Debris flow and landslide positioning and classifying method based on natural images

Abstract

The invention relates to a debris flow and landslide positioning and classifying method based on natural images. The method comprises the following steps: firstly, acquiring natural image data of debris flow and landslide, and preprocessing the natural image data of debris flow and landslide to obtain disaster data; then, a disaster range positioning model is built, and the disaster range positioning model comprises a U-Net network module and an SE module; then, inputting the disaster data into the disaster range positioning model to obtain a disaster range image; and finally, performing feature extraction and iterative feature selection based on the disaster range image, and performing classification by adopting a machine learning model to obtain a target debris flow and landslide classification result.

Description

Technical Field

[0001] The present application relates to the technical field of disaster management, and in particular to a method for locating and classifying debris flows and landslides based on natural images. Background Art

[0002] Debris flow and landslide are two extremely destructive geological disasters, which often cause huge losses to industrial and agricultural production as well as people's lives and property. Therefore, the research on debris flow and landslide has also attracted more and more attention. Among them, accurately locating debris flow and landslide in monitoring images to determine the disaster scope and classifying the two natural disasters according to image features are crucial for disaster management.

[0003] With the application and popularization of drones and monitoring equipment, it is becoming more and more convenient to obtain natural images for different landforms and regions. Therefore, the observation and early warning research of natural geological disasters can be carried out with the help of natural images. Compared with remote sensing SAR and other images, natural images have the following advantages: First, the resolution of natural images is usually higher, which can provide clearer and more detailed landform features, which is crucial for identifying and distinguishing disaster types such as debris flows and landslides. Secondly, natural images can be obtained under different time and climate conditions, which can reflect the landform changes before and after the disaster occurs, and provide continuous time series data for disaster monitoring and assessment. In addition, natural image equipment such as drones and monitoring equipment are low-cost and easy to operate, and can be quickly deployed to the disaster site for real-time monitoring and data collection.

[0004] When a disaster occurs, the analysis of natural images can quickly identify the type of disaster and the scope of impact, and take appropriate emergency measures in a timely manner to reduce disaster losses. At the same time, natural images can also be used for post-disaster assessments to help formulate recovery and reconstruction plans. Therefore, using natural images to monitor and classify debris flows and landslides can not only improve the accuracy and efficiency of disaster identification, but also provide important data support and decision-making basis for disaster management and mitigation.

[0005] In the prior art, the research on the location and classification of debris flow and landslide images is relatively limited. Many studies rely on traditional image processing methods based on manual feature extraction and rule formulation. Such methods are difficult to deal with polygonal states under different terrain and climatic conditions, which not only leads to insufficient positioning and classification accuracy, but also increases the risk of misjudgment, thus affecting the accuracy and timeliness of disaster management decisions. In addition, the image data of debris flow and landslide often contain a lot of noise and background information. Traditional methods are inefficient in processing such complex information and are prone to inaccurate location of disaster scope. In emergency response and risk assessment, inaccurate positioning may lead to the misallocation of resources and inaccurate response measures. In severe cases, it may delay rescue time and increase loss of life and property.

[0006] Therefore, in the relevant technology, there is an urgent need for a method that can improve the accuracy and reliability of debris flow and landslide disaster positioning and classification based on widely acquired natural images. Summary of the invention

[0007] Based on this, it is necessary to provide a debris flow and landslide positioning and classification method based on natural images that can improve the accuracy and reliability of debris flow and landslide disaster positioning and classification in response to the above technical problems.

[0008] In a first aspect, the present application provides a method for locating and classifying debris flows and landslides based on natural images. The method comprises:

[0009] Acquire natural image data of debris flow and landslide, and pre-process the natural image data of debris flow and landslide to obtain disaster data;

[0010] Building a disaster range positioning model, wherein the disaster range positioning model includes a U-Net network module and a SE module;

[0011] Inputting the disaster data into the disaster range positioning model to obtain a disaster range image;

[0012] Feature extraction and iterative feature selection are performed based on the disaster range image, and classification is performed using a machine learning model to obtain target debris flow and landslide classification results.

