CIM platform-based hill collapse early warning method and system

Abstract

The invention relates to a collapse early warning method and system based on a CIM platform, and the method comprises the steps: collecting historical data of collapse influence factors, transmitting the historical data to the CIM platform, and carrying out the preprocessing and significant influence feature selection of the historical data; constructing an early warning network, and training the early warning network by using the historical data after preprocessing and significant influence feature selection; and after real-time data of the collapse influence factors are collected and subjected to preprocessing and significant influence feature selection, grade parameters of the corresponding risk areas are output through the trained early warning network, and the CIM platform gives out collapse risk early warning according to the grade parameters. According to the method, the CIM platform is effectively used for processing data, the early warning network is constructed to predict the collapse risk area and send out collapse risk early warning, an automatic early warning mechanism is formed, a large number of monitoring personnel such as videos and broadcasts are not needed, and the labor cost is effectively reduced; and real-time early warning is realized by collecting and processing data in real time.

Description

Technical Field

[0001] The present application relates to the field of ridge collapse warning technology, and specifically to a ridge collapse warning method and system based on a CIM platform. Background Art

[0002] Collapse erosion refers to a special erosional landform formed by the action of water and gravity in the hilly areas of southern China. It consists of five parts: water-collecting slope, collapse wall, collapsing body, ditch and alluvial fan. There are many factors affecting collapse erosion. Human activities, topography, climate, water system and vegetation provide the driving force for the initiation and development of collapse gullies. The development process of collapse gullies can be divided into the formation stage of gully erosion or other landforms developing into collapse gullies, and the development stage of collapse gullies with a certain scale, shape and erosion mechanism continuing to complete the entire life cycle. Collapse erosion causes great harm, cutting the surface, breaking mountains, destroying roads, farmland and houses, silting up lakes, and seriously endangering people's lives and property.

[0003] Although the theoretical research on gully collapse erosion has achieved fruitful results, its formation and development mechanism is complex and changeable, and traditional prediction and display methods are still insufficient. At present, the early warning management of gully collapse in most communities, parks, and scenic spots in China is still in its initial stage, with only simple means such as video monitoring, broadcast notifications, and on-site processing by staff. The level of intelligence is low, the labor cost is high, and it takes a long time; the site equipment and service management are still in a decentralized state, the lack of interconnection between terminals, and the difficulty in sharing data. The regional control center has not formed a dynamic control of "people, objects, events, land, and vehicles", resulting in the early warning being conveyed in a timely manner when facing disasters; the lack of early warning mechanism and the ability to intervene quickly makes it impossible to conduct emergency prevention drills. Summary of the invention

[0004] The present application provides a landslide early warning method and system based on a CIM platform, which can solve the problems of high labor costs, unrealistic early warnings, and lack of early warning mechanisms in current landslide early warning technologies.

[0005] To achieve the above objectives, in a first aspect, the present application provides a landslide early warning method based on a CIM platform, the method comprising:

[0006] The historical data of factors affecting landslides are collected and transmitted to the CIM platform, and the historical data are preprocessed and features with significant impact are selected.

[0007] Construct an early warning network and use historical data after preprocessing and significant impact feature selection to train the early warning network.

[0008] After collecting real-time data on factors affecting landslides and preprocessing and selecting significant influencing features, the trained early warning network outputs the level parameters of the corresponding risk area, and the CIM platform issues a landslide risk warning based on the level parameters.

[0009] Furthermore, in one embodiment, the data collected on factors affecting ridge collapse include data on vegetation density, topography, land use, and data on climate, population density, and geological soil.

[0010] The data on vegetation density, topography, and land use are collected using remote sensing image technology; the data on climate, population density, and geological soil are monitored and collected in real time using IoT sensing devices.

[0011] Furthermore, in one embodiment, historical data and real-time data are preprocessed, including correcting the data by deleting duplicate data, filling missing values ​​in the data, identifying and processing outliers, and integrating relevant data from different data sources, and labeling the corrected data with risk level parameters.

[0012] Furthermore, in one embodiment, the significantly affecting feature selection includes:

[0013] Get the original features of the data after preprocessing operations and create corresponding shadow features.

[0014] The random forest model is trained using the original features and shadow features, and the significantly influential features are output.

