Environmental ground disaster emergency early warning method, device and equipment and readable storage medium

Abstract

The invention discloses an environmental disaster emergency early warning method, device and equipment and a readable storage medium. The method comprises the following steps: acquiring a historical disaster cause map and a real-time disaster cause map of different types of disaster cause features; segmenting the historical geological disaster cause map, marking the historical segmented image according to the geological disaster cause features, and constructing different types of data sets; training a preset deep convolutional neural network model to obtain different types of feature models with different model weights; inputting the real-time segmented image into a feature model of a corresponding type; fusing the outputs of different types of feature models to obtain a fused image; restoring the fused image according to the position information; performing expansion processing on the feature confidence map; carrying out region merging processing on different types of feature region maps; and according to a preset dangerous case label criterion and the dangerous case feature map, performing judgment to obtain ground disaster dangerous case information. The method is high in early warning efficiency, low in labor cost and capable of achieving real-time early warning.

Description

Technical Field

[0001] The present invention relates to the field of geological disaster emergency warning technology, and in particular to an environmental geological disaster emergency warning method, device, equipment and readable storage medium. Background Art

[0002] With the influence of unstable climate in recent years, the geological environment has been deteriorating, and geological disasters have also increased. Faced with the increasingly severe threat of geological disasters, early manual inspections or traditional early warning methods for geological disaster investigations not only require a lot of manpower costs, but also have problems such as high latency and easy omission of dangerous information, making it difficult to report dangerous situations in a timely and accurate manner, arrange emergency evacuation, and protect people's lives and property. At present, deep learning technology has matured in image recognition and detection. Applying deep learning technology to the scene of geological disaster warning can reduce the complexity of the system design of geological disaster warning. At the same time, through the distributed processing structure, the scope of geological disaster monitoring is wider, the monitoring is more comprehensive, and the timeliness of warning is greatly improved.

[0003] At present, mainstream geological disaster early warning technologies are based on traditional methods, such as geomechanics principles and statistical analysis. Although they can achieve certain results, traditional methods are difficult to model and complex, and require high manpower costs to analyze and process a large amount of data. In the early warning stage, the early warning efficiency is low and the speed is slow. At the same time, the traditional hierarchical early warning method has certain lags and other problems.

[0004] Therefore, an environmental disaster emergency warning method, device, equipment and readable storage medium are developed to solve the above problems. Summary of the invention

[0005] The present invention proposes an environmental disaster emergency warning method, device, equipment and readable storage medium to solve the problems of low warning efficiency, high labor cost and delayed warning in existing warning methods.

[0006] The present invention achieves the above-mentioned purpose through the following technical solutions: The present invention provides an environmental disaster emergency warning method, comprising: Obtain historical and real-time geological disaster cause maps of different types of geological disaster cause characteristics; Segmenting the historical geological disaster cause map to obtain historical segmented images, annotating the historical segmented images according to the geological disaster cause characteristics to obtain historical sample labels, and constructing different types of data sets according to the historical sample labels and the historical segmented images; The preset deep convolutional neural network models are trained respectively by using different types of data sets to obtain feature models of different types and with different model weights; Segmenting the real-time geological disaster cause map to obtain real-time segmented images, recording the position information of each real-time segmented image, inputting the real-time segmented image into the corresponding feature model, and outputting the confidence label of the real-time segmented image corresponding to the geological disaster cause feature; fusing the outputs of the feature models of different types to obtain a fused image; Restoring the fused image according to the position information to obtain different types of feature confidence maps; Performing expansion processing on the feature confidence map to obtain different types of feature region maps; Performing region merging processing on the characteristic region maps of different types, and generating new confidence labels for the merged regions respectively, to obtain a dangerous situation characteristic map; The geological disaster hazard information is obtained according to the preset hazard label criteria and the hazard characteristic map.

[0007] Furthermore, historical geological disaster cause maps and real-time geological disaster cause maps of different types of geological disaster cause characteristics are obtained, including: Collect historical and real-time geological disaster cause data; Performing statistical integration on the historical geological disaster cause data to generate a historical geological disaster cause map that can characterize the characteristics of geological disasters under the same resolution and scale conditions; The real-time geological disaster cause data are statistically integrated and statistically analyzed to generate a real-time geological disaster cause map that can characterize the characteristics of the geological disaster under the conditions of the same resolution and the same scale.

