Landslide prediction method, device, equipment and storage medium based on Internet of Things

Abstract

The invention relates to the technical field of data processing, and discloses a landslide prediction method, device, equipment and storage medium based on the Internet of Things. By configuring a distributed edge environment, a landslide prediction model is constructed and trained by using a cloud server, and then sent to a shared edge storage library of the distributed edge environment. The Internet of Things end node sends the collected landslide-related perception data to a target edge device in the distributed edge environment. According to the landslide-related perception data, a landslide prediction model currently deployed or stored in a shared edge storage library is used to perform landslide prediction, and the most suitable landslide prediction model is matched for the target edge device to perform landslide prediction of the Internet of Things end node. Then, a landslide data update log is sent to the shared edge storage library, so as to help the shared edge storage library to perform incremental data retraining on the stored landslide prediction model and construct a new landslide prediction model of the cloud server, thereby improving the accuracy and real-time performance of landslide prediction and reducing system delay.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a landslide prediction method, device, equipment and storage medium based on the Internet of Things. Background Art

[0002] Landslides cause huge damage to people, landscapes, and ecosystems. If the government can be alerted before a landslide occurs, then some action plan can be made to evacuate people from the affected areas. Recently, a lot of work has been done in the field of landslide monitoring using IoT cloud architecture. Several studies have been conducted on landslide prediction and detection using artificial intelligence (AI). With the integration of AI with IoT cloud architecture, both “model building” and “model inference” can be performed on cloud-based servers with powerful elastic processors and sufficient storage capabilities.

[0003] However, there are many problems in the research of IoT cloud architecture based on landslide prediction, such as excessive delays due to network congestion and long-distance data transmission. Cloud computing cannot meet the deterministic delay requirements, especially for time-critical applications such as real-time landslide monitoring and prediction. In addition, during the monsoon season, due to heavy rainfall, most of the network infrastructure is unavailable, resulting in a complete interruption of communication with the cloud server. This makes the system unreliable because the cloud cannot be accessed when it is most needed, as landslides usually occur during the monsoon season. In addition, it is difficult to maintain an uninterrupted communication link between the IoT end nodes and the cloud. A high-speed and reliable architecture is required. Therefore, how to improve the accuracy of landslide prediction, reduce system latency, and ensure the real-time nature of landslide prediction is a technical problem that needs to be solved urgently. Summary of the invention

[0004] The present invention provides a landslide prediction method, device, equipment and storage medium based on the Internet of Things, aiming to solve at least one of the above technical problems.

[0005] To achieve the above object, the present invention provides a landslide prediction method based on the Internet of Things, comprising the following steps:

[0006] The cloud server sends several pre-built landslide prediction models to the shared edge repository in the distributed edge environment and stores them in the shared edge repository;

[0007] When the IoT end node collects the landslide-related perception data of the current prediction period, it determines the target edge device for executing the landslide prediction of each IoT end node in the current prediction period among several edge devices in the distributed edge environment according to the communication signal strength parameter;

[0008] Each IoT end node sends the landslide-related perception data collected in the current prediction period to the target edge device, so that the target edge device performs landslide prediction using the landslide-related perception data according to the currently deployed landslide prediction model or the landslide prediction model stored in the shared edge repository. The target edge device generates a landslide data update log based on the landslide prediction result, the received landslide-related perception data and the landslide monitoring data, and sends it to the shared edge repository.

[0009] Several edge devices in a distributed edge environment perform incremental data retraining based on the landslide data update log stored in a shared edge repository to update the landslide prediction model stored in the shared edge repository.

[0010] Optionally, the cloud server sends several pre-built landslide prediction models to a shared edge repository in a distributed edge environment, and performs storage steps in the shared edge repository, specifically including:

[0011] The cloud server obtains several landslide prediction models stored in the shared edge repository and landslide data update logs uploaded by several edge devices within the most recent target number of prediction cycles in the past;

[0012] When the cloud server constructs each artificial intelligence model, it selects the landslide data update log of the training edge device from a number of edge devices in a circular selection manner, extracts the landslide data features and inputs them into the artificial intelligence model for training, and obtains the target landslide prediction model;

[0013] The prediction accuracy of the target landslide prediction model is compared with several landslide prediction models stored in the shared edge repository. When the prediction accuracy of the target landslide prediction model is higher than that of any landslide prediction model, the target landslide prediction model is sent to the shared edge repository for storage.

[0014] Optionally, when the IoT end node collects the landslide-related perception data of the current prediction period, the target edge device step for executing the landslide prediction of each IoT end node in the current prediction period is determined among a plurality of edge devices in the distributed edge environment according to the communication signal strength parameter, specifically including:

[0015] When each IoT end node collects landslide-related perception data for the current prediction period, it broadcasts prediction request information to several edge devices in the distributed edge environment;

[0016] The edge device that receives the prediction request information feeds back confirmation information to the corresponding IoT end node, and the IoT end node extracts the RSSI value in the received confirmation information. The IoT end node uses the edge device with the strongest RSSI value as the target edge device for performing landslide prediction in the current prediction period;

[0017] Among them, each IoT end node includes a sensor group for performing landslide-related perception data detection, a communication module for performing information interaction with edge devices, and a control module for selecting a target edge device according to an RSSI value.

[0018] Optionally, each IoT end node sends the landslide-related perception data collected in the current prediction period to the target edge device, so that the target edge device uses the landslide-related perception data to perform landslide prediction steps according to the currently deployed landslide prediction model or the landslide prediction model stored in the shared edge repository, specifically including:

[0019] Each IoT end node sends the landslide-related perception data collected in the current prediction period to the target edge device, and the target edge device determines whether the landslide prediction model currently deployed by the target edge device is adapted to the IoT end node based on the landslide impact parameter of the first monitoring area of ​​the IoT end node in the current prediction period and the landslide impact parameter of the second monitoring area of ​​the landslide prediction model currently deployed by the target edge device; wherein the landslide impact parameter of the second monitoring area is configured as the landslide impact parameter of the monitoring area of ​​the training edge device corresponding to the landslide data update log used by the landslide prediction model currently deployed by the target edge device when training on the cloud server;

[0020] If so, the landslide prediction model currently deployed on the target edge device is driven to perform landslide prediction using the landslide-related perception data; if not, the target edge device is driven to call the most suitable landslide prediction model from the shared edge repository to perform landslide prediction using the landslide-related perception data.