[0013] Optionally, in one embodiment of the present application, the preprocessing of the debris flow and landslide natural image data includes:

[0014] The debris flow and landslide natural image data are normalized by using a Z-score normalization method;

[0015] Image data enhancement is performed on the normalized debris flow and landslide natural image data.

[0016] Optionally, in one embodiment of the present application, the performing image data enhancement on the normalized debris flow and landslide natural image data comprises:

[0017] The normalized debris flow and landslide natural image data are rotated, translated, scaled and flipped.

[0018] Optionally, in one embodiment of the present application, the U-Net network module includes an encoder structure and a decoder structure, the encoder structure is used to extract features of image context information, and the decoder structure is used to recover information lost during image decoding.

[0019] Optionally, in one embodiment of the present application, the encoder structure includes a convolution layer and a pooling layer, the SE module is integrated between the convolution layer and the pooling layer, and the decoder structure includes a deconvolution operation and a convolution layer.

[0020] Optionally, in one embodiment of the present application, the iterative feature selection based on the disaster range image includes:

[0021] The SHAP value was calculated to evaluate the importance of each feature, and important features were selected based on the importance.

[0022] Optionally, in one embodiment of the present application, the method further includes:

[0023] The classification results of the target debris flow and landslide were evaluated based on Dice coefficient, intersection-over-union ratio, sensitivity and specificity.

[0024] In a second aspect, the present application also provides a debris flow and landslide location classification device based on natural images. The device comprises:

[0025] A data acquisition and processing module is used to acquire natural image data of debris flow and landslide, and pre-process the natural image data of debris flow and landslide to obtain disaster data;

[0026] A disaster range positioning model building module is used to build a disaster range positioning model, wherein the disaster range positioning model includes a U-Net network module and a SE module;

[0027] A disaster range positioning module, used for inputting the disaster data into the disaster range positioning model to obtain a disaster range image;

[0028] The classification module is used to perform feature extraction and iterative feature selection based on the disaster range image, and use a machine learning model for classification to obtain target debris flow and landslide classification results.

[0029] In a third aspect, the present application further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and the processor executes the steps of the methods described in the above embodiments.

[0030] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method described in each of the above embodiments are implemented.

[0031] The above-mentioned method for locating and classifying debris flows and landslides based on natural images, first, obtains the natural image data of debris flows and landslides, pre-processes the natural image data of debris flows and landslides, and obtains disaster data. Afterwards, a disaster range positioning model is built, and the disaster range positioning model includes a U-Net network module and an SE module; then, the disaster data is input into the disaster range positioning model to obtain a disaster range image; finally, feature extraction and iterative feature selection are performed based on the disaster range image, and a machine learning model is used for classification to obtain the classification results of target debris flows and landslides. In other words, when locating and classifying debris flows and landslide disasters, by integrating the SE module in the U-Net network, the feature representation capability is effectively enhanced, the positioning accuracy is improved, and the feature set is optimized in combination with the iterative feature selection method, so that the model can maintain high accuracy while reducing the computational complexity, effectively reducing feature redundancy, and improving the generalization ability of the classification model. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 An application environment diagram of a debris flow and landslide location classification method based on natural images in one embodiment;

[0033] Figure 2 A schematic diagram of a flow chart of a method for locating and classifying debris flows and landslides based on natural images in one embodiment;

[0034] Figure 3 A schematic diagram of the structure of a disaster range positioning model in one embodiment;

[0035] Figure 4 is a schematic diagram of an iterative feature selection process in one embodiment;

[0036] Figure 5 A flowchart of specific steps of a method for locating and classifying debris flows and landslides based on natural images in one embodiment;

[0037] Figure 6 is a structural block diagram of a debris flow and landslide location and classification device based on natural images in one embodiment;

[0038] Figure 7 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0039] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application 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 application and are not used to limit the present application.

