Disaster emergency warning method, system, equipment and medium based on Internet of Things technology

Abstract

The present invention relates to the field of geological disaster early warning technology, and specifically to a disaster emergency early warning method, system, equipment and medium based on the Internet of Things technology. Image data of a terrain area to be monitored is collected through remote sensing images, laser radar and drone photography, and corresponding remote sensing image sets, radar data sets and drone image sets are respectively established; and physical data of geological structures in the terrain area to be monitored are collected to establish a geological structure data set; terrain features are extracted to establish a terrain feature database; geological structure features of the terrain area to be monitored are obtained to establish a geological structure feature database; a three-dimensional site model of the terrain area to be monitored is constructed; sensor data of the terrain area to be monitored is obtained based on the Internet of Things technology, and visual interaction is performed with the three-dimensional site model to establish a disaster emergency early warning platform; disaster early warning of the terrain area to be monitored is performed by real-time acquisition of remote sensing images, sensor collection data and historical disaster data.

Description

Technical Field

[0001] The present invention relates to the field of geological disaster early warning technology, and specifically to a disaster emergency early warning method, system, equipment and medium based on Internet of Things technology. Background Art

[0002] The sudden occurrence of natural disasters will cause huge losses of life and property to human society. Strengthening disaster prevention and mitigation warning work is the key task of emergency management departments. Taking precautions against natural disasters can minimize post-disaster losses, so emergency management departments are gradually shifting from post-disaster warnings to pre-disaster warnings. However, there are still some problems with traditional pre-disaster warnings and emergency management, mainly including the following aspects:

[0003] (1) Information transmission delay: Traditional disaster warning methods usually rely on manual monitoring and reporting. The information transmission link is long, resulting in delays in the transmission and response of warning information, which affects the effectiveness of emergency rescue and emergency response.

[0004] (2) Low accuracy: Traditional disaster warning systems often rely on human judgment and experience, lack scientific data support and analysis, and have certain deficiencies in the accuracy of disaster prediction, which may result in false warnings or missed reports.

[0005] (3) Limited emergency response capabilities: Traditional disaster emergency response usually relies on manual coordination and resource allocation, lacks intelligent and automated support, and results in the need to improve the efficiency and accuracy of emergency response.

[0006] (4) Lack of real-time monitoring: Traditional disaster warning systems often rely on limited monitoring methods and are unable to achieve real-time and comprehensive monitoring of disaster events, making it difficult to obtain timely information about the occurrence of disasters.

[0007] (5) Vulnerabilities and weak links: Traditional disaster warning and emergency response have some vulnerabilities and weak links, such as untimely equipment maintenance and distorted information transmission, which affect the reliability and stability of the entire warning and response system. Summary of the invention

[0008] In order to solve the above problems, the first aspect of the present invention provides a disaster emergency warning method based on Internet of Things technology, comprising:

[0009] S1. Collect image data of the terrain area to be monitored through remote sensing images, laser radar and drone photography, and establish corresponding remote sensing image sets, radar data sets and drone image sets respectively; and collect geological structure physical data of the terrain area to be monitored to establish a geological structure data set;

[0010] S2. Use deep learning technology to perform feature recognition on the image data of remote sensing image sets, radar data sets and drone image sets, extract terrain features and establish a terrain feature database;

[0011] S3. Use deep learning technology to identify and process geological structure data sets, obtain geological structure characteristics of the terrain area to be monitored, and establish a geological structure characteristic database;

[0012] S4. Construct a three-dimensional site model of the terrain area to be monitored based on the terrain feature database and the geological structure feature database;

[0013] S5. Based on the Internet of Things technology, sensor data of the terrain area to be monitored is obtained, and visual interaction is carried out with the three-dimensional site model to establish a disaster emergency warning platform; disaster warnings are carried out in the terrain area to be monitored by acquiring remote sensing images, sensor data, and historical disaster data in real time.

[0014] The S2 also includes building a machine learning model and performing deep learning training, the specific method is as follows:

[0015] Collect image data of three-dimensional sites and urban agglomerations to establish sample data sets for model training;

[0016] Build machine learning models;

[0017] Perform deep learning training on the machine learning model through sample data sets, determine the loss function, optimizer and learning rate parameters of the machine learning model, evaluate and tune the machine learning model, and obtain a trained machine learning model.

