Method and system for zoning control of water and soil environmental pollution risk in karst cities

Abstract

This application discloses a karst urban water and soil environmental pollution risk zoning control method and system, which uses remote sensing images of the monitored area to characterize the water and soil environmental characteristics of the monitored area, and then uses an artificial intelligence algorithm based on deep learning to extract features and conduct comprehensive feature analysis on the remote sensing images of the monitored area to obtain the risk level assessment results of the controlled area. In this way, it can help decision makers and environmental management departments understand the risk status of the water and soil environment in karst areas, formulate control measures in a targeted manner, protect and improve the quality of the water and soil environment, and reduce the impact of environmental pollution risks on human health and ecosystems.

Description

Technical Field

[0001] The present application relates to the field of intelligent management and control, and more specifically, to a method and system for zoning management of water and soil environmental pollution risks in karst cities. Background Art

[0002] The karst urban water and soil environmental pollution risk zoning control plan is a comprehensive governance measure formulated based on the idea and method of risk management for the special water and soil environmental problems in karst areas. The purpose of building such a plan is to protect the water resources and ecosystems in karst areas and prevent or reduce the impact of water and soil environmental pollution on human health and social and economic development.

[0003] In the process of building such a plan, determining the risk level of the controlled area is an important step. The risk level is a relative indicator obtained through comprehensive evaluation based on the water and soil environmental pollution status of the controlled area and the potential degree of harm caused by pollutants to receptors (such as groundwater, soil, vegetation, humans, etc.). Determining the risk level helps to identify pollution problems and governance needs in different areas, so as to formulate corresponding control measures and priorities.

[0004] However, there are many technical difficulties in determining the risk level of the controlled area. First, the water and soil environment in karst areas is complex and dynamic. Second, the water and soil environment in karst areas is affected by many factors, such as geology, climate, land use, human activities, etc., which leads to the need to consider information in multiple dimensions and levels in risk assessment.

[0005] Therefore, an optimized karst urban water and soil environmental pollution risk zoning control plan is expected. Summary of the invention

[0006] In order to solve the above-mentioned technical problems, the present application is proposed. The embodiment of the present application provides a karst urban soil and water environmental pollution risk zoning control method and system, which characterizes the soil and water environmental characteristics of the monitored area through the remote sensing image of the monitored area, and then uses the artificial intelligence algorithm based on deep learning to extract features and comprehensively analyze the features of the remote sensing image of the monitored area to obtain the risk level assessment result of the controlled area. In this way, it can help decision makers and environmental management departments understand the risk status of the soil and water environment in karst areas, formulate control measures in a targeted manner, protect and improve the quality of the soil and water environment, and reduce the impact of environmental pollution risks on human health and ecosystems.

[0007] According to one aspect of the present application, a method for zoning control of water and soil environmental pollution risks in karst cities is provided, which includes:

[0008] Obtain multiple remote sensing images of the controlled area;

[0009] Extracting a plurality of local area remote sensing features of the surface from the plurality of remote sensing images;

[0010] Determining remote sensing essential semantic features of the monitored area based on the multiple local area remote sensing features of the surface; and

[0011] Based on the remote sensing essential semantic features of the monitored area, the risk level classification result of the controlled area is determined.

[0012] According to another aspect of the present application, a karst urban water and soil environmental pollution risk zoning management and control system is provided, which includes:

[0013] An image acquisition module is used to acquire multiple remote sensing images of the controlled area;

[0014] A surface local area remote sensing feature extraction module, used to extract multiple surface local area remote sensing features from the multiple remote sensing images;

[0015] A monitored area remote sensing essential semantic extraction module, used to determine the remote sensing essential semantic features of the monitored area based on the multiple local area remote sensing features of the surface; and

[0016] The risk classification result generating module is used to determine the risk classification result of the controlled area based on the remote sensing essential semantic features of the monitored area.