[0008] Furthermore, the confidence labels include precipitation, wind speed, vegetation coverage and soil conditions, and the confidence value range of each confidence label is [0, 1].

[0009] Furthermore, the real-time segmented image is input into the feature model of the corresponding type, and a predicted image is output, including: Using convolution kernels of three scales of 1*1, 3*3, and 5*5 and 64 channels to extract features from the real-time segmented image, to obtain a first feature matrix; Then, the first feature matrix is ​​superimposed, max pooled, and batch normalized to obtain a second feature matrix; the second feature matrix is ​​subjected to feature extraction using convolution kernels of three scales of 1*1, 3*3, and 5*5 and a channel number of 128 to obtain a third feature matrix; Performing superposition, maximum pooling and BN operations on the third feature matrix to obtain a fourth feature matrix; The fourth feature matrix is ​​subjected to feature extraction using convolution kernels of three scales, 1*1, 3*3, and 5*5, and with a channel number of 256, to obtain a fifth feature matrix; Performing superposition, maximum pooling and BN operations on the fifth feature matrix to complete feature extraction, thereby obtaining a sixth feature matrix; Perform feature extraction on the sixth feature matrix using a convolution kernel with a scale of 3*3 and a channel number of 512 to obtain a seventh feature matrix; Performing maximum pooling and BN operations on the seventh feature matrix to extract features, to obtain an eighth feature matrix; Perform feature extraction on the eighth feature matrix using a 3*3 convolution kernel with a channel number of 1024 to obtain a ninth feature matrix; Perform maximum pooling and BN operations on the ninth feature matrix to extract features, and obtain the tenth feature matrix; The tenth feature matrix is ​​subjected to feature extraction using a 3*3 convolution kernel with a channel number of 2048, and then the feature matrix is ​​subjected to maximum pooling and BN operations to obtain a final feature matrix; Finally, the final feature matrix is ​​flattened and fully connected to a feature vector with a length of 1, and nonlinearly processed using a Sigmoid function to obtain the predicted image.

[0010] Further, the characteristic area maps of different types are subjected to regional merging processing, and new confidence labels are respectively generated for the merged areas according to the characteristics of the causes of different types of geological disasters to obtain dangerous situation characteristic maps, including: the characteristic confidence map is expanded to obtain different types of characteristic area maps, including: Setting a sliding window on the feature confidence map based on the neighbor mean principle; The confidence average boundary dilation processing is performed on the center position of the sliding window to obtain the feature area map.

[0011] Furthermore, it also includes: Storing the geological disaster risk information; Distributing the geological disaster risk information to different types of terminal devices according to a distributed system, and outputting corresponding risk alerts through the terminal devices, the terminal devices including command and management terminals, public equipment terminals and personal equipment terminals; The danger alarm is sent to the corresponding danger response terminal set through different types of terminal devices. The danger response terminal set includes a command and management danger response terminal set, a public equipment danger response terminal set, and a personal equipment danger response terminal set. The command and management danger response terminal set includes the emergency command center large screen and the danger management data center, the public equipment danger response terminal set includes broadcasting, access control and alarms, and the personal equipment danger response terminal set includes computers, collection and bracelets.

[0012] The present invention also provides an environmental disaster emergency warning device, comprising: An acquisition module, the acquisition module is used to acquire historical geological disaster cause maps and real-time geological disaster cause maps of different types of geological disaster cause characteristics; A construction module, wherein the construction module is used to segment the historical geological disaster cause map to obtain historical segmented images, annotate the historical segmented images according to the geological disaster cause characteristics to obtain historical sample labels, and construct different types of data sets according to the historical sample labels and the historical segmented images; A training module, wherein the training module is used to train a preset deep convolutional neural network model using different types of data sets to obtain feature models of different types and with different model weights; A prediction module, the prediction module is used to segment the real-time geological disaster cause map to obtain real-time segmented images, record the position information of each real-time segmented image, input the real-time segmented image into the corresponding feature model, and output a predicted image, wherein the predicted image has a confidence label of the real-time segmented image corresponding to the geological disaster cause feature; A fusion module, the fusion module is used to fuse the outputs of the feature models of different types to obtain a fused image; A restoration module, the restoration module is used to restore the fused image according to the position information to obtain different types of feature confidence maps; An expansion module, the expansion module is used to expand the feature confidence map to obtain different types of feature region maps; A merging module, the merging module is used to perform region merging processing on the characteristic region maps of different types, and generate new confidence labels for the merged regions respectively, to obtain a dangerous situation characteristic map; A judgment module is used to judge and obtain geological disaster hazard information according to a preset hazard label criterion and the hazard characteristic map.