[0021] Optionally, the target edge device determines whether the landslide prediction model currently deployed by the target edge device is compatible with the IoT end node according to the landslide impact parameter of the first monitoring area of ​​the IoT end node in the current prediction period and the landslide impact parameter of the second monitoring area of ​​the landslide prediction model currently deployed by the target edge device, specifically including:

[0022] The target edge device extracts the landslide impact parameters of the first monitoring area of ​​the IoT end node in the current prediction period and the landslide impact parameters of the second monitoring area of ​​the landslide prediction model currently deployed by the target edge device; wherein the landslide impact parameters of the first monitoring area and the landslide impact parameters of the second monitoring area include several geological condition parameters and several climate condition parameters;

[0023] Calculating the variance values ​​of several geological condition parameters and several climatic condition parameters in the landslide impact parameters of the first monitoring area and the corresponding geological condition parameters and corresponding climatic condition parameters in the landslide impact parameters of the second monitoring area, and determining whether the variance value exceeds a preset value;

[0024] If so, the landslide prediction model currently deployed by the target edge device is compatible with the IoT end node; if not, the landslide prediction model currently deployed by the target edge device is not compatible with the IoT end node.

[0025] Optionally, the landslide-associated sensing data includes: rainfall data collected by a rainfall sensor and displacement data collected by a displacement sensor; the landslide-associated sensing data is used to perform landslide prediction, and the target edge device generates a landslide data update log according to the landslide prediction result, the received landslide-associated sensing data and the landslide monitoring data, and sends the log to the shared edge storage repository, specifically including:

[0026] Extracting rainfall data from landslide-related perception data and constructing rainfall data features; wherein the rainfall data features include a plurality of leading rainfall data with different leading times at a target prediction moment within a current prediction period;

[0027] Inputting the rainfall data feature into a landslide prediction model currently deployed by the target edge device or into a landslide prediction model called from a shared edge repository to perform landslide prediction and obtain a landslide prediction result;

[0028] Extract the displacement data at the target prediction time from the landslide-associated perception data, use the displacement data as landslide monitoring data, and generate a landslide data update log based on the landslide monitoring data, landslide prediction results and received landslide-associated perception data and send it to the shared edge storage repository.

[0029] Optionally, the distributed edge environment further includes an MQTT server, which is configured to connect a shared edge repository and a plurality of edge devices; driving the target edge device to call the most suitable landslide prediction model step from the shared edge repository specifically includes:

[0030] The target edge device sends a landslide prediction model call request to the shared edge repository, and the MQTT server extracts the landslide monitoring data in the landslide data update log uploaded by each IoT end node within the most recent target number of prediction cycles from the shared edge repository;

[0031] Based on the landslide monitoring data, the MQTT server calculates the proportion of displacement exceeding the target value within the most recent target number prediction periods, and classifies the frequency of landslide activities of the IoT end nodes according to the proportion; according to the frequency of landslide activities of the IoT end nodes, according to the relationship of negative correlation changes, the pruning algorithm of the corresponding model pruning rate is matched for the target edge device corresponding to the IoT end node in the pruning algorithm library;

[0032] Calculate the variance values ​​of several geological condition parameters and several climatic condition parameters in the landslide impact parameters of the first monitoring area of ​​the IoT end node in the current prediction period and several geological condition parameters and several climatic condition parameters in the landslide impact parameters of the third monitoring area of ​​each landslide prediction model stored in the shared edge repository, and sort them according to the size of the variance value;

[0033] The product of the standard prediction time quantization value of each landslide prediction model stored in the shared edge repository and the model pruning rate of the pruning algorithm is estimated as the actual prediction time quantization value of each landslide prediction model, and the actual prediction time quantization value is less than the preset target time quantization value as a constraint condition, and the variance value of the landslide prediction model stored in the shared edge repository is minimized as the goal, and the most suitable landslide prediction model in the shared edge repository is optimized and solved;

[0034] The shared edge repository uses a matching pruning algorithm to prune the most suitable landslide prediction model, and sends the pruned landslide prediction model to the corresponding target edge device.

[0035] In addition, in order to achieve the above-mentioned purpose, the present invention also provides a landslide prediction device based on the Internet of Things, comprising:

[0036] A construction module is used for the cloud server to send several pre-built landslide prediction models to the shared edge storage repository in the distributed edge environment and store them in the shared edge storage repository;

[0037] A determination module is used for determining, when the IoT end node collects the landslide-related perception data of the current prediction period, a target edge device for executing landslide prediction for each IoT end node in the current prediction period among a plurality of edge devices in a distributed edge environment according to a communication signal strength parameter;

[0038] The prediction module is used for each IoT end node to send the landslide-related perception data collected in the current prediction period to the target edge device, so that the target edge device performs landslide prediction using the landslide-related perception data according to the currently deployed landslide prediction model or the landslide prediction model stored in the shared edge repository. The target edge device generates a landslide data update log based on the landslide prediction result, the received landslide-related perception data and the landslide monitoring data, and sends it to the shared edge repository;

[0039] The update module is used for several edge devices in a distributed edge environment to retrain incremental data according to the landslide data update log stored in the shared edge repository to update the landslide prediction model stored in the shared edge repository.

[0040] In addition, in order to achieve the above-mentioned purpose, the present invention also provides a landslide prediction device based on the Internet of Things, which includes: a memory, a processor, and a landslide prediction program based on the Internet of Things stored in the memory and executable on the processor. When the landslide prediction program based on the Internet of Things is executed by the processor, the steps of the landslide prediction method based on the Internet of Things as described above are implemented.

[0041] In addition, in order to achieve the above-mentioned purpose, the present invention also provides a storage medium, on which a landslide prediction program based on the Internet of Things is stored. When the landslide prediction program based on the Internet of Things is executed by a processor, the steps of the above-mentioned landslide prediction method based on the Internet of Things are implemented.