[0018] The specific method of collecting the image data of the three-dimensional site and the urban agglomeration and establishing the sample data set for model training is:

[0019] Data preparation: collect image data of three-dimensional sites and urban agglomerations to establish sample data sets;

[0020] Data labeling: labeling or annotating the image data in the sample data set, wherein the annotation information uses a frame or mask to represent the location and shape of the target;

[0021] Data processing: Perform data scaling, normalization, and data enhancement on each image in the sample data set to match the image data with the characteristics of the three-dimensional sites and urban agglomerations.

[0022] In some implementations of the first aspect, the machine learning model adopts a Faster R-CNN model, including a region extraction network and a target recognition network. The region extraction network is used to extract image features to generate candidate target regions; the target recognition network is used to classify and regress the candidate target regions to determine the location and category of the final target.

[0023] The Faster R-CNN model training also includes distributed training by building a deep learning framework TensorFlow.

[0024] In some implementations of the first aspect, in S2, feature recognition is performed on the image data of the remote sensing image set, the radar data set, and the drone image set, and terrain features are extracted to establish a terrain feature database. The specific method is:

[0025] S2.1. Import remote sensing image sets, radar datasets and drone image sets respectively;

[0026] S2.2, process remote sensing image sets to extract ground feature information; process radar data sets to generate point cloud data and extract terrain information; process drone image sets to extract building outlines and crowded areas;

[0027] S2.3. The extracted ground feature information, terrain information, building outlines and crowded areas are integrated with multi-source data to establish a terrain feature database.

[0028] In some implementations of the first aspect, the IoT disaster warning method further includes real-time collection of remote sensing images, laser radar, and image data taken by drones, and extracting new features to update the three-dimensional site model.

[0029] The second aspect provides a disaster emergency warning system based on Internet of Things technology, including:

[0030] The acquisition module is used to collect image data of the terrain area to be monitored through remote sensing images, laser radar and drone photography, and to establish corresponding remote sensing image sets, radar data sets and drone image sets respectively; and to collect geological structure physical data of the terrain area to be monitored and establish a geological structure data set;

[0031] The recognition and extraction module is used to use machine learning and deep learning technologies to perform feature recognition on the image data of remote sensing image sets, radar data sets, and drone image sets, extract terrain features, and establish a terrain feature database;

[0032] The identification and extraction module is also used to use machine learning and deep learning technology to identify and process the geological structure data set, obtain the geological structure characteristics of the terrain area to be monitored, and establish a geological structure characteristic database;

[0033] A three-dimensional site construction module is used to construct a three-dimensional site model of the terrain area to be monitored based on the terrain feature database and the geological structure feature database;

[0034] The monitoring and early warning module is used to obtain sensor data of the terrain area to be monitored based on the Internet of Things technology, conduct visual interaction with the three-dimensional site model, and establish a disaster emergency warning platform; it provides disaster warnings in the terrain area to be monitored by acquiring remote sensing images, sensor collection data, and historical disaster data in real time.

[0035] The third aspect provides a disaster emergency warning device based on Internet of Things technology, including a processor and a memory, wherein the processor implements the disaster emergency warning method based on Internet of Things technology as described above when executing program data stored in the memory.

[0036] A fourth aspect provides a readable storage medium for storing control program data, wherein the control program data, when executed by a processor, implements the disaster emergency warning method based on the Internet of Things technology as described above.

[0037] The beneficial effects are as follows: the present invention rapidly constructs a three-dimensional site model based on machine learning and deep learning technologies, and realizes efficient modeling of cities, population gathering areas and important building structures under different terrain conditions by automatically identifying terrain features and geological structure features; at the same time, it combines the Internet of Things technology to collect data in real time, and builds a disaster emergency warning platform supported by the Internet of Things technology in combination with the three-dimensional site model, so as to realize timely updating of the three-dimensional site model, provide high-precision and real-time site information, and perform disaster prediction and classification through remote sensing images, sensor data, and historical disaster data, so as to realize pre-disaster warning and forecast of site information. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 This is a flow chart of the disaster emergency warning method based on Internet of Things technology. DETAILED DESCRIPTION

[0039] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings.