[0013] The present invention also provides an environmental disaster emergency warning device, comprising: Memory for storing computer programs; A processor is used to implement the steps of the environmental geological disaster emergency warning method when executing the computer program.

[0014] The present invention also provides a readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the environmental geological disaster emergency warning are implemented.

[0015] The beneficial effects of the present invention are: The environmental geological disaster emergency warning method and device proposed in the present invention solve the problems of low warning efficiency, high labor cost and delayed warning in existing warning methods. It obtains geological disaster cause data, combines deep learning technology to analyze, model and monitor environmental geological disaster data, and realizes real-time geological disaster warning. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A method flow chart of an environmental disaster emergency warning method in an embodiment of the present invention; Figure 2 A schematic diagram of a prediction process in an embodiment of the present invention; Figure 3 A schematic diagram of constructing a data set in an embodiment of the present invention; Figure 4 Schematic diagram of the structure of a deep convolutional neural network in an embodiment of the present invention; Figure 5 It is a schematic diagram of the feature processing and feature area merging process in an embodiment of the present invention; Figure 6 This is an example diagram of feature fusion representation in an embodiment of the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0018] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0019] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0020] In the description of the present invention, it should be understood that the terms "upper", "lower", "inside", "outside", "left", "right", etc. indicate directions or positional relationships based on the directions or positional relationships shown in the accompanying drawings, or are directions or positional relationships in which the product of the invention is usually placed when in use, or are directions or positional relationships commonly understood by those skilled in the art. These directions or positional relationships are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as a limitation on the present invention.

[0021] Furthermore, the terms “first”, “second”, etc. are merely used for distinguishing descriptions and should not be understood as indicating or implying relative importance.

[0022] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms such as "setting" and "connection" should be understood in a broad sense. For example, "connection" can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or an indirect connection through an intermediate medium, or it can be the internal communication of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0023] The specific implementation modes of the present invention are described in detail below in conjunction with the accompanying drawings.

[0024] The embodiment of the present invention selects four causes of geological disasters: rainfall, wind, vegetation and soil for monitoring and data collection as examples for relevant explanation, including but not limited to using the causes of geological disasters as examples.

[0025] The present invention discloses an environmental disaster emergency warning method, comprising the following steps: Obtain historical and real-time geological disaster cause maps of different types of geological disaster cause characteristics; Specifically, firstly, in the area that needs to be monitored, according to the monitoring range of the equipment, equipment with different causes of geological disasters are deployed to implement monitoring and collect relevant data on the causes of geological disasters; the collected data on the causes of geological disasters in various places are integrated, and statistics are generated to generate k types of geological disaster cause maps that can characterize the characteristics of geological disasters under the conditions of the same resolution and scale.

[0026] Segmenting the historical geological disaster cause map to obtain historical segmented images, annotating the historical segmented images according to the geological disaster cause characteristics to obtain historical sample labels, and constructing different types of data sets according to the historical sample labels and the historical segmented images; Specifically, the images are divided into w*h sizes according to the same scale requirements, the factors affecting geological disasters in each segmented image are marked, and the source image to which each segmented image belongs and its row and column position are recorded. Four different datasets representing geological disasters are produced based on the labeled images.

[0027] The preset deep convolutional neural network models are trained respectively by using different types of data sets to obtain different types of feature models; Segmenting the real-time geological disaster cause map to obtain real-time segmented images, recording the position information of each real-time segmented image, inputting the real-time segmented image into the corresponding feature model, and outputting a predicted image, wherein the predicted image has a confidence label of the real-time segmented image corresponding to the geological disaster cause feature; Fusing the predicted images of the feature models of different types to obtain a fused image; Restoring the fused image according to the position information to obtain different types of feature confidence maps; Performing expansion processing on the feature confidence map to obtain different types of feature region maps; Specifically, the predicted images of each feature model are fused, and each type of image is restored according to the position information of the segmented image to obtain the feature region distribution set of the source image. Each feature set is expanded using image processing technology to obtain the feature region map of each type of image.

