the construction point;

[0084] The construction topology map is constructed according to the location data of each construction point, the topological relationship between each construction point and the environmental data of each construction point.

[0085] The embodiment of the present application provides a method for constructing a construction topology map, firstly, the location data and topological relationship of each construction point are obtained based on GIS technology. Since the amount of data involved in the environmental data of the ocean is very large, the embodiment of the present application considers using big data technology to collect environmental data, and in the data collection process, the location data of each construction point can assist big data technology in data positioning and data screening, improve the efficiency and accuracy of environmental data collection, and thus improve the accuracy of marine safety assessment.

[0086] In a possible implementation, in step S3, the environmental condition prediction model predicts the environmental conditions of each construction site, such as Figure 2 As shown, steps S301-S304 are included:

[0087] S301, using a preset encoder-decoder structure network based on a self-attention mechanism to extract two-dimensional features of the construction topology map;

[0088] S302, using a preset long short-term memory convolutional network to respectively extract the three-dimensional features of each of the real-time image data;

[0089] S303, constructing a feature matrix and an adjacency matrix according to the two-dimensional features, the three-dimensional features and the construction topology map, wherein the feature matrix is ​​used to store each of the two-dimensional features and each of the three-dimensional features, and the adjacency matrix is ​​used to store the topological relationship between the various construction points;

[0090] S304. The feature matrix and the adjacency matrix are jointly input into a preset graph convolutional network model, so that the graph convolutional network model performs feature fusion on the two-dimensional features and the three-dimensional features, and then predicts the environmental conditions of each construction point based on the topological relationship between the construction points.

[0091] The embodiment of the present application provides a method for predicting the environmental conditions of each construction site. Since the embodiment of the present application predicts by combining the construction topology map with the real-time image data of each device, and the construction topology map belongs to two-dimensional data, and the real-time image data belongs to three-dimensional data, it is necessary to perform different feature extraction on the construction topology map and the real-time image data respectively. In the embodiment of the present application, an encoder-decoder structure network is used to extract two-dimensional features, and a long short-term memory convolutional network is used to extract three-dimensional features, wherein the self-attention mechanism in the encoder-decoder structure network can effectively capture the dependencies between different parts of the input data, so that the network can accurately extract the two-dimensional features in the construction topology map. The long short-term memory convolutional network can effectively capture and understand the complex dependencies in long sequences, avoid the influence of a large number of parameters, and use the long short-term memory convolutional network to extract the three-dimensional features of real-time image data, which can effectively improve the efficiency of three-dimensional feature extraction. Finally, the embodiment of the present application uses a graph convolutional network model to perform feature fusion processing on each three-dimensional feature and two-dimensional feature. The graph convolutional network model is good at processing graph or network type data, and the real-time image data and construction topology map obtained in the embodiment of the present application are graph type data or grid type data. Therefore, the embodiment of the present application uses a graph convolutional network model to perform the final prediction to improve the accuracy of environmental condition prediction.

[0092] In a possible implementation, the step of obtaining the equipment condition prediction model based on the machine learning model includes:

[0093] Construct an initial equipment condition prediction model based on the support vector machine model;

[0094] Obtain the historical operation data and operation and maintenance records of each device;

[0095] Manually annotate the historical operation data of each device based on the operation and maintenance records of each device to construct a training data set;

[0096] The initial equipment condition prediction model is supervisedly learned using the training data set, and the model parameters in the initial equipment condition prediction model are updated through a preset interval loss function to obtain the equipment condition prediction model.

[0097] In a possible implementation, in step S7, when risk transfer is performed on any first construction point among the construction points according to the risk level of each construction point in the initial construction risk topology map and the topological relationship between the construction points, risk transfer is performed on each construction point according to the risk level of each construction point in the initial construction risk topology map and the topological relationship between the construction points, including:

[0098] Determining a number of second construction points for risk transfer to the first construction point according to the initial construction risk topology map and a preset number of risk transfer layers;

[0099] Calculate the transmission risk level of each second construction point according to the risk level of each second construction point, the number of layers between each second construction point and the first construction point, and the relative distance between each second construction point and the first construction point;

[0100] By combining the transferred risk level of each second construction point and the inherent risk level of the first construction point, a comprehensive safety assessment result of the first construction point is calculated to complete the risk transfer to the first construction point.