[0040] This embodiment provides a disaster emergency warning method based on Internet of Things technology, and the specific implementation steps are as follows:

[0041] S1. Collect image data of the terrain area to be monitored through remote sensing images, laser radar and drone photography, and establish corresponding remote sensing image sets, radar data sets and drone image sets respectively; and collect geological structure physical data of the terrain area to be monitored to establish a geological structure data set;

[0042] S2. Use deep learning technology to perform feature recognition on the image data of remote sensing image sets, radar data sets and drone image sets, extract terrain features and establish a terrain feature database;

[0043] S3. Use deep learning technology to identify and process geological structure data sets, obtain geological structure characteristics of the terrain area to be monitored, and establish a geological structure characteristic database;

[0044] S4. Construct a three-dimensional site model of the terrain area to be monitored based on the terrain feature database and the geological structure feature database;

[0045] S5. Based on the Internet of Things technology, sensor data of the terrain area to be monitored is obtained, and visual interaction is carried out with the three-dimensional site model to establish a disaster emergency warning platform; disaster warnings are carried out in the terrain area to be monitored by acquiring remote sensing images, sensor data, and historical disaster data in real time.

[0046] Based on the Internet of Things technology, the collected data of various sensors in the terrain area to be monitored are obtained, and the collected sensor data are imported into the three-dimensional site model for visual interaction, and an intelligent disaster emergency warning platform is established to realize real-time monitoring and early warning of various disasters (floods, earthquakes, mudslides, etc.).

[0047] By acquiring remote sensing images, sensor data, and historical disaster data for data analysis, the disaster type and probability of the terrain area to be monitored are predicted, specifically:

[0048] Through remote sensing technology, high-resolution, multi-angle, and multi-temporal remote sensing image data are collected from the monitoring area, and target features are extracted through image processing to determine the remote sensing image prediction results;

[0049] The collected data of the sensor is obtained through the Internet of Things technology, and the collected data is imported into the three-dimensional site model for simulation to obtain the simulation prediction results of the sensor data;

[0050] Conduct data mining on historical disaster data, establish a disaster type classification model based on historical disaster data, and determine data mining prediction results based on sensor data;

[0051] The disaster prediction results of the terrain area to be monitored are determined through remote sensing image prediction results, sensor data simulation prediction results and data mining prediction results. According to the disaster prediction results, the disaster emergency warning platform issues a disaster alarm.

[0052] Sample Code

[0053] #Use remote sensing images for image prediction

[0054] defimage_classification(image_data,labels):

[0055] #Image processing and target feature extraction

[0056] processed_data=image_preprocessing(image_data)

[0057] # Divide into training set and test set

[0058] X_train,X_test,y_train,y_test=train_test_split(processed_data,labels,test_size=0.2,random_state=42)

[0059] #Training random forest classifier

[0060] rf_classifier=RandomForestClassifier()

[0061] rf_classifier.fit(X_train,y_train)

[0062] #Predict the category of the test set image

[0063] predicted_labels=rf_classifier.predict(X_test)

[0064] #Return the prediction result

[0065] return predicted_labels

[0066] #Using sensor data for disaster prediction

[0067] defdisaster_prediction(sensor_data):

[0068] #Data preprocessing and feature extraction

[0069] processed_data=preprocess_sensor_data(sensor_data)

[0070] # Load the 3D site model

[0071] model=tf.keras.models.load_model('disaster_prediction_model.h5')

[0072] #Predicted disaster type

[0073] predictions=model.predict(processed_data)

[0074] #Return emergency response suggestions based on the prediction results

[0075] emergency_response=get_emergency_response(predictions)

[0076] return emergency_response

[0077] #Use data mining to classify and predict historical disaster data

[0078] defdata_mining(data):

[0079] #Data cleaning process

[0080] cleaned_data=clean_data(data)

[0081] #Feature extraction and data conversion

[0082] transformed_data=transform_data(cleaned_data)

[0083] #Train disaster type classification model

[0084] model=train_model(transformed_data)

[0085] # Make predictions

[0086] predictions=model.predict(transformed_data)

[0087] #Return the prediction result

[0088] return predictions

[0089] #Main function

[0090] defmain():

[0091] #Get remote sensing image data

[0092] image_data,labels=get_remote_sensing_data()

[0093] #Use remote sensing images for image prediction

[0094] classified_labels=image_classification(image_data,labels)

[0095] #Get sensor data

[0096] sensor_data=get_sensor_data()

[0097] #Using sensor data for disaster prediction

[0098] emergency_response=disaster_prediction(sensor_data)

[0099] #Get historical disaster data

[0100] historical_data=get_historical_data()

[0101] #Use data mining to classify and predict historical disaster data

[0102] predictions=data_mining(historical_data)

[0103] # Output prediction results

[0104] print("Image prediction results:", classified_labels)

[0105] print("Disaster prediction emergency response suggestions:", emergency_response)

[0106] print("Historical disaster data classification prediction results:",predictions)

[0107] if __name__ == '__main__':

[0108] main().