[0042] The beneficial effects of the present invention are as follows: a landslide prediction method, device, equipment and storage medium based on the Internet of Things are proposed, by configuring a distributed edge environment between a cloud server and an Internet of Things end node, using the cloud server to build and train a landslide prediction model, and then sending it to a shared edge repository of the distributed edge environment, the Internet of Things end node sends the collected landslide-related perception data to a target edge device determined based on communication signal strength in the distributed edge environment, and the target edge device performs landslide prediction based on the landslide-related perception data using the landslide prediction model currently deployed or stored in the shared edge repository, taking into account the landslide prediction model stored in the shared edge repository. The model adaptability of the target edge device to the current corresponding IoT end node is measured, as well as the deployment delay and computational burden of the landslide prediction model called to the target edge device for landslide prediction. The most suitable landslide prediction model is matched to the target edge device to execute the landslide prediction of the current IoT end node. Then the landslide prediction results, landslide-related perception data and landslide monitoring data are sent to the shared edge repository as a landslide data update log to help the shared edge repository to perform incremental data retraining on the stored landslide prediction model and build a new landslide prediction model on the cloud server, thereby improving the accuracy of landslide prediction, reducing system latency and ensuring the real-time performance of landslide prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 A schematic diagram of the device structure of the hardware operating environment involved in the embodiment of the present invention;

[0044] Figure 2 It is a flow chart of an embodiment of a landslide prediction method based on the Internet of Things of the present invention;

[0045] Figure 3 The figure is a structural block diagram of a landslide prediction device based on the Internet of Things in an embodiment of the present invention.

[0046] Description of reference numerals:

[0047] 1001 - processor; 1002 - communication bus; 1003 - user interface; 1004 - network interface; 1005 - memory; 10 - construction module; 20 - determination module; 30 - prediction module; 40 - update module.

[0048] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0049] It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0050] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0051] like Figure 1 As shown, Figure 1 It is a schematic diagram of the device structure of the hardware operating environment involved in the embodiment of the present invention.

[0052] like Figure 1 As shown, the device may include: a processor 1001, such as a CPU, a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. Among them, the communication bus 1002 is used to realize the connection and communication between these components. The user interface 1003 may include a display screen (Display), an input unit such as a keyboard (Keyboard), and the optional user interface 1003 may also include a standard wired interface and a wireless interface. The network interface 1004 may optionally include a standard wired interface and a wireless interface (such as a WI-FI interface). The memory 1005 may be a high-speed RAM memory, or it may be a stable memory (non-volatile memory), such as a disk memory. The memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0053] Those skilled in the art will understand that Figure 1 The structure of the device shown in the figure does not constitute a limitation of the device, and may include more or less components than shown in the figure, or combine certain components, or arrange the components differently.

[0054] like Figure 1 As shown, the memory 1005 as a computer storage medium may include an operating system, a network communication module, a user interface module, and a landslide prediction program based on the Internet of Things.

[0055] exist Figure 1In the terminal shown, the network interface 1004 is mainly used to connect to the backend server and perform data communication with the backend server; the user interface 1003 is mainly used to connect to the client (user end) and perform data communication with the client; and the processor 1001 can be used to call the landslide prediction program based on the Internet of Things stored in the memory 1005, and perform the following operations:

[0056] The cloud server sends several pre-built landslide prediction models to the shared edge repository in the distributed edge environment and stores them in the shared edge repository;

[0057] When the IoT end node collects the landslide-related perception data of the current prediction period, it determines the target edge device for executing the landslide prediction of each IoT end node in the current prediction period among several edge devices in the distributed edge environment according to the communication signal strength parameter;

[0058] Each IoT end node sends the landslide-related perception data collected in the current prediction period to the target edge device, so that the target edge device performs landslide prediction using the landslide-related perception data according to the currently deployed landslide prediction model or the landslide prediction model stored in the shared edge repository. The target edge device generates a landslide data update log based on the landslide prediction result, the received landslide-related perception data and the landslide monitoring data, and sends it to the shared edge repository.

[0059] Several edge devices in a distributed edge environment perform incremental data retraining based on the landslide data update log stored in a shared edge repository to update the landslide prediction model stored in the shared edge repository.

[0060] The specific embodiments of the present invention applied to the device are basically the same as the embodiments of the landslide prediction method based on the Internet of Things described below, and will not be described in detail here.

[0061] The embodiment of the present invention provides a landslide prediction method based on the Internet of Things, referring to Figure 2 , Figure 2 The figure is a flow chart of an embodiment of a landslide prediction method based on the Internet of Things of the present invention.

[0062] In this embodiment, the landslide prediction method based on the Internet of Things includes the following steps:

[0063] S100: The cloud server sends several pre-built landslide prediction models to the shared edge storage repository in the distributed edge environment, and stores them in the shared edge storage repository;

[0064] S200: When the IoT end node collects the landslide-related perception data of the current prediction period, the target edge device for executing the landslide prediction of each IoT end node in the current prediction period is determined from among a plurality of edge devices in the distributed edge environment according to the communication signal strength parameter;

[0065] S300: Each IoT end node sends the landslide-related perception data collected in the current prediction period to the target edge device, so that the target edge device performs landslide prediction using the landslide-related perception data according to the currently deployed landslide prediction model or the landslide prediction model stored in the shared edge repository. The target edge device generates a landslide data update log according to the landslide prediction result, the received landslide-related perception data and the landslide monitoring data, and sends the log to the shared edge repository.

[0066] S400: Several edge devices in a distributed edge environment perform incremental data retraining according to a landslide data update log stored in a shared edge repository to update a landslide prediction model stored in the shared edge repository.

[0067] It should be noted that the research on IoT cloud architecture for landslide prediction has many problems, such as excessive delays due to network congestion and long-distance data transmission. Cloud computing cannot meet the deterministic latency requirements, especially for time-critical applications such as real-time landslide monitoring and prediction. In addition, during the monsoon season, due to heavy rainfall, the network infrastructure is mostly unavailable, resulting in a complete interruption of communication with the cloud server. This makes the system unreliable because the cloud cannot be accessed when it is most needed because landslides usually occur during the monsoon season. In addition, it is difficult to maintain an uninterrupted communication link between the IoT end nodes and the cloud. A high-speed and reliable architecture is required.