[0101] The embodiment of the present application provides a method for risk transfer based on topological relationships. Taking the first construction point as an example, before risk transfer, several second construction points around the first construction point for risk transfer to the first construction point are determined according to the number of risk transfer layers, wherein the risk transfer range of each construction point can be controlled by the number of risk transfer layers to avoid unlimited risk transfer and ensure the rationality of marine safety assessment. This parameter can be flexibly set according to the complexity of the connection between each construction point in the actual production process and the actual scale of the marine project. Then, in the process of risk transfer, the embodiment of the present application further considers the influence of the relative distance between each second construction point and the first construction point on risk transfer, and calculates the transfer risk level of each second construction point. Finally, according to the risk level of the first construction point itself and the transfer risk level of each second construction point, the comprehensive safety assessment result of the first construction point is calculated, which realizes the multi-dimensional safety assessment of the marine project and improves the accuracy and comprehensiveness of the marine safety assessment.

[0102] Among them, the meaning of the number of risk transfer layers is the maximum number of risk transfer layers in the shortest path between the first construction point and the second construction point minus 1 construction point. For example, when the number of risk transfer layers is 3, the maximum allowed interval construction points is 2. The shortest path from construction point A to construction point B is (A, C, D, B). There are interval construction points C and D on the path, which meets the requirements of the number of risk transfer layers. In this case, the risk of A can be transferred to B.

[0103] Furthermore, the transfer risk level of each second construction point is calculated based on the risk level of each second construction point, the number of layers between each second construction point and the first construction point, and the relative distance between each second construction point and the first construction point. The specific formula is:

[0104]

[0105] in, represents the transfer risk level of the k-th second construction point, represents the risk level of the k-th second construction point, Indicates the preset number of risk transfer layers, represents the number of interval construction points on the shortest path between the kth second construction point and the first construction point, D represents the Dijkstra algorithm, represents the shortest distance from the kth second construction point to the first construction point j, N represents the number of second construction points, represents the shortest distance from the i-th second construction point to the first construction point j;

[0106] The comprehensive safety assessment result of the first construction point is calculated by combining the transmission risk level of each second construction point and the risk level of the first construction point itself. The specific formula is:

[0107]

[0108] in, is the comprehensive safety assessment result of the first construction point j, is the transfer risk level of the i-th second construction point, is the risk level of the first construction site j, is the preset risk transfer weight coefficient, It is the preset risk weight factor.

[0109] In a preferred embodiment, part of the initial construction risk topology map is as follows: Figure 3 As shown in the figure, it includes construction points A, B, C, D and their corresponding risk levels. The number of risk transfer layers is set to 2, that is, there is at most one construction point in the shortest path between the first construction point and the second construction point. The risk transfer weight coefficient and the risk weight coefficient are both set to 0.5. Take A as the first construction point to show the risk transfer process, and transfer the risk levels of B, C, and D to A. At this time, it is necessary to calculate the transfer risk levels of B, C, and D respectively according to the above formula, and finally calculate the comprehensive safety assessment result of A to be 4.375. The risk transfer process is as follows: Figure 4 shown.

[0110] In a preferred embodiment, based on the initial construction risk topology map and the construction risk topology map, a dynamic display of risk transfer can be performed on a designated screen, so that users can more intuitively discover the risk changes of each construction point. At the same time, based on the numerical comparison of the risk level of each construction point, the transfer risk level and the comprehensive safety assessment results, a risk transfer analysis is performed to determine the construction points that are prone to risk transfer. When these construction points have a higher level of their own risk level, they often have a greater impact on other construction points. Therefore, these construction points are taken as the focus of attention, so that operation and maintenance engineers give priority to solving the safety risks of these construction points and improve the safety and stability of marine engineering operations.