[0017] Compared with the prior art, the present application provides a method and system for zoning control of water and soil environmental pollution risks in karst cities. It uses remote sensing images of the monitored area to characterize the water and soil environmental characteristics of the monitored area, and then uses an artificial intelligence algorithm based on deep learning to extract features and conduct comprehensive feature analysis on the remote sensing images of the monitored area to obtain the risk level assessment results of the controlled area. In this way, it can help decision makers and environmental management departments understand the risk status of the water and soil environment in karst areas, formulate control measures in a targeted manner, protect and improve the quality of the water and soil environment, and reduce the impact of environmental pollution risks on human health and ecosystems. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0019] Figure 1 It is a flow chart of a method for zoning control of water and soil environmental pollution risks in karst cities according to an embodiment of the present application;

[0020] Figure 2 A system architecture diagram of a method for controlling the risk of karst urban water and soil environmental pollution by zoning according to an embodiment of the present application;

[0021] Figure 3 It is a flowchart of sub-step S3 of the method for zoning control of water and soil environmental pollution risk in karst cities according to an embodiment of the present application;

[0022] Figure 4 It is a flowchart of sub-step S32 of the method for zoning control of water and soil environmental pollution risk in karst cities according to an embodiment of the present application;

[0023] Figure 5 It is a block diagram of a karst urban water and soil environmental pollution risk zoning control system according to an embodiment of the present application. DETAILED DESCRIPTION

[0024] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0025] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0026] Although the present application makes various references to certain modules in the system according to the embodiments of the present application, any number of different modules can be used and run on the user terminal and / or server. The modules are only illustrative, and different aspects of the system and method can use different modules.

[0027] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously as required. Meanwhile, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0028] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0029] In the technical solution of the present application, a method for zoning control of water and soil environmental pollution risks in karst cities is proposed. Figure 1 It is a flow chart of the method for zoning control of karst urban water and soil environmental pollution risks according to an embodiment of the present application. Figure 2 : is a system architecture diagram of the method for controlling the risk of karst urban water and soil environmental pollution zoning according to an embodiment of the present application. Figure 1 and Figure 2 As shown, the method for karst urban water and soil environmental pollution risk zoning control according to the embodiment of the present application includes the steps of: S1, acquiring multiple remote sensing images of the controlled area; S2, extracting multiple surface local area remote sensing features from the multiple remote sensing images; S3, determining the remote sensing essential semantic features of the monitored area based on the multiple surface local area remote sensing features; and, S4, determining the risk level division result of the controlled area based on the remote sensing essential semantic features of the monitored area.

[0030] In particular, the S1 obtains multiple remote sensing images of the controlled area. It should be understood that remote sensing images are image data of the earth's surface obtained by satellites, aircraft or other remote sensing technologies, which can provide large-scale, high-resolution surface information. In particular, in the technical solution of the present application, in the karst urban water and soil environmental pollution risk zoning, the multiple remote sensing images can provide the topography, land use, vegetation coverage and other surface information of the controlled area. This information is very important for assessing the risk of water and soil environmental pollution and can reveal the characteristics and changes of the surface.

[0031] In particular, the S2 extracts multiple local surface area remote sensing features from the multiple remote sensing images. That is, in the technical solution of the present application, the multiple remote sensing images are respectively passed through the surface state feature extractor of the control area based on the convolutional neural network model to obtain multiple local surface area remote sensing feature maps. Here, the purpose of processing multiple remote sensing images through the surface state feature extractor of the control area based on the convolutional neural network model to obtain multiple local surface area remote sensing feature maps is to capture and extract key feature information of the surface. It should be known to those of ordinary skill in the art that the convolutional neural network (CNN) is a deep learning model with good feature extraction and image recognition capabilities. By inputting the remote sensing image into a trained convolutional neural network, the network can automatically learn and extract surface features in the image, such as landforms, land use types, vegetation coverage, etc. These surface features are of great significance for assessing the risk of water and soil environmental pollution. In a specific example of the present application, the surface state feature extractor of the control area based on the convolutional neural network model is a pyramid network. Specifically, each layer of the controlled area surface state feature extractor based on the convolutional neural network model performs the following operations on the input data in the forward pass of the layer: convolution processing on the input data to obtain a convolution feature map; pooling on the convolution feature map based on the local feature matrix to obtain a pooled feature map; and nonlinear activation on the pooled feature map to obtain an activated feature map; wherein the output of the last layer of the controlled area surface state feature extractor based on the convolutional neural network model is the multiple local surface area remote sensing feature maps, and the input of the first layer of the controlled area surface state feature extractor based on the convolutional neural network model is the multiple remote sensing images.