[0109] The IoT intelligent disaster warning method further includes building a machine learning model and performing deep learning training before performing S2, and the specific operations are as follows:

[0110] A. Collect image data of 3D sites and urban agglomerations and build sample datasets for model training;

[0111] Data preparation: collect image data of three-dimensional sites and urban agglomerations to establish sample data sets;

[0112] Data labeling: label or annotate the image data in the sample data set, such as roads, buildings, crowded areas, etc. The annotation information uses a frame or mask to represent the location and shape of the target;

[0113] Data processing: scaling, normalization, and data enhancement are performed on each image in the sample data set to match the image data with the characteristics of the three-dimensional site and urban agglomeration;

[0114] B. Build machine learning models;

[0115] In a specific implementation, the machine learning model adopts the Faster R-CNN model, which is mainly composed of two main parts, including a region proposal network (RPN) and an object recognition network (Object Detection Network). The RPN is used to extract image features to generate candidate target regions; the object recognition network is used to classify and regress candidate target regions to determine the location and category of the final target.

[0116] C. Perform deep learning training on the machine learning model through the sample data set, determine the loss function, optimizer and learning rate parameters of the machine learning model, evaluate and tune the machine learning model, and obtain a trained machine learning model;

[0117] Divide the sample data set into training set, test set and validation set, and input the labeled or annotated image data in the training set into the machine learning model for training;

[0118] The test set is input into the machine learning model for testing, the accuracy, recall and precision of the model are calculated, and the loss function, optimizer and learning rate parameters of the machine learning model are adjusted to improve the accuracy and robustness of the machine learning model;

[0119] Evaluate the performance and effect of machine learning models and optimize parameters of deep learning models;

[0120] The unlabeled or annotated image data is input into the trained machine learning model through the validation set for target detection, and the output target detection data is used for image data verification to obtain a trained machine learning model.

[0121] At the same time, the Faster R-CNN model training process is time-consuming and occupies a lot of computer resources. Therefore, the Faster R-CNN model is trained in a distributed manner by building a deep learning framework TensorFlow.

[0122] The operations of constructing the deep learning framework TensorFlow for distributed training are specifically as follows:

[0123] Prepare the pre-trained Faster R-CNN model and load it in the deep learning framework TensorFlow;

[0124] Set the cluster configuration for distributed training, including setting the master node and worker node information, communication protocol, and port;

[0125] Select a distributed training strategy based on the cluster configuration for distributed training;

[0126] Prepare the dataset for training and use data parallel strategy to perform distributed training on Faster R-CNN model;

[0127] The accuracy and loss value of the model during the training process are collected, and the Faster R-CNN model is evaluated to obtain the trained Faster R-CNN model.

[0128] This application deploys the trained machine learning model for S2 operation, uses the machine learning model Faster R-CNN and the deep learning framework TensorFlow to perform feature recognition on the image data of the remote sensing image set, radar data set, and drone image set, extracts terrain features, and establishes a terrain feature database, specifically:

[0129] S2.1. Import remote sensing image sets, radar datasets and drone image sets respectively;

[0130] S2.2, process remote sensing image sets to extract ground feature information; process radar data sets to generate point cloud data and extract terrain information; process drone image sets to extract building outlines and crowded areas;

[0131] S2.3. The extracted ground feature information, terrain information, building outlines and crowded areas are integrated with multi-source data to establish a terrain feature database. Specifically:

[0132] Collect data from different data sources such as ground feature information, terrain information, building outlines, and crowded areas, and clean, process, and format the data to ensure data consistency;

[0133] Correlate and match data information from different data sources to ensure the relevance between data;

[0134] By extracting the respective information features from different data sources, such as extracting the type of landforms from landform information and extracting elevation information from terrain information, the elevation information and building outlines associated with the landform types are integrated based on the relevance of data information to achieve feature fusion of multi-source data;

[0135] The fused data are verified and corrected, and finally a terrain feature database is established.