[0111] Embodiment 2:

[0112] like Figure 5 As shown, embodiment 2 provides a marine safety assessment system based on real-time images, including a first acquisition module 10, a second acquisition module 20, an environmental condition prediction module 30, an equipment condition prediction module 40, a self-risk assessment module 50, a risk topology map construction module 60 and a safety assessment module 70;

[0113] Wherein, the first acquisition module 10 is used to acquire real-time image data and real-time operation data of each engineering equipment in the marine engineering;

[0114] The second acquisition module 20 is used to acquire a construction topology map of the marine engineering project, wherein the construction topology map is constructed based on big data technology and includes location data of each construction point, environmental data, and topological relationships between each construction point;

[0115] The environmental condition prediction module 30 is used to input each of the real-time image data and the construction topology map into a preset environmental condition prediction model, so that the environmental condition prediction model predicts the environmental condition of each construction point, wherein the environmental condition prediction model is constructed based on a graph convolutional network model, a two-dimensional feature extraction network, and a three-dimensional feature extraction network;

[0116] The equipment condition prediction module 40 is used to input the real-time operation data into a preset equipment condition prediction model so that the equipment condition prediction model predicts the operation status of each engineering equipment, wherein the equipment condition prediction model is constructed based on a machine learning model;

[0117] The self-risk assessment module 50 is used to assess the self-risk level of each construction site according to the environmental conditions of each construction site and the operating conditions of each engineering equipment;

[0118] The risk topology map construction module 60 is used to construct an initial construction risk topology map according to the risk level of each construction point and in combination with the construction topology map;

[0119] The safety assessment module 70 is used to transfer the risk of each construction point according to the risk level of each construction point and the topological relationship between each construction point in the initial construction risk topology map, and obtain and output a construction risk topology map containing comprehensive safety assessment results of each construction point.

[0120] Furthermore, the construction topology map is obtained based on big data technology, including:

[0121] Based on GIS technology, we can obtain the location data of each construction point and the topological relationship between them;

[0122] According to the location data of each construction point, environmental data of each construction point is obtained based on big data technology, wherein the environmental data includes temperature, wind direction, wind force level, light intensity and tidal data of the location of the construction point;

[0123] The construction topology map is constructed according to the location data of each construction point, the topological relationship between each construction point and the environmental data of each construction point.

[0124] In a possible implementation, the environmental condition prediction model predicts the environmental conditions of each construction site, including:

[0125] Using a preset encoder-decoder structure network based on a self-attention mechanism to extract two-dimensional features of the construction topology map;

[0126] Using a preset long short-term memory convolutional network to respectively extract the three-dimensional features of each of the real-time image data;

[0127] Constructing a feature matrix and an adjacency matrix according to the two-dimensional features, the three-dimensional features and the construction topology map, wherein the feature matrix is ​​used to store each of the two-dimensional features and each of the three-dimensional features, and the adjacency matrix is ​​used to store the topological relationship between each of the construction points;

[0128] The feature matrix and the adjacency matrix are jointly input into a preset graph convolutional network model, so that the graph convolutional network model performs feature fusion on the two-dimensional features and the three-dimensional features, and then predicts the environmental conditions of each construction point based on the topological relationship between the construction points.

[0129] In a possible implementation, the step of obtaining the equipment condition prediction model based on the machine learning model includes:

[0130] Construct an initial equipment condition prediction model based on the support vector machine model;

[0131] Obtain the historical operation data and operation and maintenance records of each device;

[0132] Manually annotate the historical operation data of each device based on the operation and maintenance records of each device to construct a training data set;

[0133] The initial equipment condition prediction model is supervisedly learned using the training data set, and the model parameters in the initial equipment condition prediction model are updated through a preset interval loss function to obtain the equipment condition prediction model.