[0032] It is worth mentioning that convolutional neural network (CNN) is a deep learning model specially designed for processing data with grid structure, such as images and speech. The core idea of ​​CNN is to extract the features of input data through convolution operation, and to perform high-level representation and abstraction of features by stacking layers. The following are the basic components and working principles of CNN: Convolution layer: The convolution layer is the core component of CNN, which is used to extract the features of input data. It performs convolution operation on the input data by applying a set of learnable convolution kernels (filters). The convolution operation can capture local patterns and features in the input data and generate a series of feature maps; Activation function: After the convolution layer, a nonlinear activation function such as ReLU is usually applied. The activation function introduces nonlinear features, enabling the network to learn more complex patterns and representations; Pooling layer: The pooling layer is used to reduce the size and number of parameters of the feature map and extract the most important features. Common pooling operations include maximum pooling and average pooling; Fully connected layer: After a series of convolutional layers and pooling layers, some fully connected layers are usually added. The fully connected layer converts the feature map of the previous layer into an output result, such as classification or regression; Dropout: In order to prevent overfitting, Dropout technology is often used in CNN. Dropout randomly discards a part of neurons during the training process to reduce the dependency between neurons and improve the generalization ability of the model. Through the back propagation algorithm, CNN can automatically learn to extract features from the input data and optimize according to the training objectives. During the training process, CNN adjusts the network parameters by minimizing the loss function so that the output results are as close to the true labels as possible.

[0033] In particular, the S3 determines the remote sensing essential semantic features of the monitored area based on the multiple local area remote sensing features of the surface. In particular, in a specific example of the present application, Figure 3 As shown, the S3 includes: S31, using a spatial perception feature fusion module to perform spatial dimension expression significance on the multiple local surface area remote sensing feature maps to obtain multiple spatial dimension significant local surface area remote sensing feature maps; S32, performing essential feature extraction on the multiple spatial dimension significant local surface area remote sensing feature maps to obtain a remote sensing essential semantic feature vector of the monitored area as the remote sensing essential semantic feature of the monitored area.

[0034] Specifically, the S31 uses a spatial perception feature fusion module to perform spatial dimension expression saliency on the multiple local surface area remote sensing feature maps to obtain multiple spatial dimension saliency local surface area remote sensing feature maps. That is, in the technical solution of the present application, the spatial perception feature fusion module is used to perform spatial dimension expression saliency on the multiple local surface area remote sensing feature maps to obtain multiple spatial dimension saliency local surface area remote sensing feature maps. In the karst urban water and soil environmental pollution risk zoning, different local surface area remote sensing feature maps contain different surface information, such as land use type, vegetation coverage, terrain, etc., and there is a certain correlation and spatial relationship between these feature maps. Therefore, in the technical solution of the present application, the spatial perception feature fusion module is further used to capture the important feature information of the multiple local surface area remote sensing feature maps in the spatial domain. Specifically, in the technical solution of the present application, the spatial perception feature fusion module performs spatial dimension expression saliency on the multiple local surface area remote sensing feature maps to obtain the multiple spatial dimension saliency local surface area remote sensing feature maps; wherein, the formula is:

[0035]

[0036] Among them, F t for each of the plurality of local surface area remote sensing feature maps, and is a one-dimensional convolution layer, and Pool(·) represents a pooling layer. That is, in the data processing process of the spatial perception feature fusion module, three 3×3 convolution layers are used to expand the receptive field and capture richer spatial information. The feature matrix after the convolution operation is mapped into a spatial attention matrix using the Sigmoid function. The result of the dot product of the input feature and the attention matrix is ​​the local area remote sensing feature map of the surface that is salient in the multiple spatial dimensions.