[0040] The embodiment of the present application provides a method for locating and classifying debris flows and landslides based on natural images, which can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. Among them, the terminal 102 can be but is not limited to various personal computers, laptops, smart phones, tablets, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The server 104 can be implemented with an independent server or a server cluster consisting of multiple servers.

[0041] In one embodiment, Figure 2 As shown in the figure, a method for locating and classifying debris flows and landslides based on natural images is provided. Figure 1 The server in the example is used to illustrate the following steps:

[0042] S201: Obtain natural image data of debris flow and landslide, and pre-process the natural image data of debris flow and landslide to obtain disaster data.

[0043] In the embodiment of the present application, first, the natural image data of debris flow and landslide is obtained. The natural image data of debris flow and landslide mainly includes the relevant data of the debris flow and landslide-prone areas and their surrounding areas, including data sources, interpretation methods, data content and geographical scope, and involves multiple earthquake core areas and coastal areas. After that, the natural image data of debris flow and landslide is preprocessed, such as normalization by methods such as minimum-maximum scaling, Z-score standardization and mean normalization, and the image data is enhanced to obtain disaster data.

[0044] S203: Building a disaster range positioning model, wherein the disaster range positioning model includes a U-Net network module and a SE module.

[0045] In the embodiment of the present application, a disaster range positioning model is constructed, and the disaster range positioning model mainly includes a U-Net network module and a SE module, such as Figure 3 As shown in the figure, the U-Net network module is a convolutional neural network architecture with a symmetrical U-shaped structure, including a contraction path (encoder) and an expansion path (decoder), which enables it to capture contextual information and make accurate pixel-level predictions. The SE (Squeeze-and-Excitation) module is designed to recalibrate channel-level feature responses by showing the interdependencies between modeling channels.

[0046] S205: Input the disaster data into the disaster range positioning model to obtain a disaster range image.

[0047] In an embodiment of the present application, the disaster data obtained after preprocessing is input into a disaster range positioning model, and after multi-layer downsampling, upsampling and feature fusion, the final disaster range image is output, that is, the accurate image disaster range positioning.

[0048] S207: Perform feature extraction and iterative feature selection based on the disaster range image, and use a machine learning model for classification to obtain target debris flow and landslide classification results.

[0049] In the present application embodiment, finally, in the process of classification experiment, feature selection is a key step. The main goal of feature selection is to filter out the features with the highest predictive power for the target variable from the original data, so as to improve the performance and generalization ability of the classification model. In the feature selection process, iterative feature selection is performed based on accurate image disaster range positioning, and iterative feature selection optimizes the feature set by gradually adding or removing features to evaluate their impact on model performance. Specifically, features include geometric features (area, length, aspect ratio), texture features (contrast, dissimilarity, homogeneity, energy), color features (color histogram, gradient color features) and grayscale features (average grayscale, grayscale standard deviation). And machine learning models are used for classification to obtain the final target debris flow and landslide classification results, wherein the machine learning models include k nearest neighbor (KNN), support vector machine (SVM), random forest, naive Bayes and decision tree.

[0050] In the above-mentioned method for locating and classifying debris flows and landslides based on natural images, first, the natural image data of debris flows and landslides are obtained, and the natural image data of debris flows and landslides are preprocessed to obtain disaster data. Afterwards, a disaster range positioning model is built, and the disaster range positioning model includes a U-Net network module and an SE module; then, the disaster data is input into the disaster range positioning model to obtain a disaster range image; finally, feature extraction and iterative feature selection are performed based on the disaster range image, and a machine learning model is used for classification to obtain the classification results of target debris flows and landslides. In other words, when locating and classifying debris flows and landslide disasters, by integrating the SE module in the U-Net network, the feature representation capability is effectively enhanced, the positioning accuracy is improved, and the feature set is optimized in combination with the iterative feature selection method, so that the model can maintain high accuracy while reducing the computational complexity, effectively reducing feature redundancy, and improving the generalization ability of the classification model.

[0051] In one embodiment of the present application, the preprocessing of the debris flow and landslide natural image data includes:

[0052] S301: normalizing the debris flow and landslide natural image data using a Z-score normalization method.