[0134] Further, when risk transfer is performed on any first construction point among the construction points according to the risk level of each construction point in the initial construction risk topology map and the topological relationship between the construction points, the safety assessment module performs risk transfer on each construction point according to the risk level of each construction point in the initial construction risk topology map and the topological relationship between the construction points, including:

[0135] Determining a number of second construction points for risk transfer to the first construction point according to the initial construction risk topology map and a preset number of risk transfer layers;

[0136] Calculate the transmission risk level of each second construction point according to the risk level of each second construction point, the number of layers between each second construction point and the first construction point, and the relative distance between each second construction point and the first construction point;

[0137] By combining the transferred risk level of each second construction point and the inherent risk level of the first construction point, a comprehensive safety assessment result of the first construction point is calculated to complete the risk transfer to the first construction point.

[0138] Furthermore, the transfer risk level of each second construction point is calculated based on the risk level of each second construction point, the number of layers between each second construction point and the first construction point, and the relative distance between each second construction point and the first construction point. The specific formula is:

[0139]

[0140] in, represents the transfer risk level of the k-th second construction point, represents the risk level of the k-th second construction point, Indicates the preset number of risk transfer layers, represents the number of interval construction points on the shortest path between the kth second construction point and the first construction point, D represents the Dijkstra algorithm, represents the shortest distance from the kth second construction point to the first construction point j, N represents the number of second construction points, represents the shortest distance from the i-th second construction point to the first construction point j;

[0141] The comprehensive safety assessment result of the first construction point is calculated by combining the transmission risk level of each second construction point and the risk level of the first construction point itself. The specific formula is:

[0142]

[0143] in, is the comprehensive safety assessment result of the first construction point j, is the transfer risk level of the i-th second construction point, is the risk level of the first construction site j, is the preset risk transfer weight coefficient, It is the preset risk weight factor.

[0144] The embodiment of the present application provides a marine safety assessment system based on real-time images, which predicts the environmental conditions of each construction point based on the real-time image data of each device. At the same time, considering that in the marine environment, there is a large linkage between the environmental conditions of each construction point, and the change of environmental conditions will not occur independently at a certain point, but radiate in the form of a surface. Therefore, the embodiment of the present application combines the construction topology of the marine engineering with the real-time image data of each device for prediction, thereby improving the accuracy of the environmental condition prediction. Further, the environmental conditions of each construction point are not the only factor affecting the safety of marine engineering, and the operating conditions of each engineering equipment in the marine engineering are also important for the safety of marine engineering. Therefore, while predicting the environmental conditions, the embodiment of the present application predicts the operating conditions of each engineering equipment according to the real-time operating data of each engineering equipment, and comprehensively evaluates the risk level of each construction point through the operating conditions and environmental conditions, thereby improving the accuracy and comprehensiveness of the marine safety assessment. Finally, the embodiment of the present application introduces the concept of risk transfer, that is, the risk level of each construction point will affect and transfer each other, and the risk is transferred based on the topological relationship between each construction point. After risk transfer based on the initial construction risk topology map, a construction risk topology map containing the comprehensive safety assessment results of each construction point can be obtained, completing the safety assessment of each offshore construction point and improving the accuracy and comprehensiveness of offshore safety assessment.

[0145] The more detailed working principle and step flow of this embodiment can refer to, but are not limited to, the relevant records of Embodiment 1.

[0146] The specific embodiments described above further describe the purpose, technical solutions and beneficial effects of the present application in detail. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the scope of protection of the present application. It is particularly pointed out that for those skilled in the art, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of protection of the present application.

Claims

1. A marine safety assessment method based on real-time images, characterized in that: include: Obtain real-time image data and real-time operation data of various engineering equipment in marine engineering; Obtaining a construction topology map of the marine engineering project, wherein the construction topology map is constructed based on big data technology and includes location data and environmental data of each construction point and topological relationships between each construction point; Input each of the real-time image data and the construction topology map into a preset environmental condition prediction model, so that the environmental condition prediction model predicts the environmental condition of each construction point, wherein the environmental condition prediction model is constructed based on a graph convolutional network model, a two-dimensional feature extraction network, and a three-dimensional feature extraction network; Inputting the real-time operation data into a preset equipment condition prediction model so that the equipment condition prediction model predicts the operation status of each engineering equipment, wherein the equipment condition prediction model is constructed based on a machine learning model; Assess the risk level of each construction site based on the environmental conditions of each construction site and the operating conditions of each engineering equipment; According to the risk level of each construction point, combined with the construction topology map, an initial construction risk topology map is constructed; According to the inherent risk level of each construction point in the initial construction risk topology map and the topological relationship between each construction point, risk is transferred to each construction point, and a construction risk topology map including comprehensive safety assessment results of each construction point is obtained and output.