[0037] Specifically, the step S32 extracts essential features from the local area remote sensing feature maps of the multiple spatial dimensions to obtain the remote sensing essential semantic feature vector of the monitored area as the remote sensing essential semantic feature of the monitored area. Figure 4 As shown, the S32 includes: S321, performing overall optimization of feature distribution of the remote sensing feature maps of local areas of the surface that are significantly significant in multiple spatial dimensions to obtain multiple optimized remote sensing feature maps of local areas of the surface that are significantly significant in spatial dimensions; S322, using an essential feature extractor network to perform essential feature extraction on the multiple optimized remote sensing feature maps of local areas of the surface that are significantly significant in spatial dimensions to obtain the remote sensing essential semantic feature vector of the monitored area.

[0038] More specifically, in S321, the characteristic distribution of the plurality of spatially dimensionally significant local area remote sensing characteristic maps of the surface is comprehensively optimized to obtain a plurality of optimized spatially dimensionally significant local area remote sensing characteristic maps of the surface. It should be understood that the multiple local surface area remote sensing feature maps respectively express the image semantic features of the multiple remote sensing images, and after the spatial dimension expression of the multiple local surface area remote sensing feature maps is significantly expressed using the spatial perception feature fusion module, the spatial distribution of local image semantic features in the feature matrix can be strengthened. In this way, for the multiple spatial dimensionally significantly expressed local surface area remote sensing feature maps, considering the image source semantic distribution differences of the multiple remote sensing images and the local spatial distribution differences of image semantic features brought by the spatial perception feature fusion module, the multiple spatial dimensionally significantly expressed local surface area remote sensing feature maps are taken as a whole. The image semantic feature distribution differences of each feature matrix of the multiple spatial dimensionally significantly expressed local surface area remote sensing feature maps will directly affect the distribution integrity of the multiple spatial dimensionally significantly expressed local surface area remote sensing feature maps in units of feature matrix feature distribution, thereby affecting the expression effect of the monitored area remote sensing essential semantic feature vector obtained by using the essential feature extractor network to extract the essential features of the multiple spatial dimensionally significantly expressed local surface area remote sensing feature maps in units of the overall feature distribution of the feature matrix. Based on this, the applicant of the present application first converts each feature matrix in the multiple spatial dimension significant surface local area remote sensing feature maps into a square matrix through linear transformation, and then corrects the multiple spatial dimension significant surface local area remote sensing feature maps based on the optimization of adjacent feature matrices with feature matrices as units, which is specifically expressed as follows;

[0039]

[0040] Among them, M i and M i+1 are respectively the i-th and i+1-th feature matrices in the multiple spatial dimensions of the local area remote sensing feature map of the significant surface, and and They are the feature matrices M i and M i+1 The global mean of represents matrix multiplication, ⊙ represents point multiplication by position, M′ i+1is the i+1th feature matrix of the plurality of optimized spatial dimensionally significant local area remote sensing feature maps. Here, by taking the center of the feature matrix of the plurality of spatial dimensionally significant local area remote sensing feature maps along the channel distribution as the seed point of the scene transfer under the channel dimension, each eigenvalue of the feature matrix of the plurality of spatial dimensionally significant local area remote sensing feature maps is robustly aggregated and sub-sampled by matrix multiplication, so that the distribution bounding box transfer directional constraints of adjacent feature matrices are performed on the basis of the participation of each eigenvalue of the feature matrix of the plurality of spatial dimensionally significant local area remote sensing feature maps, so as to improve the integrity of the feature representation of the plurality of spatial dimensionally significant local area remote sensing feature maps on the basis of the overall contextual association of the plurality of spatial dimensionally significant local area remote sensing feature maps along the channel dimension from bottom to top, thereby improving the expression effect of the remote sensing essential semantic feature vector of the monitored area, so as to improve the accuracy of the classification result obtained by the classifier.