[0015] Furthermore, in one embodiment,

[0016] The construction of the early warning network, using the historical data after preprocessing and significant influence feature selection to train the early warning network, includes:

[0017] Use machine learning algorithms to build an early warning network, set the number of nodes in the input layer, hidden layer, and output layer, initialize weights and biases, and find the optimal weights and biases through optimization algorithms.

[0018] Initialize the early warning network using the number of nodes, optimal weights and biases.

[0019] The historical data after preprocessing and significant influencing feature selection are divided into training set, validation set and test set, and the initialized early warning network is trained, parameter adjusted and generalization performance tested respectively.

[0020] Furthermore, in one embodiment, the CIM platform receives the level parameters of the risk area corresponding to the output of the early warning network, renders the risk area into the three-dimensional model scene generated by the CIM platform, and displays the risk area by level while issuing a landslide risk warning.

[0021] Furthermore, in one embodiment, the CIM platform displays risk areas of different levels using different colors in the three-dimensional model scene.

[0022] Furthermore, in one embodiment, the CIM platform presets safe routes based on risk areas displayed in the three-dimensional model scene.

[0023] Generate an emergency plan based on safe routes and predicted risk areas, and simulate the emergency response process when landslides occur based on the emergency plan.

[0024] Furthermore, in one embodiment, after the CIM platform generates the emergency plan, it monitors the changes in sections of the safe route in real time and automatically adjusts the emergency plan according to the changes in sections.

[0025] In a second aspect, based on the above-mentioned landslide warning method based on the CIM platform, the present application provides a landslide warning system of the landslide warning method based on the CIM platform, the system comprising:

[0026] The data preparation module is used to collect historical data of factors affecting landslides and transmit them to the CIM platform, and to preprocess the historical data and select features with significant impact.

[0027] A network construction module is used to construct an early warning network.

[0028] The network training module is used to train the early warning network using historical data after preprocessing and significant influence feature selection.

[0029] The prediction module is used to collect real-time data on factors affecting landslides and, after preprocessing and selection of significantly affecting features, output the level parameters of the corresponding risk area through a trained early warning network. The CIM platform issues a landslide risk warning based on the level parameters.

[0030] The beneficial effects brought by the technical solution provided in the embodiments of the present application include:

[0031] This application collects historical data of factors affecting hillock collapse and transmits it to the CIM platform, preprocesses the historical data and selects significant influencing features, builds an early warning network, uses the preprocessed and significantly influencing feature selection historical data to train the early warning network, collects real-time data of factors affecting hillock collapse and after preprocessing and significantly influencing feature selection, outputs the level parameters of the corresponding risk area through the trained early warning network, and the CIM platform issues a hillock collapse risk warning based on the level parameters. By using the CIM platform to process data, build an early warning network to predict hillock collapse risk areas and issue hillock collapse risk warnings, an automated early warning mechanism is formed, and a large number of video and broadcast monitoring personnel are not required, which effectively reduces labor costs; real-time early warnings are achieved through real-time data collection and processing.

[0032] In addition, when issuing a landslide risk warning, the CIM platform uses different colors to display risk areas of different levels in a three-dimensional model scene, and can preset safe routes based on the displayed risk areas to provide safety guarantees for transportation. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 This is a flow chart of a method for early warning of a landslide based on a CIM platform according to an embodiment of the present application.

[0034] Figure 2 This is a block diagram of a landslide warning system based on a CIM platform according to an embodiment of the present application.

[0035] Figure 3 This is a block diagram of the architecture of the landslide warning system based on the CIM platform in an embodiment of the present application. DETAILED DESCRIPTION

[0036] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings 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.

[0037] This application protects a landslide warning method based on the CIM (City Information Modeling) platform. As a new thing developed based on the new generation of information technology, CIM officially entered the construction stage in China in 2018 with the emergence of pilot cities. The proposal of the national new infrastructure strategy in 2020 promoted the development of my country's CIM industry. In 2021, with the introduction of relevant policies and standards represented by the "Technical Guidelines for the Basic Platform of Urban Information Modeling (CIM)", while regulating the development of the CIM industry, it also ushered in a blowout period. From the application point of view, the disaster warning and emergency response technology of CIM is still in the research stage, and as a whole, a mature route method, unified standards and requirements have not yet been formed.