[0068] In order to solve the above problems, this embodiment configures a distributed edge environment between a cloud server and an IoT end node, uses the cloud server to build and train a landslide prediction model, and then sends it to a shared edge repository of the distributed edge environment. The IoT end node sends the collected landslide-related perception data to a target edge device determined based on communication signal strength in the distributed edge environment. The target edge device uses the landslide prediction model currently deployed or stored in the shared edge repository to perform landslide prediction based on the landslide-related perception data. Considering the model adaptability of the landslide prediction model stored in the shared edge repository and the model adaptability of the IoT end node currently corresponding to the target edge device, as well as the deployment delay and computational burden of the landslide prediction model called to the target edge device for landslide prediction, the most suitable landslide prediction model is matched for the target edge device to perform landslide prediction of the current IoT end node. Then, the landslide prediction result, the landslide-related perception data and the landslide monitoring data are sent to the shared edge repository together as a landslide data update log, so as to help the shared edge repository to perform incremental data retraining on the stored landslide prediction model and build a new landslide prediction model on the cloud server, thereby improving the accuracy of landslide prediction, reducing system delay, and ensuring the real-time performance of landslide prediction.

[0069] In a preferred embodiment, the cloud server sends several pre-built landslide prediction models to a shared edge repository in a distributed edge environment, and performs storage steps in the shared edge repository, specifically including:

[0070] S110: The cloud server obtains a plurality of landslide prediction models stored in the shared edge repository and a plurality of landslide data update logs uploaded by edge devices within the most recent target number of prediction cycles in the past;

[0071] S120: When the cloud server constructs each artificial intelligence model, the landslide data update log of the training edge device is selected from a plurality of edge devices in a circular selection manner, and the landslide data features are extracted and input into the artificial intelligence model for training to obtain a target landslide prediction model;

[0072] S130: comparing the prediction accuracy of the target landslide prediction model with several landslide prediction models stored in the shared edge repository, and when the prediction accuracy of the target landslide prediction model is higher than that of any landslide prediction model, sending the target landslide prediction model to the shared edge repository for storage.

[0073] In this embodiment, the cloud server is configured as a server with sufficient memory and resource pool running on a scalable Linux system, capable of using a Postgres database server to store incoming data and maintain historical data log files for each landslide location, integrating Jupyter Notebook, TensorFlow, and Keras platforms for developing classification and regression AI models for landslide applications. Although edge servers can retrain AI models that have been developed in the cloud using incremental data sets, the computing power of commercial edge servers is insufficient to develop complete AI models from scratch. Therefore, developing state-of-the-art AI models from scratch still requires a high-computation cloud environment. Each time a new model is developed in a cloud server, its performance accuracy is compared with the compressed AI model running on the edge. If the accuracy of the new cloud-developed AI model is found to be better, the Docker container image containing the new model and related support packages is distributed to the edge server.

[0074] In a preferred embodiment, when the IoT end node collects the landslide-related perception data of the current prediction period, the target edge device step for executing the landslide prediction of each IoT end node in the current prediction period is determined among several edge devices in the distributed edge environment according to the communication signal strength parameter, specifically including:

[0075] S210: when each IoT end node collects landslide-related perception data of the current prediction period, it broadcasts prediction request information to a number of edge devices in the distributed edge environment;

[0076] S220: The edge device that receives the prediction request information feeds back confirmation information to the corresponding IoT end node, and the IoT end node extracts the RSSI value in the received confirmation information. The IoT end node uses the edge device with the strongest RSSI value as the target edge device for performing landslide prediction in the current prediction period;

[0077] Among them, each IoT end node includes a sensor group for performing landslide-related perception data detection, a communication module for performing information interaction with edge devices, and a control module for selecting a target edge device according to an RSSI value.

[0078] In this embodiment, applications running in an uncontrolled and resource-constrained environment have link establishment / connection problems. As a result, multiple data packets are dropped when randomly offloading data to nearby edge nodes, and energy is wasted every time one of these lost data packets must be resent; in addition, the deterioration of channel conditions leads to higher energy consumption for transmitting the same amount of data, which can lead to high latency and excessive battery consumption. To address these issues, this embodiment selects an RSSI-based data offloading method in the proposed architecture. RSSI is an indicator for evaluating the signal strength received by the edge device. The stronger the RSSI value, the better the link quality, so the power consumption and latency are smaller. When offloading data to nearby edge devices, the IoT end node scans available edge devices. All available edge devices around the IoT end node run the previously deployed AI model. The IoT end node broadcasts a connection request to the available edge devices through the LoRa / Zigbee communication module. After receiving the request, the edge device sends a confirmation back to the IoT end node. The IoT end node determines the RSSI value of the confirmation message received from the edge device and compares all RSSI values. It then offloads the data to the edge device with the strongest RSSI value.

[0079] In a preferred embodiment, each IoT end node sends the landslide-related perception data collected in the current prediction period to the target edge device, so that the target edge device uses the landslide-related perception data to perform landslide prediction steps according to the currently deployed landslide prediction model or the landslide prediction model stored in the shared edge repository, specifically including:

[0080] S310: Each IoT end node sends the landslide-related perception data collected in the current prediction period to the target edge device, and the target edge device determines whether the landslide prediction model currently deployed by the target edge device is adapted to the IoT end node according to the landslide impact parameter of the first monitoring area of ​​the IoT end node in the current prediction period and the landslide impact parameter of the second monitoring area of ​​the landslide prediction model currently deployed by the target edge device; wherein the landslide impact parameter of the second monitoring area is configured as the landslide impact parameter of the monitoring area of ​​the training edge device corresponding to the landslide data update log used by the landslide prediction model currently deployed by the target edge device when training on the cloud server;

[0081] S320: If yes, drive the landslide prediction model currently deployed on the target edge device to perform landslide prediction using the landslide associated perception data; if no, drive the target edge device to call the most suitable landslide prediction model from the shared edge repository to perform landslide prediction using the landslide associated perception data.