[0041] More specifically, in S322, the intrinsic feature extractor network is used to extract the intrinsic features of the multiple optimized spatial dimension salient local area remote sensing feature maps of the surface to obtain the remote sensing intrinsic semantic feature vector of the monitored area. That is, in the technical solution of the present application, the intrinsic feature extraction network is used to extract the intrinsic features of the multiple spatial dimension salient local area remote sensing feature maps of the surface to obtain the remote sensing intrinsic semantic feature vector of the monitored area. In the remote sensing image, the essential features of the surface include key information related to the environmental pollution risk, such as soil quality, water pollution, vegetation status, etc. The intrinsic feature extraction network is a deep learning network specially designed to extract high-level semantic features from images. By inputting multiple spatial dimension salient local area remote sensing feature maps of the surface into the intrinsic feature extraction network, the network can automatically learn and extract the essential features of the surface. These essential features can be used to describe the environmental pollution risk situation existing in the remote sensing image of the monitored area. By merging the essential features of multiple local areas of the surface into a feature vector, the remote sensing essential semantic feature vector of the monitored area can be obtained, which is used to evaluate and monitor the degree of water and soil environmental pollution risk. In the technical solution of the present application, the essential feature extractor network extracts the essential features of the remote sensing feature maps of the local area of ​​the surface that are significantly optimized in multiple spatial dimensions using the following formula; wherein the formula is:

[0042]

[0043] Among them, v i and v j The characteristic vector obtained by expanding the local area remote sensing characteristic map of the surface in the multiple optimized spatial dimensions, V kis the set of feature vectors obtained by expanding the remote sensing feature maps of the local area of ​​the surface in the above multiple optimized spatial dimensions, ‖·‖ 1 is the norm of the vector, M is the number of vectors in the set of the optimized channel-by-channel response multi-parameter time series correlation feature vectors - 1, D i The characteristic values ​​of each position in the characteristic vector obtained by expanding the remote sensing characteristic map of the local area of ​​the surface in the multiple optimized spatial dimensions, N s is the length of the remote sensing essential semantic feature vector of the monitored area, P k is the remote sensing essential semantic feature vector of the monitored area, and exp(·) is an exponential operation.

[0044] It is worth mentioning that in other specific examples of the present application, the multiple spatial dimensionally significant surface local area remote sensing feature maps can also be extracted by other methods to obtain the remote sensing essential semantic feature vector of the monitored area as the remote sensing essential semantic feature of the monitored area, for example: collect remote sensing feature map data of multiple surface local areas containing the monitored area. These feature maps can be obtained by different band combinations, different feature extraction methods or different preprocessing methods; each feature map is processed by spatial dimension significance to highlight the important features of the local surface area. This can be achieved by applying image enhancement technology, filters, local contrast enhancement and other methods; for the feature map processed by spatial dimension significance, image processing and computer vision technology are used to extract features. Various feature extraction methods can be used, such as color features, texture features, shape features, spatial features, etc. These features can be extracted by statistical methods, frequency domain analysis, morphological operations, etc.; the features extracted from each feature map are combined into a feature vector. Features can be spliced ​​together using simple splicing operations, or more complex feature fusion methods such as principal component analysis, multimodal fusion, etc. The generated feature vectors are normalized to ensure that the numerical ranges of different features are similar. Common normalization methods include maximum and minimum normalization, Z-score normalization, etc. The generated feature vectors are used as the remote sensing essential semantic features of the monitored area.