[0028] Performing region merging processing on the characteristic region maps of different types, and generating new confidence labels for the merged regions respectively, to obtain a dangerous situation characteristic map; The geological disaster hazard information is obtained according to the preset hazard label criteria and the hazard characteristic map.

[0029] Specifically, the characteristic regional maps of various types of geological disasters are processed by regional intersection and merging, and four-dimensional new labels <precipitation: a, wind force: b, vegetation coverage: c, soil condition: d> are generated for each processed area (where a, b, c, d are the confidence levels of each label). The fusion results are judged for danger according to the designed labeling criteria to generate basic danger results. Based on the basic danger results and confidence levels, the severity of the geological disaster danger is divided to generate danger information.

[0030] Storing the geological disaster risk information; Distributing the geological disaster risk information to different types of terminal devices according to a distributed system, and outputting corresponding risk alerts through the terminal devices, the terminal devices including command and management terminals, public equipment terminals and personal equipment terminals; The danger alarm is sent to the corresponding danger response terminal set through different types of terminal devices. The danger response terminal set includes a command and management danger response terminal set, a public equipment danger response terminal set, and a personal equipment danger response terminal set. The command and management danger response terminal set includes the emergency command center large screen and the danger management data center, the public equipment danger response terminal set includes broadcasting, access control and alarms, and the personal equipment danger response terminal set includes computers, collection and bracelets.

[0031] like Figure 1 As shown, it is a schematic diagram of the method flow of the present invention. Within the scope of geological disaster monitoring, the geological disaster cause data collection and monitoring terminal equipment concerned by the research institute is established and deployed, and the collected data is divided into real-time monitoring data and historical data. Among them, the historically stored source data is trained with a deep learning model, and the real-time monitoring data is input into the trained prediction network model through a data preprocessing process to predict geological disaster characteristics and generate dangerous information. Through the dangerous information correction and distributed system, the early warning data is stored and the dangerous information is sent to various terminal devices in a timely and targeted manner for specialized early warning and response.

[0032] like Figure 2 The figure shows the prediction process of the geological disaster feature warning model. In this process, the real-time geological disaster data is segmented and normalized, and then predicted through the network model and weight file. The weight file is the model weight. The generated feature information is restored and regionalized according to the segmentation position to generate the geological disaster feature confidence map. These confidence maps represent the impact range of different causes of geological disasters. Then, the regional fusion of different features and the designed dangerous situation label criterion are combined to realize the dangerous situation information monitoring. Among them, the design of the label criterion is determined by the fusion label of each segmented image in this example: <Precipitation: a, Wind speed: b, Vegetation coverage: c, Soil quality: d> Among them, a, b, c, d are values ​​of [0, 1]. The more extreme the value, the higher the confidence, that is, the more obvious the corresponding representation situation. In this implementation case, the characteristics of geological disasters are combined, including but not limited to designing and listing the following geological disaster situations. The preset hazard label criteria are shown in Table 1 below: Table 1

[0033] The prediction process of the present invention needs to be based on a deep learning convolutional neural network and a weight file, and the weight file is obtained by inputting a customized data set into a deep convolutional neural network for training; wherein the customized data set is as follows Figure 3As shown, it is necessary to monitor and collect data on the causes of geological disasters, generate geological disaster cause images through statistical analysis, segment the geological disaster cause images at a specified scale, annotate the segmented image samples based on whether they contain geological disaster cause features and make sample labels, and finally organize the sample label files and segmented image samples into a data set.