[0082] Furthermore, the target edge device determines whether the landslide prediction model currently deployed by the target edge device is compatible with the IoT end node according to the landslide impact parameter of the first monitoring area of ​​the IoT end node in the current prediction period and the landslide impact parameter of the second monitoring area of ​​the landslide prediction model currently deployed by the target edge device, and the steps specifically include:

[0083] S311: The target edge device extracts the landslide impact parameters of the first monitoring area of ​​the IoT end node in the current prediction period and the landslide impact parameters of the second monitoring area of ​​the landslide prediction model currently deployed by the target edge device; wherein the landslide impact parameters of the first monitoring area and the landslide impact parameters of the second monitoring area include a plurality of geological condition parameters and a plurality of climate condition parameters;

[0084] S312: Calculate the variance value between a plurality of geological condition parameters and a plurality of climatic condition parameters in the landslide impact parameters of the first monitoring area and the corresponding geological condition parameters and the corresponding climatic condition parameters in the landslide impact parameters of the second monitoring area, and determine whether the variance value exceeds a preset value;

[0085] S313: If so, the landslide prediction model currently deployed by the target edge device is compatible with the IoT end node; if not, the landslide prediction model currently deployed by the target edge device is not compatible with the IoT end node.

[0086] In this embodiment, this stage will be initiated when the data is offloaded to an available edge device. First, the edge device checks the required AI model in its own storage. If the model is not available, the edge device will start the communication pipeline and request to establish a connection with the MQTT agent. After the connection, the edge device uses the request-response mechanism of MQTT-5 to request a suitable model from the shared repository. Specifically, considering that the landslide prediction model obtained by training the cloud server based on the landslide-related perception data collected by different IoT end nodes in different environments (including geological conditions and climatic conditions) has higher adaptability in this environment (the main reason is that the landslide prediction model obtained by training only through the feature of rainfall is more obviously affected by the landslide influencing parameters in the monitored area), it is predicted that the landslide prediction model obtained in this environment or similar environments has better prediction effect.

[0087] Based on this, when judging whether the landslide prediction model currently deployed by the target edge device is compatible with the current IoT terminal node, this embodiment mainly explores the variance value between the landslide impact parameters of the first monitoring area of ​​the IoT terminal node in the current prediction period and the landslide impact parameters of the second monitoring area of ​​the landslide prediction model currently deployed by the target edge device, and judges whether the variance value exceeds a preset value. If so, it indicates that the landslide prediction model currently deployed by the target edge device has a large difference in the environment of the landslide impact parameters of the monitoring area of ​​the corresponding training edge device when the landslide data update log used in the cloud server training. At this time, using the currently deployed landslide prediction model for landslide prediction will have the problem of poor prediction effect. It is necessary to call the most suitable landslide prediction model from the shared edge repository and perform landslide prediction using landslide-related perception data.

[0088] In a preferred embodiment, the landslide-related sensing data includes: rainfall data collected by a rainfall sensor and displacement data collected by a displacement sensor; the landslide-related sensing data is used to perform landslide prediction, and the target edge device generates a landslide data update log according to the landslide prediction result, the received landslide-related sensing data and the landslide monitoring data, and sends the log to the shared edge storage repository, specifically including:

[0089] S330: extracting rainfall data from the landslide-related perception data and constructing rainfall data features; wherein the rainfall data features include a plurality of leading rainfall data with different leading times at the target prediction moment within the current prediction period;

[0090] S340: Inputting the rainfall data feature into the landslide prediction model currently deployed by the target edge device or the landslide prediction model called from the shared edge storage library to perform landslide prediction and obtain a landslide prediction result;

[0091] S350: extract the displacement data at the target prediction time from the landslide-associated perception data, use the displacement data as landslide monitoring data, and generate a landslide data update log based on the landslide monitoring data, landslide prediction results and received landslide-associated perception data, and send it to the shared edge storage repository.

[0092] Furthermore, the distributed edge environment further includes an MQTT server, which is configured to connect a shared edge repository and a plurality of edge devices; driving the target edge device to call the most suitable landslide prediction model from the shared edge repository, specifically including:

[0093] S321: The target edge device sends a landslide prediction model call request to the shared edge repository, and the MQTT server extracts the landslide monitoring data in the landslide data update log uploaded by each IoT end node within the most recent target number of prediction cycles from the shared edge repository;

[0094] S322: The MQTT server calculates the proportion of displacement exceeding the target value within the most recent target number of prediction periods based on the landslide monitoring data, and classifies the frequency of landslide activities of the IoT end nodes according to the proportion; and matches the pruning algorithm of the corresponding model pruning rate for the target edge device corresponding to the IoT end node in the pruning algorithm library according to the frequency of landslide activities of the IoT end node and the relationship of negative correlation change;

[0095] S323: Calculate the variance values ​​of several geological condition parameters and several climatic condition parameters in the landslide impact parameters of the first monitoring area of ​​the IoT end node in the current prediction period and several geological condition parameters and several climatic condition parameters in the landslide impact parameters of the third monitoring area of ​​each landslide prediction model stored in the shared edge repository, and sort them according to the size of the variance values;

[0096] S324: the product of the standard prediction time quantization value of each landslide prediction model stored in the shared edge repository and the model pruning rate of the pruning algorithm is estimated as the actual prediction time quantization value of each landslide prediction model, and the actual prediction time quantization value is less than the preset target time quantization value as a constraint condition, and the variance value of the landslide prediction model stored in the shared edge repository is minimized as the goal, and the most suitable landslide prediction model in the shared edge repository is optimized and solved;

[0097] S325: The shared edge repository performs pruning operations on the most suitable landslide prediction model using a matching pruning algorithm, and sends the pruned landslide prediction model to the corresponding target edge device.