[0045] It is worth mentioning that in other specific examples of the present application, the remote sensing essential semantic features of the monitored area can also be determined based on the remote sensing features of the multiple local surface areas in other ways, for example: collect remote sensing image data of multiple local surface areas containing the monitored area; select multiple representative local surface areas in the monitored area according to actual needs and research purposes; for each local surface area, use image processing and computer vision technology to extract a series of remote sensing features. These features may include color features (such as color histogram, color moment), texture features (such as grayscale co-occurrence matrix, wavelet texture), shape features (such as contour, area), spatial features (such as relative position, neighboring relationship), etc.; fuse the features extracted from different local surface areas to obtain a comprehensive feature representation; according to actual needs, manually annotate the remote sensing images of the monitored area and assign corresponding semantic labels to each image position; use the remote sensing feature data of the local surface area with semantic labels to train a classification or regression model. Commonly used models include support vector machines, random forests, deep learning models, etc.; use the evaluation data set to evaluate the trained model; use the trained model to predict the remote sensing images of the monitored area and generate corresponding semantic features.

[0046] In particular, the S4 determines the risk level classification result of the controlled area based on the remote sensing essential semantic features of the monitored area. That is, in the technical solution of the present application, the remote sensing essential semantic feature vector of the monitored area is passed through a control area risk level classifier based on a classifier to obtain a risk level classification result, wherein the risk level classification result is used to represent the risk level of the controlled area. The purpose of risk level classification of the remote sensing essential semantic feature vector of the monitored area through a control area risk level classifier based on a classifier is to determine the degree of environmental pollution risk of the controlled area and represent it as a corresponding risk level. In a specific example, the remote sensing essential semantic feature vector of the monitored area is passed through a control area risk level classifier based on a classifier to obtain a risk level classification result, including: using multiple fully connected layers of the classifier to fully connect the remote sensing essential semantic feature vector of the monitored area to obtain a coded classification feature vector; and passing the coded classification feature vector through the Softmax classification function of the classifier to obtain the classification result. More specifically, the control area risk level classifier based on a classifier is a trained classifier that can judge the risk level of the area based on the input feature vector. This classifier can classify the monitored area into different risk levels, such as low risk, medium risk and high risk, based on the existing training data and feature vector information. By inputting the remote sensing essential semantic feature vector of the monitored area into the risk level classifier, the corresponding risk level classification result can be obtained. This result can represent the risk level of the controlled area and help environmental management departments understand and identify potential pollution risk areas. According to the risk level classification results, management departments can formulate corresponding control strategies and measures to reduce environmental pollution risks and protect the health of water and soil resources.

[0047] In summary, the method for zoning control of water and soil environmental pollution risks in karst cities according to the embodiment of the present application is explained, which uses the remote sensing images of the monitored area to characterize the water and soil environmental characteristics of the monitored area, and then uses the artificial intelligence algorithm based on deep learning to extract features and conduct comprehensive feature analysis on the remote sensing images of the monitored area to obtain the risk level assessment results of the controlled area. In this way, it can help decision makers and environmental management departments understand the risk status of the water and soil environment in karst areas, formulate control measures in a targeted manner, protect and improve the quality of the water and soil environment, and reduce the impact of environmental pollution risks on human health and ecosystems.

[0048] Furthermore, a karst urban water and soil environmental pollution risk zoning management and control system is also provided.

[0049] Figure 5 FIG. 1 is a block diagram of a karst urban water and soil environmental pollution risk zoning control system according to an embodiment of the present application. Figure 5As shown, according to the karst urban water and soil environmental pollution risk zoning control system 300 of the embodiment of the present application, it includes: an image acquisition module 310, used to acquire multiple remote sensing images of the controlled area; a local surface area remote sensing feature extraction module 320, used to extract multiple local surface area remote sensing features from the multiple remote sensing images; a monitored area remote sensing essential semantic extraction module 330, used to determine the remote sensing essential semantic features of the monitored area based on the multiple local surface area remote sensing features; and a level classification result generation module 340, used to determine the risk level classification result of the controlled area based on the remote sensing essential semantic features of the monitored area.