[0053] S303: Performing image data enhancement on the normalized debris flow and landslide natural image data.

[0054] In one embodiment of the present application, the Z-score normalization method is used to normalize the natural image data of debris flow and landslide, and the normalization formula is as follows:

[0055]

[0056] Here, x is the mean of the data, μ is the standard deviation, and z is the standard score.

[0057] After that, the normalized debris flow and landslide natural image data are enhanced. When the amount of data is limited, image data enhancement can not only increase the size of the data set, but also generate images with different features, so that the model can better adapt to diverse data, which helps to improve the generalization ability of the model and prevent overfitting. In specific applications, the most commonly used methods include rotation, translation, scaling and flipping.

[0058] In this embodiment, the Z-score normalization method is used to normalize the debris flow and landslide natural image data, and image data enhancement is performed on the normalized debris flow and landslide natural image data, so that the scale of the data set can be effectively increased.

[0059] In one embodiment of the present application, the step of performing image data enhancement on the normalized debris flow and landslide natural image data comprises:

[0060] The normalized debris flow and landslide natural image data are rotated, translated, scaled and flipped.

[0061] In one embodiment of the present application, rotation refers to rotating the image at a certain angle, which can be random or fixed. By rotating the image, the model can learn features that are invariant to different angles, so that it has better adaptability when processing real data. The mathematical formula of the rotation transformation is as follows: For a point (x, y), the rotation angle is θ, and the new coordinates (x′, y′) after rotation are calculated as follows:

[0062]

[0063] Translation refers to moving an image horizontally or vertically by a certain distance to generate image views at different positions. Translation transformation can help the model learn position invariance and enhance its ability to recognize objects in various positions. The formula for translation transformation is: for a point (x, y), the translation amounts are Δx and Δy, and the new coordinates (x′, y′) after translation are calculated as follows:

[0064] x′=x+Δx

[0065] y′=y+Δy

[0066] Among them, Δx and Δy refer to the displacement of the image on the x-axis and y-axis when the image is translated, Δx represents the moving distance in the horizontal direction, and Δy represents the moving distance in the vertical direction.

[0067] Scaling refers to changing the size of an image, either by enlarging or reducing it. By scaling the image, the model can learn to be robust to different object sizes, enabling it to better handle real-world situations where object sizes vary. The formula for the scaling transformation is: For a point (x, y), the scaling factor is s x and y , the new scaled coordinates (x′, y′) are calculated as follows:

[0068] x′=s x ·x

[0069] y′=s y ·y

[0070] Among them, s x and y are scaling factors used to scale the image on the x-axis and y-axis respectively. These two factors determine the ratio of the image to be enlarged or reduced in the horizontal and vertical directions. Specifically, s x is the scaling factor in the x-axis direction. If s x >1, the image is enlarged in the x-axis direction; if s x <1, the image is reduced in the x-axis direction. y is the scaling factor in the y-axis direction. If s y >1, the image is enlarged in the y-axis direction; if s y <1, the image is reduced in the y-axis direction.

[0071] Flipping includes horizontal flipping and vertical flipping, which are symmetrical transformations of the image along the horizontal axis or vertical axis respectively. These transformations can generate mirror images, helping the model learn the invariance of symmetrical changes. The formula for horizontal flipping is: For a point (x, y), the new coordinates (x', y') after horizontal flipping are calculated as follows:

[0072] x′=ω-x

[0073] y′=y

[0074] Where ω is the width of the image. The formula for vertical flipping is: For a point (x, y), the new coordinates (x', y') after vertical flipping are calculated as follows:

[0075] x′=x

[0076] y′=hy

[0077] Where h is the height of the image.

[0078] In this embodiment, by rotating, translating, scaling and flipping the normalized debris flow and landslide natural image data, and performing geometric transformation on the original image to generate new training samples, the generalization ability and robustness of the model can be enhanced.