[0098] In this embodiment, a distributed edge environment is configured between a cloud server and an IoT end node, a landslide prediction model is constructed and trained using the cloud server, and then sent to a shared edge repository of the distributed edge environment. The IoT end node sends the collected landslide-related perception data to a target edge device determined based on communication signal strength in the distributed edge environment. The target edge device uses the landslide prediction model currently deployed or stored in the shared edge repository to perform landslide prediction based on the landslide-related perception data. Considering the model adaptability of the landslide prediction model stored in the shared edge repository and the model of the IoT end node currently corresponding to the target edge device, as well as the deployment delay and computational burden of the landslide prediction model called to the target edge device for landslide prediction, the landslide prediction model of the shared edge repository is reasonably selected and pruned, so that the prediction effect, deployment time and prediction delay of the landslide prediction model allocated to the target edge device are optimal as a whole, and then the most suitable landslide prediction model is matched to the target edge device to perform landslide prediction of the current IoT end node.

[0099] After that, the landslide prediction results, landslide-related perception data and landslide monitoring data are sent to the shared edge repository as a landslide data update log to help the shared edge repository to perform incremental data retraining on the stored landslide prediction model and build a new landslide prediction model on the cloud server, thereby improving the accuracy of landslide prediction, reducing system latency and ensuring the real-time nature of landslide prediction.

[0100] Reference Figure 3 , Figure 3 It is a structural block diagram of an embodiment of a landslide prediction device based on the Internet of Things of the present invention.

[0101] like Figure 3 As shown, the landslide prediction device based on the Internet of Things proposed in the embodiment of the present invention includes:

[0102] A construction module 10 is used for the cloud server to send a plurality of pre-built landslide prediction models to a shared edge storage repository in a distributed edge environment, and store them in the shared edge storage repository;

[0103] The determination module 20 is used for determining, when the IoT end node collects the landslide-related perception data of the current prediction period, a target edge device for executing the landslide prediction of each IoT end node in the current prediction period from among a plurality of edge devices in the distributed edge environment according to the communication signal strength parameter;

[0104] The prediction module 30 is used for each IoT end node to send the landslide-related perception data collected in the current prediction period to the target edge device, so that the target edge device performs landslide prediction using the landslide-related perception data according to the currently deployed landslide prediction model or the landslide prediction model stored in the shared edge storage repository. The target edge device generates a landslide data update log according to the landslide prediction result, the received landslide-related perception data and the landslide monitoring data, and sends it to the shared edge storage repository;

[0105] The updating module 40 is used for a plurality of edge devices in a distributed edge environment to perform incremental data retraining according to the landslide data update log stored in the shared edge repository, so as to update the landslide prediction model stored in the shared edge repository.

[0106] Other embodiments or specific implementations of the landslide prediction device based on the Internet of Things of the present invention can refer to the above-mentioned method embodiments, which will not be repeated here.

[0107] In addition, the present invention also proposes a landslide prediction device based on the Internet of Things, which includes: a memory, a processor, and a landslide prediction program based on the Internet of Things stored in the memory and executable on the processor. When the landslide prediction program based on the Internet of Things is executed by the processor, the steps of the landslide prediction method based on the Internet of Things as described above are implemented.

[0108] The specific implementation of the landslide prediction device based on the Internet of Things in the present application is basically the same as the various embodiments of the landslide prediction method based on the Internet of Things mentioned above, and will not be repeated here.

[0109] In addition, the present invention also proposes a readable storage medium, which includes a computer-readable storage medium on which a landslide prediction program based on the Internet of Things is stored. The readable storage medium may be Figure 1 The memory 1005 in the terminal may also be at least one of a ROM (Read-Only Memory) / RAM (Random Access Memory), a magnetic disk, and an optical disk. The readable storage medium includes a number of instructions for enabling a landslide prediction device based on the Internet of Things with a processor to execute the landslide prediction method based on the Internet of Things described in various embodiments of the present invention.

[0110] The specific implementation methods in the readable storage medium of the present application are basically the same as the above-mentioned embodiments of the landslide prediction method based on the Internet of Things, and will not be repeated here.

[0111] It is understood that, in the description of this specification, the description with reference to the terms "one embodiment", "another embodiment", "other embodiments", or "first to Nth embodiments" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0112] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0113] The serial numbers of the above embodiments of the present invention are only for description and do not represent the advantages or disadvantages of the embodiments.

[0114] 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 invention 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, magnetic disk, optical disk) as described above, and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present invention.