[0050] As described above, the karst city water and soil environmental pollution risk zoning control system 300 according to the embodiment of the present application can be implemented in various wireless terminals, such as a server with a karst city water and soil environmental pollution risk zoning control algorithm. In a possible implementation, the karst city water and soil environmental pollution risk zoning control system 300 according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or hardware module. For example, the karst city water and soil environmental pollution risk zoning control system 300 can be a software module in the operating system of the wireless terminal, or it can be an application developed for the wireless terminal; of course, the karst city water and soil environmental pollution risk zoning control system 300 can also be one of the many hardware modules of the wireless terminal.

[0051] Alternatively, in another example, the karst city water and soil environmental pollution risk zoning control system 300 and the wireless terminal may also be separate devices, and the karst city water and soil environmental pollution risk zoning control system 300 may be connected to the wireless terminal via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0052] The embodiments of the present disclosure have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for zoning control of water and soil environmental pollution risks in karst cities, characterized in that: include: Obtain multiple remote sensing images of the controlled area; Extracting a plurality of local area remote sensing features of the surface from the plurality of remote sensing images; Determining remote sensing essential semantic features of the monitored area based on the multiple local area remote sensing features of the surface; as well as Determine the risk level classification result of the controlled area based on the remote sensing essential semantic features of the monitored area; Based on the multiple local area remote sensing features of the surface, determining the remote sensing essential semantic features of the monitored area includes: Using a spatial perception feature fusion module to perform spatial dimension expression saliency on the plurality of local surface area remote sensing feature maps to obtain a plurality of spatial dimension saliency local surface area remote sensing feature maps; Extracting essential features from the remote sensing feature maps of the local area of ​​the surface that are significantly represented in the multiple spatial dimensions to obtain a remote sensing essential semantic feature vector of the monitored area as the remote sensing essential semantic feature of the monitored area; Extracting essential features from the multiple spatially salient local area surface remote sensing feature maps to obtain a remote sensing essential semantic feature vector of the monitored area as the remote sensing essential semantic feature of the monitored area includes: Performing overall optimization of feature distribution on the plurality of spatially dimensionally significant local surface remote sensing feature maps to obtain a plurality of optimized spatially dimensionally significant local surface remote sensing feature maps; Using an essential feature extractor network to extract essential features from the multiple optimized spatial dimension-signalized local area remote sensing feature maps to obtain the remote sensing essential semantic feature vector of the monitored area; The characteristic distribution of the plurality of spatially significant local area remote sensing characteristic maps of the surface is comprehensively optimized to obtain a plurality of optimized spatially significant local area remote sensing characteristic maps of the surface, including: Each feature matrix in the plurality of spatially dimensionally significant local area remote sensing feature maps is converted into a square matrix through a linear transformation, and then the plurality of spatially dimensionally significant local area remote sensing feature maps are corrected based on the optimization of adjacent feature matrices with feature matrices as units, which is specifically expressed as follows; in and They are the first and feature matrix, and and They are the feature matrices and The global mean of represents matrix multiplication, It means point multiplication by position. The multiple optimized spatial dimensions are used to significantly optimize the local area remote sensing feature map of the surface. feature matrix, express The transposed matrix of express Take the reciprocal of each eigenvalue in .

2. The method for zoning control of karst urban water and soil environmental pollution risk according to claim 1 is characterized in that: Extracting multiple local area remote sensing features of the surface from the multiple remote sensing images includes: The multiple remote sensing images are respectively passed through a control area surface state feature extractor based on a convolutional neural network model to obtain multiple surface local area remote sensing feature maps as the multiple surface local area remote sensing features.

3. The method for zoning control of karst urban water and soil environmental pollution risk according to claim 2 is characterized in that: The control area surface state feature extractor based on the convolutional neural network model is a pyramid network.

4. The method for zoning control of karst urban water and soil environmental pollution risk according to claim 3 is characterized in that: Using a spatial perception feature fusion module to perform spatial dimension expression saliency on the multiple local surface area remote sensing feature maps to obtain multiple spatial dimension salient local surface area remote sensing feature maps, including: using the spatial perception feature fusion module to perform spatial dimension expression saliency on the multiple local surface area remote sensing feature maps using the following spatial perception formula to obtain the multiple spatial dimension salient local surface area remote sensing feature maps; Wherein, the spatial perception formula is: in, for each of the plurality of local surface area remote sensing feature maps, , and is a one-dimensional convolutional layer, represents the pooling layer, is the spatial attention matrix, Indicates based on Nonlinear activation processing of functions.