[0038] CIM technology is a comprehensive technology based on digital city data. It is based on three-dimensional GIS (Geographic Information System), building information model, Internet of Things and big data. By integrating data from different systems, it creates a comprehensive information model of the city. This model accurately restores the physical entities of the city in digital form, including buildings, infrastructure and natural environment, providing a comprehensive and visual perspective for urban planning, management and decision-making. CIM technology supports data integration, visualization and spatial analysis, becoming the foundation of smart city construction and providing scientific and reliable basic data and analysis tools for urban applications in various fields.

[0039] The application of CIM platform in disaster early warning and emergency management is to implement CIM technology in the construction and operation of various places such as parks, scenic spots, and communities, coordinate the application of regional overall systems, implement regional data interconnection with data as the core, and use components of different granularities in the region as carriers to complete the "checkable, manageable, rehearsal, and early warning" of disaster events in the region. It can intuitively and realistically display the formation and evolution process of collapsing ridges in three dimensions, so that regional managers and government departments can more accurately predict the disaster process of collapsing ridges, so as to take effective measures in advance and minimize the losses of the masses.

[0040] In order to make the objectives, technical solutions and advantages of the present application more clear, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0041] In a first aspect, an embodiment of the present application provides a landslide warning method based on a CIM platform.

[0042] In one embodiment, see Figure 1 As shown, the above-mentioned landslide warning method includes:

[0043] S1. Collect historical data of factors affecting landslides and transmit them to the CIM platform, and perform preprocessing and significant influencing feature selection on the historical data.

[0044] S2. Build an early warning network and use historical data after preprocessing and significant impact feature selection to train the early warning network.

[0045] S3. Real-time data of factors affecting landslides are collected and preprocessed and features with significant impact are selected. The level parameters of the corresponding risk areas are output through the trained early warning network. The CIM platform issues landslide risk warnings based on the level parameters.

[0046] This application uses machine learning, big data, CIM three-dimensional modeling and other technologies to collect data on factors affecting ridge collapse and transmit it to the CIM platform. By using the CIM platform to process data and build an early warning network to predict ridge collapse risk areas and issue ridge collapse risk warnings, an automated early warning mechanism is formed, and there is no need for a large number of video and broadcast monitoring personnel, which effectively reduces labor costs. Through real-time collection and processing of data, real-time warnings are achieved to provide safety guarantees for transportation.

[0047] Furthermore, in one embodiment, in the above step S1, collecting data on factors affecting ridge collapse includes: using remote sensing image technology to collect data on vegetation density, topography, and land use, and using IoT sensing devices to monitor and collect data on climate, population density, and geological soil in real time, wherein the IoT sensing devices include meteorological monitoring sensors, traffic sensors, and soil moisture sensors, and the data on factors affecting ridge collapse include historical data on factors affecting ridge collapse and real-time data on factors affecting ridge collapse.

[0048] Furthermore, in one embodiment, the data is corrected by deleting duplicate data, filling missing values ​​in the data, identifying and processing outliers, and integrating relevant data from different data sources, and the corrected data is labeled with risk level parameters to achieve the process of preprocessing historical data and real-time data. The above-mentioned correction of data can produce the following effects:

[0049] By deleting duplicate data, we can ensure the uniqueness of the data and avoid the adverse effects of redundant data on subsequent model results; use the mean or median to fill in missing values ​​in the data to ensure data integrity; identify outliers through methods such as Z-score detection and delete or correct them to ensure the rationality of the data; integrate related data sets from different data sources to ensure the consistency of timestamps and geographic locations.

[0050] In this embodiment, the real-time nature, integrity and security of the data are ensured by preprocessing the historical data and the real-time data.

[0051] The risk levels in the above embodiment can be divided into: high risk, medium risk and low risk, corresponding to the high risk area, medium risk area and low risk area respectively.

[0052] Furthermore, in one embodiment, in the above step S1, the historical data of factors affecting landslides are subjected to significant influence feature selection, including:

[0053] S11. Obtain the original features of the historical data of the landslide influencing factors after preprocessing, and create corresponding shadow features in a randomly disrupted order for each original feature.

[0054] S12. Input the original features and shadow features into the random forest model together and output the importance score of each feature.

[0055] S13. Compare the importance score of the original feature with the importance score of the shadow feature, and apply FDR (False Discovery Rate) correction to perform statistical tests to determine the significantly influential features.

[0056] S14. Through an iterative process, features with no significant impact are removed to determine the final feature subset, which contains significant impact features including rainfall, population density, rainfall erosivity, soil moisture, elevation, area-elevation integral, temperature, lithology and soil texture.