[0034] Figure 4 This is a schematic diagram of the deep convolutional neural network structure designed in the present invention. Figure 4 In it, input represents input, conv represents convolutional layer, block represents module, contat represents concatenation operation, maxpooling represents maximum pooling layer, BN represents batch normalization layer, Flatten represents flattening layer, Dense represents fully connected layer, sigmoid represents sigmoid activation function, 2class represents binary classification output, and leakyReLu is a variant of the activation function. The deep convolutional neural network first uses convolution kernels with three scales of 1*1, 3*3, and 5*5 and 64 channels to extract features from the input image matrix, and then performs superposition, maximum pooling, and BN (batch normalization) operations on the feature matrix to complete feature extraction; uses convolution kernels with three scales of 1*1, 3*3, and 5*5 and 128 channels to extract features from the obtained feature matrix, and then performs superposition, maximum pooling, and BN (batch normalization) operations to complete feature extraction; continues to use convolution kernels with three scales of 1*1, 3*3, and 5*5 and 256 channels to extract features from the obtained feature matrix, and then performs superposition, maximum pooling, and BN (batch normalization) operations to complete feature extraction; uses 3* 3. The convolution kernel with the number of channels of 512 is used for feature extraction, and then the feature matrix is ​​subjected to the maximum pooling and BN (batch normalization) operation for feature extraction; the obtained feature matrix is ​​subjected to the convolution kernel with the number of channels of 3*3 and 1024 for feature extraction, and then the feature matrix is ​​subjected to the maximum pooling and BN (batch normalization) operation for feature extraction; the obtained feature matrix is ​​subjected to the convolution kernel with the number of channels of 3*3 and 2048 for feature extraction, and then the feature matrix is ​​subjected to the maximum pooling and BN (batch normalization) operation for feature extraction; finally, the feature matrix is ​​flattened and fully connected to the feature vector with a length of 1, and the Sigmoid function is used for nonlinear processing, and the category label is output as required. The Sigmoid function is the activation function in the neural network. In order to prevent the "cold start" problem, the LeakyReLu activation function is selected for nonlinear transformation after each convolution operation.

[0035] like Figure 5As shown in the figure, the results of model prediction are processed with features and merged with feature regions. First, the source image data is spliced ​​and restored according to the predicted image of the segmented image and the location information of the segmentation to generate an effective feature confidence map. In order to smooth the confidence map boundary, the neighbor mean principle is used. A k*k sliding window is set, and the confidence mean boundary expansion is performed on the center position of the sliding window to obtain the feature region map. Next, the four representation confidence maps are superimposed, and the label of each segmented image unit is reconstructed and feature fused to generate a dangerous situation feature map. Compared with the original source image data, each segmented area of ​​size w*h contains the confidence of each local disaster representation, that is, the confidence of the geological disaster representation, which is in the form of <precipitation: a, wind force: b, vegetation coverage: c, soil condition: d> (where a, b, c, d are the confidences of each label).

[0036] like Figure 6 As shown in the figure, it is an example of a dangerous feature map. The positions of the four types of feature area maps are mapped one by one, and the confidence labels are reconstructed in multiple dimensions. For example, the gray segmented area in the figure is reconstructed into the confidence label of the orange area, <0.7, 0.2, 0.7, 0.1>, and the corresponding attributes are attached: <precipitation: 0.7, wind force: 0.2, soil condition: 0.7, vegetation coverage: 0.1>. Combined with the preset label criteria, it is judged that the vegetation coverage rate in this area is low, the precipitation is large, and the soil humidity is high. According to the geological disaster situation designed above, landslides or mudslides are very likely to occur.

[0037] The present invention also provides an environmental disaster emergency warning device, comprising: An acquisition module, the acquisition module is used to acquire historical geological disaster cause maps and real-time geological disaster cause maps of different types of geological disaster cause characteristics; A construction module, wherein the construction module is used to segment the historical geological disaster cause map to obtain historical segmented images, annotate the historical segmented images according to the geological disaster cause characteristics to obtain historical sample labels, and construct different types of data sets according to the historical sample labels and the historical segmented images; A training module, wherein the training module is used to train a preset deep convolutional neural network model using different types of data sets to obtain feature models of different types and with different model weights; A prediction module, the prediction module is used to segment the real-time geological disaster cause map to obtain real-time segmented images, record the position information of each real-time segmented image, input the real-time segmented image into the corresponding feature model, and output a predicted image, wherein the predicted image has a confidence label of the real-time segmented image corresponding to the geological disaster cause feature; A fusion module, the fusion module is used to fuse the outputs of the feature models of different types to obtain a fused image; A restoration module, the restoration module is used to restore the fused image according to the position information to obtain different types of feature confidence maps; An expansion module, the expansion module is used to expand the feature confidence map to obtain different types of feature region maps; A merging module, the merging module is used to perform region merging processing on the characteristic region maps of different types, and generate new confidence labels for the merged regions respectively, to obtain a dangerous situation characteristic map; A judgment module is used to judge and obtain geological disaster hazard information according to a preset hazard label criterion and the hazard characteristic map.