[0136] Sample Code

[0137] #Import the required data sources

[0138] import numpy as np

[0139] import cv2#for image processing

[0140] import open3d as o3d#for point cloud processing and 3D visualization

[0141] #Image data processing and feature extraction

[0142] defextract_features_terrain(data):

[0143] #Data sources include remote sensing image sets, radar data sets, and drone image sets

[0144] #Processing remote sensing image set

[0145] satellite_image=data['satellite_image']

[0146] #Use image processing to preprocess remote sensing image data

[0147] processed_image=preprocess_satellite_image(satellite_image)

[0148] #Extract ground feature information, such as roads, buildings, etc.

[0149] features_from_satellite=extract_features_from_image(processed_image)

[0150] #Process the lidar data of the radar dataset

[0151] lidar_data=data['lidar_data']

[0152] #Process the lidar data

[0153] processed_lidar_data=preprocess_lidar_data(lidar_data)

[0154] #Generate point cloud based on lidar data

[0155] point_cloud_from_lidar=generate_point_cloud(processed_lidar_data)

[0156] #Extract terrain information, such as ground elevation, etc.

[0157] terrain_features_from_lidar=extract_terrain_features(point_cloud_from_lidar)

[0158] #Processing drone imagery

[0159] drone_image=data['drone_image']

[0160] #Use image processing to preprocess drone image data

[0161] processed_drone_image=preprocess_drone_image(drone_image)

[0162] #Extract features, such as building outlines, crowded areas, etc.

[0163] features_from_drone=extract_features_from_image(processed_drone_image)

[0164] #Fuse the features of multi-source data

[0165] fused_features=fuse_features(features_from_satellite,terrain_features_from_lidar,features_from_drone)

[0166] #Return the fused features

[0167] return fused_features.

[0168] The machine learning model Faster R-CNN and the deep learning framework TensorFlow are used to identify and process the geological structure data set to obtain the geological structure characteristics of the terrain area to be monitored, such as faults, lithology and other attributes, and establish a geological structure characteristic database.

[0169] A three-dimensional site model is constructed using a rapid modeling algorithm based on a terrain feature database and a geological structure feature database. At the same time, the S4 also includes real-time acquisition of remote sensing images, laser radar, and image data taken by drones, so as to extract new features to update the three-dimensional site model to adapt to the dynamic changes of the terrain area to be monitored;

[0170] Sample Code

[0171] #Dynamic Update

[0172] defupdate_model(realtime_data,current_model):

[0173] #Real-time data collection

[0174] #Process real-time data, such as obtaining the latest remote sensing image data, lidar data, and drone image data

[0175] new_data=process_realtime_data(realtime_data)

[0176] #Extract features of new data

[0177] new_features=extract_features_terrain(new_data)

[0178] #Update the new features with the current model

[0179] updated_model=update_features(current_model,new_features)

[0180] # Return the updated model

[0181] return updated_model.

[0182] In addition, the present invention also provides a disaster emergency warning system based on Internet of Things technology, including:

[0183] The acquisition module is used to collect image data of the terrain area to be monitored through remote sensing images, laser radar and drone photography, and to establish corresponding remote sensing image sets, radar data sets and drone image sets respectively; and to collect geological structure physical data of the terrain area to be monitored and establish a geological structure data set;

[0184] The recognition and extraction module is used to use machine learning and deep learning technologies to perform feature recognition on the image data of remote sensing image sets, radar data sets, and drone image sets, extract terrain features, and establish a terrain feature database;

[0185] The identification and extraction module is also used to use machine learning and deep learning technology to identify and process the geological structure data set, obtain the geological structure characteristics of the terrain area to be monitored, and establish a geological structure characteristic database;

[0186] A three-dimensional site construction module is used to construct a three-dimensional site model of the terrain area to be monitored based on the terrain feature database and the geological structure feature database;

[0187] The monitoring and early warning module is used to obtain sensor data of the terrain area to be monitored based on the Internet of Things technology, conduct visual interaction with the three-dimensional site model, and establish a disaster emergency warning platform; it provides disaster warnings in the terrain area to be monitored by acquiring remote sensing images, sensor collection data, and historical disaster data in real time.

[0188] Secondly, a disaster emergency warning device based on Internet of Things technology is also provided, comprising a processor and a memory, wherein the processor implements the disaster emergency warning method based on Internet of Things technology as described above when executing program data stored in the memory.