5. The method for zoning control of karst urban water and soil environmental pollution risk according to claim 4 is characterized in that: Using an essential feature extractor network to extract essential features from the plurality of optimized spatial dimensionally significant local surface area remote sensing feature maps to obtain the monitored area remote sensing essential semantic feature vector, including: using an essential feature extractor network to extract essential features from the plurality of optimized spatial dimensionally significant local surface area remote sensing feature maps using the following formula; Wherein, the formula is: in, and The feature vectors obtained by expanding the remote sensing feature maps of the local area of ​​the surface in the multiple optimized spatial dimensions are significant, is a set of feature vectors obtained by expanding the remote sensing feature maps of the local area of ​​the surface in the multiple optimized spatial dimensions, is the one-norm of the vector, To optimize the number of vectors in the set of multi-parameter time series correlation feature vectors for channel-by-channel response, The characteristic values ​​of each position in the characteristic vector obtained by expanding the local area remote sensing characteristic map of the surface for the multiple optimized spatial dimensions are significant, is the length of the remote sensing essential semantic feature vector of the monitored area, is the remote sensing essential semantic feature vector of the monitored area, For exponential operation.

6. The method for zoning control of karst urban water and soil environmental pollution risk according to claim 5 is characterized in that: Based on the remote sensing essential semantic features of the monitored area, the risk level classification result of the controlled area is determined, including: The remote sensing essential semantic feature vector of the monitored area is passed through a control area risk level divider based on a classifier to obtain the risk level division result, wherein the risk level division result is used to represent the risk level of the controlled area.

7. A karst urban water and soil environmental pollution risk zoning control system, characterized in that: include: An image acquisition module is used to acquire multiple remote sensing images of the controlled area; A surface local area remote sensing feature extraction module, used to extract multiple surface local area remote sensing features from the multiple remote sensing images; A monitored area remote sensing essential semantic extraction module, used to determine the remote sensing essential semantic features of the monitored area based on the multiple surface local area remote sensing features; as well as A level classification result generating module is used to determine the risk level classification result of the controlled area based on the remote sensing essential semantic features of the monitored area; The remote sensing essential semantic extraction module of the monitored area is used to: Using a spatial perception feature fusion module to perform spatial dimension expression saliency on the plurality of local surface area remote sensing feature maps to obtain a plurality of spatial dimension saliency local surface area remote sensing feature maps; Extracting essential features from the remote sensing feature maps of the local area of ​​the surface that are significantly represented in the multiple spatial dimensions to obtain a remote sensing essential semantic feature vector of the monitored area as the remote sensing essential semantic feature of the monitored area; The remote sensing essential semantic extraction module of the monitored area is used to: Performing overall optimization of feature distribution on the plurality of spatially dimensionally significant local surface remote sensing feature maps to obtain a plurality of optimized spatially dimensionally significant local surface remote sensing feature maps; Using an essential feature extractor network to extract essential features from the multiple optimized spatial dimension-signalized local area remote sensing feature maps to obtain the remote sensing essential semantic feature vector of the monitored area; The remote sensing essential semantic extraction module of the monitored area is used to: Each feature matrix in the plurality of spatially dimensionally significant local area remote sensing feature maps is converted into a square matrix through a linear transformation, and then the plurality of spatially dimensionally significant local area remote sensing feature maps are corrected based on the optimization of adjacent feature matrices with feature matrices as units, which is specifically expressed as follows; in and They are the first and feature matrix, and and They are the feature matrices and The global mean of represents matrix multiplication, It means point multiplication by position. The multiple optimized spatial dimensions are used to significantly optimize the local area remote sensing feature map of the surface. feature matrix, express The transposed matrix of express Take the reciprocal of each eigenvalue in .

Images