2. A marine safety assessment method based on real-time images as claimed in claim 1, characterized in that: The construction topology map is obtained based on big data technology, including: Based on GIS technology, we can obtain the location data of each construction point and the topological relationship between them; According to the location data of each construction point, environmental data of each construction point is obtained based on big data technology, wherein the environmental data includes temperature, wind direction, wind force level, light intensity and tidal data of the location of the construction point; The construction topology map is constructed according to the location data of each construction point, the topological relationship between each construction point and the environmental data of each construction point.

3. A marine safety assessment method based on real-time images as claimed in claim 1, characterized in that: The environmental condition prediction model predicts the environmental conditions of each construction site, including: Using a preset encoder-decoder structure network based on a self-attention mechanism to extract two-dimensional features of the construction topology map; Using a preset long short-term memory convolutional network to respectively extract the three-dimensional features of each of the real-time image data; Constructing a feature matrix and an adjacency matrix according to the two-dimensional features, the three-dimensional features and the construction topology map, wherein the feature matrix is ​​used to store each of the two-dimensional features and each of the three-dimensional features, and the adjacency matrix is ​​used to store the topological relationship between each of the construction points; The feature matrix and the adjacency matrix are jointly input into a preset graph convolutional network model, so that the graph convolutional network model performs feature fusion on the two-dimensional features and the three-dimensional features, and then predicts the environmental conditions of each construction point based on the topological relationship between the construction points.

4. A marine safety assessment method based on real-time images as claimed in claim 1, characterized in that: The device condition prediction model is obtained based on the machine learning model, including: Construct an initial equipment condition prediction model based on the support vector machine model; Obtain the historical operation data and operation and maintenance records of each device; Manually annotating the historical operation data of each device based on the operation and maintenance records of each device to construct a training data set; The initial equipment condition prediction model is supervisedly learned using the training data set, and the model parameters in the initial equipment condition prediction model are updated through a preset interval loss function to obtain the equipment condition prediction model.

5. A method for marine safety assessment based on real-time images as claimed in claim 1, characterized in that: When risk transfer is performed on any first construction point among the construction points according to the risk level of each construction point in the initial construction risk topology map and the topological relationship between the construction points, risk transfer is performed on each construction point according to the risk level of each construction point in the initial construction risk topology map and the topological relationship between the construction points, including: Determining a number of second construction points for risk transfer to the first construction point according to the initial construction risk topology map and a preset number of risk transfer layers; Calculate the transmission risk level of each second construction point according to the risk level of each second construction point, the number of layers between each second construction point and the first construction point, and the relative distance between each second construction point and the first construction point; By combining the transferred risk level of each second construction point and the inherent risk level of the first construction point, a comprehensive safety assessment result of the first construction point is calculated to complete the risk transfer to the first construction point.

6. A marine safety assessment method based on real-time images as claimed in claim 5, characterized in that: The transfer risk level of each second construction point is calculated based on the risk level of each second construction point, the number of layers between each second construction point and the first construction point, and the relative distance between each second construction point and the first construction point. The specific formula is: ； in, represents the transfer risk level of the k-th second construction point, represents the risk level of the k-th second construction point, Indicates the preset number of risk transfer layers, represents the number of interval construction points on the shortest path between the kth second construction point and the first construction point, D represents the Dijkstra algorithm, represents the shortest distance from the kth second construction point to the first construction point j, N represents the number of second construction points, represents the shortest distance from the i-th second construction point to the first construction point j; The comprehensive safety assessment result of the first construction point is calculated by combining the transmission risk level of each second construction point and the risk level of the first construction point itself. The specific formula is: ； in, is the comprehensive safety assessment result of the first construction point j, is the transfer risk level of the i-th second construction point, is the risk level of the first construction site j, is the preset risk transfer weight coefficient, It is the preset risk weight factor.