[0189] Finally, a readable storage medium is provided for storing control program data, wherein the control program data, when executed by a processor, implements the disaster emergency warning method based on the Internet of Things technology as described above.

Claims

1. A disaster emergency warning method based on Internet of Things technology, characterized in that: include: S1. Collect image data of the terrain area to be monitored through remote sensing images, laser radar and drone photography, and establish corresponding remote sensing image sets, radar data sets and drone image sets respectively; and collect geological structure physical data of the terrain area to be monitored to establish a geological structure data set; S2. Build a machine learning model and conduct deep learning training, collect image data of three-dimensional sites and urban agglomerations, and establish a sample data set for model training; build a machine learning model; conduct deep learning training on the machine learning model through the sample data set, determine the loss function, optimizer and learning rate parameters of the machine learning model, evaluate and tune the machine learning model, and obtain a trained machine learning model; The machine learning model adopts the Faster R-CNN model, including a region extraction network and a target recognition network. The region extraction network is used to extract image features to generate candidate target regions; the target recognition network is used to classify and regress the candidate target regions to determine the location and category of the final target; The Faster R-CNN model training also includes distributed training by building a deep learning framework TensorFlow, specifically: Prepare the pre-trained Faster R-CNN model and load it in the deep learning framework TensorFlow; Set the cluster configuration for distributed training, including setting the master node and worker node information, communication protocol, and port; Select a distributed training strategy based on the cluster configuration for distributed training; Prepare the dataset for training and use data parallel strategy to perform distributed training on Faster R-CNN model; Collect the accuracy and loss value of the model during the training process, evaluate the Faster R-CNN model, and obtain the trained Faster R-CNN model; Using deep learning technology, we perform feature recognition on the image data of remote sensing image sets, radar data sets, and drone image sets, extract terrain features, and establish a terrain feature database; S2.

1. Import remote sensing image sets, radar datasets and drone image sets respectively; S2.2, process remote sensing image sets to extract ground feature information; process radar data sets to generate point cloud data and extract terrain information; process drone image sets to extract building outlines and crowded areas; S2.3, integrate the extracted ground feature information, terrain information, building outlines and crowded areas, and establish a terrain feature database; Collect data from different data sources including ground feature information, terrain information, building outlines and crowded areas, and clean, process and format the data from different data sources to ensure data consistency; Correlate and match data information from different data sources; By extracting the respective information features from different data sources and integrating the elevation information and building outlines associated with the land object types based on the relevance of data information, the feature fusion of multi-source data is achieved; Verify and correct the fused data, and finally establish a terrain feature database; S3. Use deep learning technology to identify and process geological structure data sets, obtain geological structure characteristics of the terrain area to be monitored, and establish a geological structure characteristic database; S4. Construct a three-dimensional site model of the terrain area to be monitored based on the terrain feature database and the geological structure feature database; S5. Based on the Internet of Things technology, obtain sensor data of the terrain area to be monitored, interact with the three-dimensional site model visually, and establish a disaster emergency warning platform; obtain remote sensing images, sensor data, and historical disaster data in real time to conduct disaster warning in the terrain area to be monitored, specifically: Through remote sensing technology, high-resolution, multi-angle, and multi-temporal remote sensing image data are collected from the monitoring area, and target features are extracted through image processing to determine the remote sensing image prediction results; The collected data of the sensor is obtained through the Internet of Things technology, and the collected data is imported into the three-dimensional site model for simulation to obtain the simulation prediction results of the sensor data; Conduct data mining on historical disaster data, establish a disaster type classification model based on historical disaster data, and determine data mining prediction results based on sensor data; The disaster prediction results of the terrain area to be monitored are determined through remote sensing image prediction results, sensor data simulation prediction results and data mining prediction results. According to the disaster prediction results, the disaster emergency warning platform issues a disaster alarm.

2. The disaster emergency warning method according to claim 1, characterized in that: The specific method of collecting the image data of the three-dimensional site and the urban agglomeration and establishing the sample data set for model training is: Data preparation: collect image data of three-dimensional sites and urban agglomerations to establish sample data sets; Data labeling: labeling or annotating the image data in the sample data set, wherein the annotation uses a frame or mask to indicate the location and shape of the target; Data processing: Perform data scaling, normalization, and data enhancement on each image in the sample data set to match the image data with the characteristics of the three-dimensional sites and urban agglomerations.