[0057] In this embodiment, a subset of significantly influential features is determined through significantly influential feature selection for use in subsequent model construction, which can improve the prediction performance of the model.

[0058] Furthermore, in one embodiment, in the above step S2, constructing an early warning network, and using historical data after preprocessing and significant influence feature selection to train the early warning network, includes:

[0059] S21. Divide the historical data after preprocessing and significant influencing feature selection into training set, validation set and test set in a ratio of seven to three. The training set is used for model training and optimization, the validation set is used to adjust the hyperparameters of the model, and the test set is used to evaluate the generalization ability of the model.

[0060] S22. Use machine learning algorithm to build an early warning network, set the number of nodes in the input layer, hidden layer and output layer, initialize weights and biases. Since BP (Back Propagation) neural network has powerful nonlinear mapping capabilities, the machine learning algorithm of this embodiment selects BP neural network.

[0061] S23. Find the optimal weights and biases through optimization algorithm.

[0062] S24. Initialize the BP neural network using the number of nodes, optimal weights and biases, and perform forward propagation, error calculation and back propagation on the training set to optimize network performance.

[0063] S25. Combined with the selected significant influencing features, the initialized early warning network is trained. The validation set is used to adjust the number of hidden layer nodes and learning rate parameters of the model. The test set is used to calculate generalization performance indicators such as accuracy and recall rate to verify the performance and effect of the model and complete the generalization performance test of the early warning network.

[0064] In this embodiment, the optimization algorithm in step S23 selects PSO (Particle Swarm Optimization) to optimize the weights and biases of the BP neural network, which specifically includes the following steps:

[0065] S231, initialize the particle swarm, each particle represents a set of weights and biases of the BP neural network.

[0066] S232. Set the parameters of PSO, including particle swarm size, number of iterations, inertia weight, individual learning factor and social learning factor.

[0067] S233. Use the training set to calculate the fitness value of each particle, that is, the mean square error, and update the individual optimum and the global optimum according to the fitness value.

[0068] S234, iteratively update the particle's velocity and position until the termination condition is met, that is, the optimal weight and bias are found, wherein the particle velocity update formula is:

[0069]

[0070] The update formula of particle position is:

[0071]

[0072] Among them, v id represents the particle speed, w represents the inertia factor, c 1 and c 2 represents the learning factor, p id and p gd represents the extreme value, x id Indicates the position of the particle.

[0073] In this embodiment, by constructing and training an early warning network, a scientific basis and decision-making support are provided for the prevention and control of landslide disasters.

[0074] Furthermore, in one embodiment, the GIS of the CIM platform is used to generate a vector layer. The CIM platform receives the level parameters of the risk area corresponding to the output of the early warning network, and in combination with the vector layer, uses DEM (Digital Elevation Model) to render the risk area into the three-dimensional model scene generated by the CIM platform. When issuing a landslide risk warning, the risk area is graded and displayed. In this example, different colors are used to display risk areas of different levels, for example, high-risk areas correspond to red, medium-risk areas correspond to orange, and low-risk areas correspond to yellow. The above-mentioned risk areas of different levels can also be displayed in other ways such as text markings.

[0075] In this embodiment, by displaying the risk area in three dimensions, real-time warning is achieved, which effectively improves the accuracy of the display of the collapsed hill position. Decision makers can monitor the risk of collapsed hill disasters in the interchange area in real time, and take corresponding prevention and control measures. In addition, this embodiment uses the three-dimensional model of the CIM platform to display the risk area, and the three-dimensional model of other platforms to display the risk area are all included in the patent protection scope of this application.

[0076] Furthermore, in one embodiment, the CIM platform presets safe routes based on the risk areas displayed in the 3D model scene, and releases relevant information to the traffic management department and the public in a timely manner. An emergency plan is generated based on the safe route and the predicted risk area, and the emergency response process when a landslide occurs is simulated based on the emergency plan. After the CIM platform generates the emergency plan, it monitors the changes in the sections of the safe route in real time and automatically adjusts the emergency plan based on the changes in the sections.

[0077] In this embodiment, by screening safe routes and publishing relevant information in a timely manner, it is convenient to take corresponding measures in a timely manner to ensure road safety. For example, the traffic management department can set up warning signs on high-risk sections to restrict vehicle traffic, and the public can choose relatively safe travel routes based on the published information to avoid entering risk areas, thereby providing safety guarantees for traffic travel. By setting the warning level and screening safe sections, the impact of landslide disasters on road safety can be effectively reduced, providing a safer and more reliable environment for people's travel.