[0079] In one embodiment of the present application, the U-Net network module includes an encoder structure and a decoder structure, wherein the encoder structure is used to extract features of image context information, and the decoder structure is used to recover information lost during image decoding.

[0080] In one embodiment of the present application, the U-Net network module mainly includes an encoder structure and a decoder structure, wherein the encoder structure realizes feature extraction of image context information, and the decoder structure recovers the information lost in the image decoding process by upsampling. In addition, a jump connection structure is used to fuse and supplement the corresponding features extracted by the encoder and decoder parts, thereby achieving a superposition effect of feature information.

[0081] In one embodiment of the present application, the encoder structure includes a convolution layer and a pooling layer, the SE module is integrated between the convolution layer and the pooling layer, and the decoder structure includes a deconvolution operation and a convolution layer.

[0082] In one embodiment of the present application, the encoder structure of the U-Net network module is divided into four downsampling parts. Image features are gradually extracted through these four downsampling modules, each of which includes two convolutional layers and a maximum pooling layer. For each downsampling operation, the number of feature map channels is doubled, while the spatial dimension is halved. The SE module is integrated after each pair of convolutional layers and before each maximum pooling layer in the encoder, performs a global average pooling operation, and then generates channel-level weights through two fully connected layers, which are used to scale the input feature map. This operation enhances the network's sensitivity to important features and suppresses less important features, thereby improving feature representation. The decoder structure consists of four upsampling parts to restore the original size of the image, each upsampling part includes a deconvolution operation and two convolutional layers. Through upsampling, the spatial dimension of the feature map is gradually restored, while the number of channels is halved. The upsampled feature map is combined with the corresponding feature map in the encoder through jump connections to promote the integration of multi-level information. These jump connections connect shallow detail information with deep semantic information to further optimize feature extraction.

[0083] In this embodiment, through multi-layer downsampling, upsampling and feature fusion, U-Net is able to maintain high resolution while utilizing multi-level feature information, thereby achieving accurate image disaster range positioning.

[0084] In one embodiment of the present application, the iterative feature selection based on the disaster range image includes:

[0085] The SHAP value was calculated to evaluate the importance of each feature, and important features were selected based on the importance.

[0086] In one embodiment of the present application, the SHAP (SHapley Additive exPlanations) value is calculated to accurately evaluate the importance of each feature. The SHAP value is an effective interpretation tool for measuring the contribution of each feature to model prediction. By calculating the SHAP value, the impact of each feature on the classification result can be identified, and which features play a key role in model prediction can be determined. Figure 4 The figure shows a schematic diagram of the iterative feature selection process. First, the SHAP value of each feature is calculated based on the training data set. Then, based on the SHAP value, the features with low contribution to the model are eliminated, and the iterative process is repeated.

[0087] In this embodiment, the importance of each feature is evaluated by calculating the SHAP value, and important features are screened based on the importance, thereby enhancing the scientificity and accuracy of feature selection and making the prediction results more transparent and reliable.

[0088] In one embodiment of the present application, the method further includes:

[0089] The classification results of the target debris flow and landslide were evaluated based on Dice coefficient, intersection-over-union ratio, sensitivity and specificity.

[0090] In one embodiment of the present application, the Dice coefficient, also known as The Dice coefficient is a statistical tool used to measure the similarity between two samples. In the context of image segmentation, it is used to compare the predicted segmentation results with the actual data. The Dice coefficient ranges from 0 to 1, where 1 indicates perfect segmentation. Its calculation formula is as follows:

[0091]

[0092] Among them, X represents the predicted segmentation result and Y represents the actual real data.

[0093] Intersection over Union (IoU) is another metric for evaluating segmentation accuracy. It calculates the ratio of the overlapping area of ​​the predicted and actual segmentation to the combined area. The IoU value ranges from 0 to 1, where 1 indicates a perfect segmentation. The calculation formula for IoU is as follows:

[0094]

[0095] Among them, X represents the predicted segmentation result and Y represents the actual real data.