[0115] The above are only preferred embodiments of the present invention, and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A landslide prediction method based on the Internet of Things, characterized in that: The following steps are involved: The cloud server sends several pre-built landslide prediction models to the shared edge storage repository in the distributed edge environment, and stores them in the shared edge storage repository; specifically, the cloud server obtains several landslide prediction models stored in the shared edge storage repository and landslide data update logs uploaded by several edge devices in the most recent target number of prediction cycles in the past; when the cloud server builds each artificial intelligence model, the landslide data update log of the training edge device is selected from several edge devices in a cyclic selection manner, and the landslide data features are extracted and input into the artificial intelligence model for training to obtain a target landslide prediction model; the prediction accuracy of the target landslide prediction model is compared with the prediction accuracy of several landslide prediction models stored in the shared edge storage repository, and when the prediction accuracy of the target landslide prediction model is higher than that of any landslide prediction model, the target landslide prediction model is sent to the shared edge storage repository for storage; When the IoT end node collects the landslide-related perception data of the current prediction period, the target edge device for executing the landslide prediction of each IoT end node in the current prediction period is determined among several edge devices in the distributed edge environment according to the communication signal strength parameter; specifically, each IoT end node broadcasts prediction request information to several edge devices in the distributed edge environment when the landslide-related perception data of the current prediction period is collected; the edge device that receives the prediction request information feeds back confirmation information to the corresponding IoT end node, the IoT end node extracts the RSSI value in the received confirmation information, and the IoT end node uses the edge device with the strongest RSSI value as the target edge device for executing the landslide prediction in the current prediction period; wherein each IoT end node includes a sensor group for executing landslide-related perception data detection, a communication module for executing information interaction with the edge device, and a control module for selecting the target edge device according to the RSSI value; Each IoT end node sends the landslide-related perception data collected in the current prediction period to the target edge device, so that the target edge device performs landslide prediction using the landslide-related perception data according to the currently deployed landslide prediction model or the landslide prediction model stored in the shared edge repository. The target edge device generates a landslide data update log based on the landslide prediction result, the received landslide-related perception data and the landslide monitoring data, and sends it to the shared edge repository. Among them, each IoT end node sends the landslide-related perception data collected in the current prediction period to the target edge device, so that the target edge device uses the landslide-related perception data to perform landslide prediction according to the currently deployed landslide prediction model or the landslide prediction model stored in the shared edge storage library, specifically including: each IoT end node sends the landslide-related perception data collected in the current prediction period to the target edge device, and the target edge device judges whether the landslide prediction model currently deployed by the target edge device is adapted to the IoT end node according to the landslide impact parameter of the first monitoring area of ​​the IoT end node in the current prediction period and the landslide impact parameter of the second monitoring area of ​​the landslide prediction model currently deployed by the target edge device; wherein the landslide impact parameter of the second monitoring area is configured as the landslide impact parameter of the monitoring area of ​​the training edge device corresponding to the landslide data update log used by the landslide prediction model currently deployed by the target edge device when training on the cloud server; if so, the landslide prediction model currently deployed by the target edge device is driven to perform landslide prediction using the landslide-related perception data; if not, the target edge device is driven to call the most suitable landslide prediction model from the shared edge storage library and perform landslide prediction using the landslide-related perception data; Wherein, the landslide associated perception data includes: rainfall data collected by a rainfall sensor and displacement data collected by a displacement sensor; landslide prediction is performed using the landslide associated perception data, and the target edge device generates a landslide data update log according to the landslide prediction result, the received landslide associated perception data and the landslide monitoring data, and sends it to the shared edge repository, specifically including: extracting rainfall data from the landslide associated perception data and constructing rainfall data features; wherein the rainfall data features include several leading rainfall data with different leading times at the target prediction moment within the current prediction period; inputting the rainfall data features into the landslide prediction model currently deployed by the target edge device or the landslide prediction model called from the shared edge repository to perform landslide prediction and obtain the landslide prediction result; extracting the displacement data at the target prediction moment from the landslide associated perception data, using the displacement data as the landslide monitoring data, and generating a landslide data update log based on the landslide monitoring data, the landslide prediction result and the received landslide associated perception data and sending it to the shared edge repository; Several edge devices in the distributed edge environment perform incremental data retraining according to the landslide data update log stored in the shared edge repository to update the landslide prediction model stored in the shared edge repository; Among them, the distributed edge environment also includes an MQTT server, which is configured to connect a shared edge repository and several edge devices; drive the target edge device to call the most suitable landslide prediction model from the shared edge repository, specifically including: the target edge device sends a landslide prediction model call request to the shared edge repository, and the MQTT server extracts the landslide monitoring data in the landslide data update log uploaded by each IoT end node in the past most recent target number of prediction cycles from the shared edge repository; based on the landslide monitoring data, the MQTT server calculates the proportion of displacement exceeding the target value in the most recent target number of prediction cycles, and divides the IoT end nodes into landslide activity frequency levels according to the proportion; according to the landslide activity frequency level of the IoT end node, according to the negative correlation change relationship, the pruning algorithm of the corresponding model pruning rate is matched for the target edge device corresponding to the IoT end node in the pruning algorithm library; calculate The variance values ​​of several geological condition parameters and several climatic condition parameters in the landslide impact parameters of the first monitoring area of ​​the IoT end node in the current prediction period and several geological condition parameters and several climatic condition parameters in the landslide impact parameters of the third monitoring area of ​​each landslide prediction model stored in the shared edge repository are sorted according to the size of the variance values; the product of the standard prediction time quantization value of each landslide prediction model stored in the shared edge repository and the model pruning rate of the pruning algorithm is estimated as the actual prediction time quantization value of each landslide prediction model, and the actual prediction time quantization value is less than the preset target time quantization value as a constraint condition, and the variance value of the landslide prediction model stored in the shared edge repository is minimized as the goal, and the most suitable landslide prediction model in the shared edge repository is optimized and solved; the shared edge repository uses a matching pruning algorithm to perform pruning operations on the most suitable landslide prediction model, and sends the pruned landslide prediction model to the corresponding target edge device.

2. The landslide prediction method based on the Internet of Things according to claim 1, characterized in that: The target edge device determines whether the landslide prediction model currently deployed by the target edge device is compatible with the IoT end node according to the landslide impact parameter of the first monitoring area of ​​the IoT end node in the current prediction period and the landslide impact parameter of the second monitoring area of ​​the landslide prediction model currently deployed by the target edge device, specifically including: The target edge device extracts the landslide impact parameters of the first monitoring area of ​​the IoT end node in the current prediction period and the landslide impact parameters of the second monitoring area of ​​the landslide prediction model currently deployed by the target edge device; wherein the landslide impact parameters of the first monitoring area and the landslide impact parameters of the second monitoring area include several geological condition parameters and several climate condition parameters; Calculating the variance values ​​of several geological condition parameters and several climatic condition parameters in the landslide impact parameters of the first monitoring area and the corresponding geological condition parameters and corresponding climatic condition parameters in the landslide impact parameters of the second monitoring area, and determining whether the variance value exceeds a preset value; If so, the landslide prediction model currently deployed by the target edge device is compatible with the IoT end node; if not, the landslide prediction model currently deployed by the target edge device is not compatible with the IoT end node.