[0038] The present invention also provides an environmental disaster emergency warning device, comprising: Memory for storing computer programs; A processor is used to implement the steps of the environmental geological disaster emergency warning method when executing the computer program.

[0039] The present invention also provides a readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the environmental geological disaster emergency warning are implemented.

[0040] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention innovatively represents the causes of geological disasters in the form of images, introduces deep learning technology, and integrates multi-dimensional representations into modeling, simplifying the difficulty of geological disaster modeling. At the same time, it enhances the representation dimension of geological disasters. Compared with the single-source method, it greatly improves the accuracy and reliability of dangerous situation warnings. Compared with traditional technologies, the process of geological disaster warnings is more reasonable and the technology is more practical.

[0041] 2. In the traditional early warning system, the transmission of early warning information often relies on a single transmission link, which is transmitted in order in a hierarchical manner. Therefore, it will take a lot of time and it is difficult to reach the terminal directly, especially in the scene where the personnel are unevenly distributed, which leads to serious lag of early warning information and may delay the best time to deal with the dangerous situation. In addition, the indiscriminate transmission of the object of early warning and the dangerous situation information is easy to cause waste of system resources and panic about the spread of dangerous situation information. Compared with the traditional early warning method, the distributed system structure has the advantages of wide deployment range of monitoring points, fast response speed, and high monitoring fault tolerance. It carries out early warning actions of different measures for different types of terminal monitoring equipment at each station in detail, helping the on-duty personnel on duty to respond to the feedback mechanism in the first time to understand the number and distribution of personnel in the station, which can well adapt to the characteristics of road and bridge engineering projects with many stations, dispersion, and large personnel flow. In addition, the present invention is only enabled during emergency warning, and generally does not affect the normal work of the construction site.

[0042] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the technical principles of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. An environmental disaster emergency warning method, characterized in that: include: Obtain historical and real-time geological disaster cause maps of different types of geological disaster cause characteristics; Segmenting the historical geological disaster cause map to obtain historical segmented images, annotating the historical segmented images according to the geological disaster cause characteristics to obtain historical sample labels, and constructing different types of data sets according to the historical sample labels and the historical segmented images; The preset deep convolutional neural network models are trained respectively by using different types of data sets to obtain feature models of different types and with different model weights; Segmenting the real-time geological disaster cause map to obtain real-time segmented images, recording the position information of each real-time segmented image, inputting the real-time segmented image into the corresponding feature model, and outputting a predicted image, wherein the predicted image has a confidence label of the real-time segmented image corresponding to the geological disaster cause feature; fusing the outputs of the feature models of different types to obtain a fused image; Restoring the fused image according to the position information to obtain different types of feature confidence maps; Performing expansion processing on the feature confidence map to obtain different types of feature region maps; Performing region merging processing on the characteristic region maps of different types, and generating new confidence labels for the merged regions respectively, to obtain a dangerous situation characteristic map; The geological disaster hazard information is obtained according to the preset hazard label criteria and the hazard characteristic map.

2. The environmental disaster emergency warning method according to claim 1, characterized in that: Obtain historical and real-time geological disaster cause maps of different types of geological disaster cause characteristics, including: Collect historical and real-time geological disaster cause data; Performing statistical integration on the historical geological disaster cause data to generate a historical geological disaster cause map that can characterize the characteristics of geological disasters under the same resolution and scale conditions; The real-time geological disaster cause data are statistically integrated and statistically analyzed to generate a real-time geological disaster cause map that can characterize the characteristics of the geological disaster under the conditions of the same resolution and the same scale.

3. The environmental disaster emergency warning method according to claim 1, characterized in that: The confidence labels include precipitation, wind speed, vegetation coverage and soil conditions, and the confidence value range of each confidence label is [0, 1].