7. A marine safety assessment system based on real-time images, characterized in that: It includes a first acquisition module, a second acquisition module, an environmental condition prediction module, an equipment condition prediction module, a self-risk assessment module, a risk topology map construction module and a safety assessment module; Wherein, the first acquisition module is used to acquire real-time image data and real-time operation data of each engineering equipment in the marine engineering; The second acquisition module is used to acquire a construction topology map of the marine engineering project, wherein the construction topology map is constructed based on big data technology and includes location data and environmental data of each construction point and a topological relationship between each construction point; The environmental condition prediction module is used to input each of the real-time image data and the construction topology map into a preset environmental condition prediction model, so that the environmental condition prediction model predicts the environmental condition of each construction point, wherein the environmental condition prediction model is constructed based on a graph convolutional network model, a two-dimensional feature extraction network, and a three-dimensional feature extraction network; The equipment condition prediction module is used to input the real-time operation data into a preset equipment condition prediction model so that the equipment condition prediction model predicts the operation status of each engineering equipment, wherein the equipment condition prediction model is constructed based on a machine learning model; The self-risk assessment module is used to assess the self-risk level of each construction site according to the environmental conditions of each construction site and the operating conditions of each engineering equipment; The risk topology map construction module is used to construct an initial construction risk topology map based on the risk level of each construction point and in combination with the construction topology map; The safety assessment module is used to transfer risks to each construction point according to the risk level of each construction point in the initial construction risk topology map and the topological relationship between each construction point, and obtain and output a construction risk topology map containing comprehensive safety assessment results of each construction point.

8. A marine safety assessment system based on real-time images as claimed in claim 7, characterized in that: The construction topology map is obtained based on big data technology, including: Based on GIS technology, we can obtain the location data of each construction point and the topological relationship between them; According to the location data of each construction point, environmental data of each construction point is obtained based on big data technology, wherein the environmental data includes temperature, wind direction, wind force level, light intensity and tidal data of the location of the construction point; The construction topology map is constructed according to the location data of each construction point, the topological relationship between each construction point and the environmental data of each construction point.

9. A marine safety assessment system based on real-time images as claimed in claim 7, characterized in that: The environmental condition prediction model predicts the environmental conditions of each construction site, including: Using a preset encoder-decoder structure network based on a self-attention mechanism to extract two-dimensional features of the construction topology map; Using a preset long short-term memory convolutional network to respectively extract the three-dimensional features of each of the real-time image data; Constructing a feature matrix and an adjacency matrix according to the two-dimensional features, the three-dimensional features and the construction topology map, wherein the feature matrix is ​​used to store each of the two-dimensional features and each of the three-dimensional features, and the adjacency matrix is ​​used to store the topological relationship between each of the construction points; The feature matrix and the adjacency matrix are jointly input into a preset graph convolutional network model, so that the graph convolutional network model performs feature fusion on the two-dimensional features and the three-dimensional features, and then predicts the environmental conditions of each construction point based on the topological relationship between the construction points.

10. A marine safety assessment system based on real-time images as claimed in claim 9, characterized in that: When risk transfer is performed on any first construction point among the construction points according to the risk level of each construction point in the initial construction risk topology map and the topological relationship between the construction points, the safety assessment module performs risk transfer on each construction point according to the risk level of each construction point in the initial construction risk topology map and the topological relationship between the construction points, including: Determining a number of second construction points for risk transfer to the first construction point according to the initial construction risk topology map and a preset number of risk transfer layers; Calculate the transmission risk level of each second construction point according to the risk level of each second construction point, the number of layers between each second construction point and the first construction point, and the relative distance between each second construction point and the first construction point; By combining the transferred risk level of each second construction point and the inherent risk level of the first construction point, a comprehensive safety assessment result of the first construction point is calculated to complete the risk transfer to the first construction point.

Images