3. A landslide prediction device based on the Internet of Things, characterized in that: include: A construction module is used for a cloud server to send several pre-built landslide prediction models to a shared edge repository in a distributed edge environment and store them in the shared edge repository; specifically, the cloud server obtains several landslide prediction models stored in the shared edge repository and landslide data update logs uploaded by several edge devices in the most recent target number of prediction cycles in the past; when the cloud server constructs each artificial intelligence model, the landslide data update log of the training edge device is selected from several edge devices in a cyclic selection manner, and the landslide data features are extracted and input into the artificial intelligence model for training to obtain a target landslide prediction model; the prediction accuracy of the target landslide prediction model is compared with the prediction accuracy of several landslide prediction models stored in the shared edge repository, and when the prediction accuracy of the target landslide prediction model is higher than that of any landslide prediction model, the target landslide prediction model is sent to the shared edge repository for storage; A determination module is used for determining the target edge device for executing landslide prediction for each IoT end node in the current prediction period among several edge devices in the distributed edge environment when the IoT end node collects the landslide-related perception data of the current prediction period, according to the communication signal strength parameter; specifically comprising: when each IoT end node collects the landslide-related perception data of the current prediction period, broadcasting prediction request information to several edge devices in the distributed edge environment; the edge device receiving the prediction request information feeds back confirmation information to the corresponding IoT end node, the IoT end node extracts the RSSI value in the received confirmation information, and the IoT end node uses the edge device with the strongest RSSI value as the target edge device for executing landslide prediction in the current prediction period; wherein each IoT end node includes a sensor group for executing landslide-related perception data detection, a communication module for executing information interaction with the edge device, and a control module for selecting the target edge device according to the RSSI value; The prediction module is used for each IoT end node to send the landslide-related perception data collected in the current prediction period to the target edge device, so that the target edge device performs landslide prediction using the landslide-related perception data according to the currently deployed landslide prediction model or the landslide prediction model stored in the shared edge repository. The target edge device generates a landslide data update log based on the landslide prediction result, the received landslide-related perception data and the landslide monitoring data, and sends it to the shared edge repository; Among them, each IoT end node sends the landslide-related perception data collected in the current prediction period to the target edge device, so that the target edge device uses the landslide-related perception data to perform landslide prediction according to the currently deployed landslide prediction model or the landslide prediction model stored in the shared edge storage library, specifically including: each IoT end node sends the landslide-related perception data collected in the current prediction period to the target edge device, and the target edge device judges whether the landslide prediction model currently deployed by the target edge device is adapted to the IoT end node according to the landslide impact parameter of the first monitoring area of ​​the IoT end node in the current prediction period and the landslide impact parameter of the second monitoring area of ​​the landslide prediction model currently deployed by the target edge device; wherein the landslide impact parameter of the second monitoring area is configured as the landslide impact parameter of the monitoring area of ​​the training edge device corresponding to the landslide data update log used by the landslide prediction model currently deployed by the target edge device when training on the cloud server; if so, the landslide prediction model currently deployed by the target edge device is driven to perform landslide prediction using the landslide-related perception data; if not, the target edge device is driven to call the most suitable landslide prediction model from the shared edge storage library and perform landslide prediction using the landslide-related perception data; Wherein, the landslide associated perception data includes: rainfall data collected by a rainfall sensor and displacement data collected by a displacement sensor; landslide prediction is performed using the landslide associated perception data, and the target edge device generates a landslide data update log according to the landslide prediction result, the received landslide associated perception data and the landslide monitoring data, and sends it to the shared edge repository, specifically including: extracting rainfall data from the landslide associated perception data and constructing rainfall data features; wherein the rainfall data features include several leading rainfall data with different leading times at the target prediction moment within the current prediction period; inputting the rainfall data features into the landslide prediction model currently deployed by the target edge device or the landslide prediction model called from the shared edge repository to perform landslide prediction and obtain the landslide prediction result; extracting the displacement data at the target prediction moment from the landslide associated perception data, using the displacement data as the landslide monitoring data, and generating a landslide data update log based on the landslide monitoring data, the landslide prediction result and the received landslide associated perception data and sending it to the shared edge repository; An update module, used for several edge devices in a distributed edge environment to perform incremental data retraining according to the landslide data update log stored in the shared edge repository, so as to update the landslide prediction model stored in the shared edge repository; Among them, the distributed edge environment also includes an MQTT server, which is configured to connect a shared edge repository and several edge devices; drive the target edge device to call the most suitable landslide prediction model from the shared edge repository, specifically including: the target edge device sends a landslide prediction model call request to the shared edge repository, and the MQTT server extracts the landslide monitoring data in the landslide data update log uploaded by each IoT end node in the past most recent target number of prediction cycles from the shared edge repository; based on the landslide monitoring data, the MQTT server calculates the proportion of displacement exceeding the target value in the most recent target number of prediction cycles, and divides the IoT end nodes into landslide activity frequency levels according to the proportion; according to the landslide activity frequency level of the IoT end node, according to the negative correlation change relationship, the pruning algorithm of the corresponding model pruning rate is matched for the target edge device corresponding to the IoT end node in the pruning algorithm library; calculate The variance values ​​of several geological condition parameters and several climatic condition parameters in the landslide impact parameters of the first monitoring area of ​​the IoT end node in the current prediction period and several geological condition parameters and several climatic condition parameters in the landslide impact parameters of the third monitoring area of ​​each landslide prediction model stored in the shared edge repository are sorted according to the size of the variance values; the product of the standard prediction time quantization value of each landslide prediction model stored in the shared edge repository and the model pruning rate of the pruning algorithm is estimated as the actual prediction time quantization value of each landslide prediction model, and the actual prediction time quantization value is less than the preset target time quantization value as a constraint condition, and the variance value of the landslide prediction model stored in the shared edge repository is minimized as the goal, and the most suitable landslide prediction model in the shared edge repository is optimized and solved; the shared edge repository uses a matching pruning algorithm to perform pruning operations on the most suitable landslide prediction model, and sends the pruned landslide prediction model to the corresponding target edge device.

4. A landslide prediction device based on the Internet of Things, characterized in that: The IoT-based landslide prediction device comprises: a memory, a processor, and an IoT-based landslide prediction program stored in the memory and executable on the processor. When the IoT-based landslide prediction program is executed by the processor, the steps of the IoT-based landslide prediction method as described in any one of claims 1 to 2 are implemented.

5. A storage medium, characterized in that: The storage medium stores a landslide prediction program based on the Internet of Things, and when the landslide prediction program based on the Internet of Things is executed by the processor, the steps of the landslide prediction method based on the Internet of Things as described in any one of claims 1 to 2 are implemented.

Images