3. The disaster emergency warning method according to claim 1, characterized in that: It also includes real-time collection of remote sensing images, lidar and image data taken by drones, and the extraction of new features to update the three-dimensional site model.

4. A disaster emergency warning system based on Internet of Things technology, characterized in that: include: The acquisition module is used to collect image data of the terrain area to be monitored through remote sensing images, laser radar and drone photography, and to establish corresponding remote sensing image sets, radar data sets and drone image sets respectively; and to collect geological structure physical data of the terrain area to be monitored and establish a geological structure data set; The recognition and extraction module is used to build a machine learning model and perform deep learning training, collect image data of three-dimensional sites and urban agglomerations, and establish a sample data set for model training; build a machine learning model; perform deep learning training on the machine learning model through the sample data set, determine the loss function, optimizer and learning rate parameters of the machine learning model, evaluate and tune the machine learning model, and obtain a trained machine learning model; the machine learning model adopts the Faster R-CNN model, including a region extraction network and a target recognition network, and the region extraction network is used to extract image features to generate candidate target regions; the target recognition network is used to classify and regress candidate target regions to determine the location and category of the final target; The Faster R-CNN model training also includes distributed training by building a deep learning framework TensorFlow, specifically: Prepare the pre-trained Faster R-CNN model and load it in the deep learning framework TensorFlow; Set the cluster configuration for distributed training, including setting the master node and worker node information, communication protocol, and port; Select a distributed training strategy based on the cluster configuration for distributed training; Prepare the dataset for training and use data parallel strategy to perform distributed training on Faster R-CNN model; Collect the accuracy and loss value of the model during the training process, evaluate the Faster R-CNN model, and obtain the trained Faster R-CNN model; Using machine learning and deep learning technologies, we perform feature recognition on remote sensing image sets, radar data sets, and drone image sets, extract terrain features, and establish a terrain feature database. Import remote sensing image sets, radar datasets, and drone image sets respectively; Process remote sensing image sets to extract ground feature information; process radar data sets to generate point cloud data and extract terrain information; process drone image sets to extract building outlines and crowded areas; The extracted ground feature information, terrain information, building outlines and crowded areas are integrated with multi-source data to establish a terrain feature database. Collect data from different data sources including ground feature information, terrain information, building outlines and crowded areas, and clean, process and format the data from different data sources to ensure data consistency; Correlate and match data information from different data sources; extract information features from different data sources, integrate elevation information and building outlines associated with land feature types based on the correlation of data information, and achieve feature fusion of multi-source data; verify and correct the fused data, and finally establish a terrain feature database; The identification and extraction module is also used to use machine learning and deep learning technology to identify and process the geological structure data set, obtain the geological structure characteristics of the terrain area to be monitored, and establish a geological structure characteristic database; A three-dimensional site construction module is used to construct a three-dimensional site model of the terrain area to be monitored based on the terrain feature database and the geological structure feature database; The monitoring and early warning module is used to obtain sensor data of the terrain area to be monitored based on the Internet of Things technology, conduct visual interaction with the three-dimensional site model, and establish a disaster emergency warning platform; it provides disaster warning for the terrain area to be monitored by acquiring remote sensing images, sensor data, and historical disaster data in real time, specifically: Through remote sensing technology, high-resolution, multi-angle, and multi-temporal remote sensing image data are collected from the monitoring area, and target features are extracted through image processing to determine the remote sensing image prediction results; The collected data of the sensor is obtained through the Internet of Things technology, and the collected data is imported into the three-dimensional site model for simulation to obtain the simulation prediction results of the sensor data; Conduct data mining on historical disaster data, establish a disaster type classification model based on historical disaster data, and determine data mining prediction results based on sensor data; The disaster prediction results of the terrain area to be monitored are determined through remote sensing image prediction results, sensor data simulation prediction results and data mining prediction results. According to the disaster prediction results, the disaster emergency warning platform issues a disaster alarm.

5. A disaster emergency warning device based on Internet of Things technology, characterized in that: It comprises a processor and a memory, wherein when the processor executes the program data stored in the memory, it implements the disaster emergency warning method based on the Internet of Things technology as described in any one of claims 1-3.

6. A readable storage medium, characterized in that: Used to store control program data, wherein the control program data, when executed by a processor, implements the disaster emergency warning method based on Internet of Things technology as described in any one of claims 1-3.

Images