4. The environmental disaster emergency warning method according to claim 1, characterized in that: Inputting the real-time segmented image into the feature model of the corresponding type and outputting a predicted image comprises: Using convolution kernels of three scales of 1*1, 3*3, and 5*5 and 64 channels to extract features from the real-time segmented image, to obtain a first feature matrix; Then, the first feature matrix is ​​superimposed, max pooled and batch normalized to obtain a second feature matrix; The second feature matrix is ​​subjected to feature extraction using convolution kernels of three scales of 1*1, 3*3, and 5*5 and with a channel number of 128 to obtain a third feature matrix; Performing superposition, maximum pooling and BN operations on the third feature matrix to obtain a fourth feature matrix; The fourth feature matrix is ​​subjected to feature extraction using convolution kernels of three scales, 1*1, 3*3, and 5*5, and with a channel number of 256, to obtain a fifth feature matrix; Performing superposition, maximum pooling and BN operations on the fifth feature matrix to complete feature extraction, thereby obtaining a sixth feature matrix; Perform feature extraction on the sixth feature matrix using a convolution kernel with a scale of 3*3 and a channel number of 512 to obtain a seventh feature matrix; Performing maximum pooling and BN operations on the seventh feature matrix to extract features, to obtain an eighth feature matrix; Perform feature extraction on the eighth feature matrix using a 3*3 convolution kernel with a channel number of 1024 to obtain a ninth feature matrix; Perform maximum pooling and BN operations on the ninth feature matrix to extract features, and obtain the tenth feature matrix; The tenth feature matrix is ​​subjected to feature extraction using a 3*3 convolution kernel with a channel number of 2048, and then the feature matrix is ​​subjected to maximum pooling and BN operations to obtain a final feature matrix; Finally, the final feature matrix is ​​flattened and fully connected to a feature vector with a length of 1, and nonlinear processing is performed using a Sigmoid function to obtain the predicted image.

5. The environmental disaster emergency warning method according to claim 1, characterized in that: The feature confidence map is expanded to obtain different types of feature region maps, including: Setting a sliding window on the feature confidence map based on the neighbor mean principle; The confidence average boundary dilation processing is performed on the center position of the sliding window to obtain the feature area map.

6. The environmental disaster emergency warning method according to claim 1, characterized in that: Also includes: Storing the geological disaster risk information; Distributing the geological disaster risk information to different types of terminal devices according to a distributed system, and outputting corresponding risk alerts through the terminal devices, the terminal devices including command and management terminals, public equipment terminals and personal equipment terminals; The danger alarm is sent to the corresponding danger response terminal set through different types of terminal devices. The danger response terminal set includes a command and management danger response terminal set, a public equipment danger response terminal set, and a personal equipment danger response terminal set. The command and management danger response terminal set includes the emergency command center large screen and the danger management data center, the public equipment danger response terminal set includes broadcasting, access control and alarms, and the personal equipment danger response terminal set includes computers, collection and bracelets.

7. An environmental disaster emergency warning device, characterized in that: include: An acquisition module, the acquisition module is used to acquire historical geological disaster cause maps and real-time geological disaster cause maps of different types of geological disaster cause characteristics; A construction module, wherein the construction module is used to segment the historical geological disaster cause map to obtain historical segmented images, annotate the historical segmented images according to the geological disaster cause characteristics to obtain historical sample labels, and construct different types of data sets according to the historical sample labels and the historical segmented images; A training module, wherein the training module is used to train a preset deep convolutional neural network model using different types of data sets to obtain feature models of different types and with different model weights; A prediction module, the prediction module is used to segment the real-time geological disaster cause map to obtain real-time segmented images, record the position information of each real-time segmented image, input the real-time segmented image into the corresponding feature model, and output a predicted image, wherein the predicted image has a confidence label of the real-time segmented image corresponding to the geological disaster cause feature; A fusion module, the fusion module is used to fuse the outputs of the feature models of different types to obtain a fused image; A restoration module, the restoration module is used to restore the fused image according to the position information to obtain different types of feature confidence maps; An expansion module, the expansion module is used to expand the feature confidence map to obtain different types of feature region maps; A merging module, the merging module is used to perform region merging processing on the characteristic region maps of different types, and generate new confidence labels for the merged regions respectively, to obtain a dangerous situation characteristic map; A judgment module is used to judge and obtain geological disaster hazard information according to a preset hazard label criterion and the hazard characteristic map.

8. An environmental disaster emergency warning device, characterized in that: include: Memory for storing computer programs; A processor is used to implement the steps of an environmental disaster emergency warning method as described in any one of claims 1 to 6 when executing the computer program.

9. A readable storage medium, characterized in that: The readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of an environmental disaster emergency warning as claimed in any one of claims 1 to 6 are implemented.

Images