[0078] In the second aspect, based on the above-mentioned embodiment of the ridge collapse warning method based on the CIM platform, an embodiment of a ridge collapse warning system based on the CIM platform is provided. Figure 2 As shown, the above system includes a data preparation module, a network construction module, a network training module, a prediction module and an emergency module. Specifically:

[0079] The data preparation module is used to collect historical data of factors affecting landslides and transmit them to the CIM platform, and to preprocess the historical data and select features with significant impact.

[0080] A network construction module is used to construct an early warning network.

[0081] The network training module is used to train the early warning network using historical data after preprocessing and significant influence feature selection.

[0082] The prediction module is used to collect real-time data on factors affecting landslides and, after preprocessing and selection of significantly affecting features, output the level parameters of the corresponding risk area through a trained early warning network. The CIM platform issues a landslide risk warning based on the level parameters.

[0083] Furthermore, in one embodiment, the above-mentioned landslide warning system based on the CIM platform can be Figure 3 The collapse warning system architecture based on the CIM platform is realized. The above system architecture includes data aggregation layer, network layer, data processing layer, CIM base layer and three-dimensional display layer. Specifically:

[0084] The data aggregation layer is used to use remote sensing image technology and GIS to collect data on vegetation density, topography, and land use in real time, and to use IoT sensing devices to monitor and collect data on climate, population density, and geological soil in real time.

[0085] The network layer uses communication protocols such as MQTT (Message Queuing Telemetry Transport), HTTP (Hypertext Transfer Protocol), Modbus (serial communication protocol), KNX (Konnex, distributed intelligent building system), OPC (OLE for Process Control, communication interface standard between industrial control system applications), BACnet (Building Automation and Control Network) and Link WAN (Internet of Things network management platform, such as LoRa Access Engine, WI-FNMEC, NBIOT, etc.) to transmit data from remote sensing image technology, GIS, and various IoT sensing devices to the CIM platform in real time.

[0086] The data processing layer is used to integrate devices, services, messages, and protocols through the IOT (Internet of Things) platform, to achieve unified management of various IoT sensing devices and unified scheduling of data flows. At the same time, the data platform is responsible for data collection, data analysis, data cleaning, and data display to ensure the accuracy and integrity of the data. The algorithm platform is responsible for modeling, training, testing, and prediction.

[0087] The CIM base layer uses advanced technologies such as 3D service engine, spatial analysis engine, spatial rendering engine, application service engine and place name and address engine to achieve in-depth mining and efficient use of spatial data. These engines work together to provide powerful 3D display, spatial analysis and application service capabilities.

[0088] The three-dimensional display layer is used to intuitively display the level of risk areas through different colors. Risk areas can be divided into high-risk areas, medium-risk areas and low-risk areas, corresponding to red, orange and yellow respectively, so that users can clearly understand the risk areas. At the same time, it also supports emergency drills, simulating the emergency response process when a landslide occurs, and improving the user's emergency handling capabilities. It can also make emergency plans based on the prediction results, monitor the changes in the road sections in real time, and automatically adjust the plans to ensure the accuracy and timeliness of the early warning system.

[0089] This application uses advanced technologies such as 3D modeling, big data, and artificial intelligence, and integrates multi-source heterogeneous data such as GIS, IOT, oblique photography, and orthophotos to build an efficient landslide early warning system. The system can predict in real time and display the risk level and specific location of landslide disasters in a 3D scene, achieving real-time early warning, thereby improving monitoring efficiency and reducing costs. The system can generate emergency plans and simulate the emergency response process when landslides occur. When a disaster occurs, emergency personnel can immediately adopt emergency management plans to effectively protect the lives and property safety of people in the area.

[0090] It should be noted that the serial numbers of the above-mentioned embodiments of the present application are only for description and do not represent the advantages or disadvantages of the embodiments.

[0091] The terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned drawings are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally includes steps or units that are not listed, or optionally includes other steps or units inherent to these processes, methods, products or devices. The terms "first", "second" and "third" are used to distinguish different objects, etc., and do not represent a sequence, nor do they limit "first", "second" and "third" to different types.