[0096] In image segmentation, sensitivity usually refers to the ability of the model to correctly identify the positive class. Higher sensitivity means fewer positive class instances are missed. The formula for sensitivity is as follows:

[0097]

[0098] Among them, TP represents true positive and FN represents false negative.

[0099] Specificity is the ability of the model to correctly identify negative classes (such as non-lesion areas). Higher specificity means a lower rate of misclassifying negative classes as positive classes. The specificity is calculated as follows:

[0100]

[0101] Among them, TN means true negative and FP means false positive.

[0102] In this embodiment, by evaluating the target debris flow and landslide classification results based on the Dice coefficient, intersection-over-union ratio, sensitivity and specificity, the accuracy and reliability of debris flow and landslide disaster positioning and classification can be improved.

[0103] The following is a specific example to illustrate the specific implementation steps of a method for locating and classifying debris flows and landslides based on natural images of the present application. Figure 5As shown, first, S401, obtain natural image data of debris flow and landslide, pre-process the natural image data of debris flow and landslide, and obtain disaster data. Specifically, S403-S405, use the Z-score standardization method to normalize the natural image data of debris flow and landslide; enhance the image data of the normalized natural image data of debris flow and landslide. S407, rotate, translate, scale and flip the normalized natural image data of debris flow and landslide.

[0104] Afterwards, S409, a disaster range location model is constructed, and the disaster range location model includes a U-Net network module and a SE module. Among them, S411, the U-Net network module includes an encoder structure and a decoder structure, the encoder structure is used to extract the features of the image context information, and the decoder structure is used to restore the information lost during the image decoding process. S413, the encoder structure includes a convolution layer and a pooling layer, the SE module is integrated between the convolution layer and the pooling layer, and the decoder structure includes a deconvolution operation and a convolution layer.

[0105] Afterwards, S415, the disaster data is input into the disaster range positioning model to obtain a disaster range image. Finally, S417, feature extraction and iterative feature selection are performed based on the disaster range image, and classification is performed using a machine learning model to obtain target debris flow and landslide classification results. S419, SHAP value is calculated to evaluate the importance of each feature, and important features are screened based on the importance. S421, the target debris flow and landslide classification results are evaluated based on the Dice coefficient, intersection-over-union ratio, sensitivity and specificity.

[0106] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0107] Based on the same inventive concept, the embodiment of the present application also provides a device for implementing the above-mentioned method for locating and classifying debris flows and landslides based on natural images. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in one or more embodiments of the device for locating and classifying debris flows and landslides based on natural images provided below can refer to the limitations of the above-mentioned method for locating and classifying debris flows and landslides based on natural images, and will not be repeated here.

[0108] In one embodiment, Figure 6 As shown, a debris flow and landslide positioning and classification device 600 based on natural images is provided, comprising: a data acquisition and processing module 601, a disaster range positioning model building module 603, a disaster range positioning module 605 and a classification module 607, wherein:

[0109] The data acquisition and processing module 601 is used to acquire natural image data of debris flow and landslide, and pre-process the natural image data of debris flow and landslide to obtain disaster data.

[0110] The disaster range positioning model building module 603 is used to build a disaster range positioning model, and the disaster range positioning model includes a U-Net network module and a SE module.

[0111] The disaster range positioning module 605 is used to input the disaster data into the disaster range positioning model to obtain a disaster range image.

[0112] The classification module 607 is used to perform feature extraction and iterative feature selection based on the disaster range image, and use a machine learning model to perform classification to obtain target debris flow and landslide classification results.

[0113] In one embodiment of the present application, the data acquisition and processing module is further used to:

[0114] The debris flow and landslide natural image data are normalized by using a Z-score normalization method;

[0115] Image data enhancement is performed on the normalized debris flow and landslide natural image data.

[0116] In one embodiment of the present application, the data acquisition and processing module is further used to:

[0117] The normalized debris flow and landslide natural image data are rotated, translated, scaled and flipped.