[0092] In the description of the embodiments of the present application, "exemplary", "for example" or "for example" are used to indicate examples, illustrations or descriptions. Any embodiment or design described as "exemplary", "for example" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary", "for example" or "for example" is intended to present related concepts in a specific way.

[0093] In some processes described in the embodiments of the present application, multiple operations or steps that appear in a specific order are included, but it should be understood that these operations or steps may not be executed in the order in which they appear in the embodiments of the present application or in parallel, and the sequence number of the operation is only used to distinguish the different operations, and the sequence number itself does not represent any execution order. In addition, these processes may include more or fewer operations, and these operations or steps may be executed in sequence or in parallel, and these operations or steps may be combined.

[0094] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, disk, CD) as described above, and includes a number of instructions for a terminal device to execute the methods described in each embodiment of the present application.

[0095] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A landslide early warning method based on the CIM platform, characterized in that: The method comprises: Collect historical data on factors affecting landslides and transmit them to the CIM platform, and perform preprocessing and significant influencing feature selection on the historical data; Construct an early warning network and train it using historical data after preprocessing and significant impact feature selection; After collecting real-time data on factors affecting landslides and preprocessing and selecting significant influencing features, the trained early warning network outputs the level parameters of the corresponding risk area, and the CIM platform issues a landslide risk warning based on the level parameters.

2. The CIM platform-based landslide early warning method according to claim 1, characterized in that: The data collected on factors affecting landslides include data on vegetation density, topography, land use, climate, population density, and geology and soil; The data on vegetation density, topography, and land use are collected using remote sensing image technology; the data on climate, population density, and geological soil are monitored and collected in real time using IoT sensing devices.

3. The CIM platform-based landslide early warning method according to claim 1, characterized in that: Preprocess historical data and real-time data, including correcting data by deleting duplicate data, filling missing values ​​in data, identifying and processing outliers, and integrating relevant data from different data sources, and labeling the corrected data with risk level parameters.

4. The CIM platform-based landslide early warning method according to claim 1, characterized in that: The significant impact on feature selection includes: Get the original features of the data after preprocessing operations and create corresponding shadow features; The random forest model is trained using the original features and shadow features, and the significantly influential features are output.

5. The method for early warning of landslide based on CIM platform as claimed in claim 1, characterized in that: The construction of the early warning network, using the historical data after preprocessing and significant influence feature selection to train the early warning network, includes: Use machine learning algorithms to build an early warning network, set the number of nodes in the input layer, hidden layer, and output layer, initialize weights and biases, and find the optimal weights and biases through optimization algorithms; Initialize the early warning network using the number of nodes, optimal weights and biases; The historical data after preprocessing and significant influencing feature selection are divided into training set, validation set and test set, and the initialized early warning network is trained, parameter adjusted and generalization performance tested respectively.

6. According to the CIM platform-based landslide warning method as described in claim 1, the CIM platform receives the level parameters of the risk area corresponding to the warning network output, renders the risk area into the three-dimensional model scene generated by the CIM platform, and displays the risk area by level while issuing a landslide risk warning.

7. According to the CIM platform-based landslide warning method as described in claim 6, the CIM platform displays risk areas of different levels in different colors in the three-dimensional model scene.

8. The CIM platform-based landslide early warning method according to claim 6, characterized in that: The CIM platform presets safe routes based on the risk areas displayed in the 3D model scene; Generate an emergency plan based on safe routes and predicted risk areas, and simulate the emergency response process when landslides occur based on the emergency plan.

9. The CIM platform-based landslide early warning method according to claim 8, characterized in that: After the CIM platform generates the emergency plan, it monitors the changes in the sections of the safe route in real time and automatically adjusts the emergency plan according to the changes in the sections.

10. A landslide early warning system based on the landslide early warning method based on the CIM platform according to any one of claims 1 to 9, characterized in that: The system comprises: A data preparation module is used to collect historical data of factors affecting landslides and transmit them to the CIM platform, and to preprocess the historical data and select significant influencing features; A network construction module, which is used to construct an early warning network; A network training module is used to train the early warning network using historical data after preprocessing and significant influence feature selection; The prediction module is used to collect real-time data on factors affecting landslides and, after preprocessing and selection of significantly affecting features, output the level parameters of the corresponding risk area through a trained early warning network. The CIM platform issues a landslide risk warning based on the level parameters.

Images