[0118] In one embodiment of the present application, the U-Net network module includes an encoder structure and a decoder structure, wherein the encoder structure is used to extract features of image context information, and the decoder structure is used to recover information lost during image decoding.

[0119] In one embodiment of the present application, the encoder structure includes a convolution layer and a pooling layer, the SE module is integrated between the convolution layer and the pooling layer, and the decoder structure includes a deconvolution operation and a convolution layer.

[0120] In one embodiment of the present application, the classification module is further used for:

[0121] The SHAP value was calculated to evaluate the importance of each feature, and important features were selected based on the importance.

[0122] The debris flow and landslide location classification device based on natural images further includes a classification result evaluation module. In one embodiment of the present application, the classification result evaluation module is used to:

[0123] The classification results of the target debris flow and landslide were evaluated based on Dice coefficient, intersection-over-union ratio, sensitivity and specificity.

[0124] Each module in the above-mentioned debris flow and landslide location classification device based on natural images can be implemented in whole or in part by software, hardware and their combination. Each of the above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0125] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 7As shown. The computer device includes a processor, a memory, a communication interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for locating and classifying debris flows and landslides based on natural images is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covered on the display screen, or a key, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0126] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0127] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0128] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0129] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0130] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0131] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.

[0132] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0133] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A method for the location and classification of debris flows and landslides based on natural images, It is characterized in that The method comprises: Acquire natural image data of debris flow and landslide, and pre-process the natural image data of debris flow and landslide to obtain disaster data; Building a disaster range positioning model, wherein the disaster range positioning model includes a U-Net network module and a SE module; Inputting the disaster data into the disaster range positioning model to obtain a disaster range image; Feature extraction and iterative feature selection are performed based on the disaster range image, and classification is performed using a machine learning model to obtain target debris flow and landslide classification results.

2. A method for locating and classifying debris flows and landslides based on natural images according to claim 1, It is characterized in that The preprocessing of the debris flow and landslide natural image data comprises: The debris flow and landslide natural image data are normalized by using a Z-score normalization method; Image data enhancement is performed on the normalized debris flow and landslide natural image data.

3. The method for locating and classifying debris flows and landslides based on natural images according to claim 2, It is characterized in that The step of performing image data enhancement on the normalized debris flow and landslide natural image data comprises: The normalized debris flow and landslide natural image data are rotated, translated, scaled and flipped.

4. The method for locating and classifying debris flows and landslides based on natural images according to claim 1, It is characterized in that The U-Net network module includes an encoder structure and a decoder structure, wherein the encoder structure is used to extract features of image context information, and the decoder structure is used to restore information lost during image decoding.

5. The method for locating and classifying debris flows and landslides based on natural images according to claim 4, It is characterized in that The encoder structure includes a convolution layer and a pooling layer, the SE module is integrated between the convolution layer and the pooling layer, and the decoder structure includes a deconvolution operation and a convolution layer.

6. The method for locating and classifying debris flows and landslides based on natural images according to claim 1, It is characterized in that The iterative feature selection based on the disaster range image includes: The SHAP value was calculated to evaluate the importance of each feature, and important features were selected based on the importance.

7. The method for locating and classifying debris flows and landslides based on natural images according to claim 1, It is characterized in that The method further comprises: The classification results of the target debris flow and landslide were evaluated based on Dice coefficient, intersection-over-union ratio, sensitivity and specificity.

8. A debris flow and landslide location and classification device based on natural images, It is characterized in that The device comprises: A data acquisition and processing module is used to acquire natural image data of debris flow and landslide, and pre-process the natural image data of debris flow and landslide to obtain disaster data; A disaster range positioning model building module is used to build a disaster range positioning model, wherein the disaster range positioning model includes a U-Net network module and a SE module; A disaster range positioning module, used for inputting the disaster data into the disaster range positioning model to obtain a disaster range image; The classification module is used to perform feature extraction and iterative feature selection based on the disaster range image, and use a machine learning model for classification to obtain target debris flow and landslide classification results.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program. It is characterized in that When the processor executes the computer program, the steps of the method according to any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images