Risk assessment method and system for natural resources

Abstract

The invention discloses a risk assessment method and system for natural resources, and the method comprises the steps: capturing public transaction data through collecting regional image data; performing feature recognition and classification statistics on natural resources to obtain natural resource data; performing risk assessment on the natural resource data to obtain risk assessment information; and constructing an evaluation optimization model, importing the risk evaluation information and the natural resource data into the evaluation optimization model, generating a growth trend of the natural resource number, and giving an optimization suggestion of the target area according to the growth trend. According to the risk assessment method for the natural resources, the natural resources in the target area can be rapidly assessed in a large range, the growth trend of the natural resources in the target area is obtained, and optimization suggestions of the target area are given according to the growth trend of the natural resources.

Description

Technical Field

[0001] The present invention belongs to the technical field of natural resource risk assessment, and particularly relates to a risk assessment method and system for natural resources. Background Art

[0002] Based on the comprehensive analysis method of remote sensing technology, environmental monitoring and artificial intelligence, it has been widely applied in the field of natural resource management. Natural resources such as forests, grasslands, water bodies and soils are important components of the ecosystem, and their distribution, status and growth trends directly affect the development of the ecological environment and human society. By standardizing the processing of regional image data and feature recognition, information such as resource types and distribution ranges can be efficiently extracted, providing basic data support for scientific management.

[0003] At the same time, environmental monitoring data (such as meteorology, pollutant concentration, etc.) can be used as important references to reflect the dynamic environmental conditions of the target area. Combining natural resource data for risk assessment helps to identify potential threats (such as drought, pollution or soil degradation), so as to take timely countermeasures.

[0004] In this context, how to integrate multi-source data and conduct in-depth analysis on it, so as to achieve the optimized management of the target area and improve the utilization efficiency of natural resources. Summary of the Invention

[0005] The purpose of the embodiments of the present invention is to provide a risk assessment method and system for natural resources, aiming to solve the problems raised in the background art.

[0006] To achieve the above purpose, the embodiments of the present invention provide the following technical solutions: A risk assessment method for natural resources, the method specifically includes the following steps: Taking satellite remote sensing as the main method and drone shooting as the auxiliary method, collect regional image data of the target natural area; Perform data preprocessing on the regional image data to obtain standard image data, and conduct feature recognition and classification statistics of natural resources to obtain natural resource data; Obtain environmental monitoring data of the target natural area, and based on the environmental monitoring data, conduct risk assessment on the natural resource data to obtain risk assessment information; Construct an evaluation optimization model, import the risk assessment information and natural resource data into the evaluation optimization model, output the growth trend of natural resources, and generate optimization suggestions for the target area according to the growth trend of natural resources.

[0007] A risk assessment system for natural resources, the system includes a relevant data collection unit, a natural resource identification unit, an environmental risk assessment unit and a model construction and optimization unit, where: A relevant data acquisition unit, which mainly uses satellite remote sensing and supplemented by drone photography to acquire regional image data of the target natural area; A natural resource identification unit, which is used to perform data preprocessing on the regional image data to obtain standard image data, and perform feature identification, classification and statistics of natural resources to obtain natural resource data; An environmental risk assessment unit, which is used to obtain environmental monitoring data of the target natural area, and based on the environmental monitoring data, perform risk assessment on the natural resource data to obtain risk assessment information; A model construction and optimization unit, which is used to construct an evaluation and optimization model, import the risk assessment information and natural resource data into the evaluation and optimization model, output the growth trend of natural resources, and generate optimization suggestions for the target area according to the growth trend of natural resources.

[0008] Compared with the prior art, the beneficial effects of the present invention are: In the embodiment of the present invention, by collecting regional image data, performing feature identification, classification and statistics of natural resources to obtain natural resource data; performing risk assessment on natural resource data to obtain risk assessment information; constructing an evaluation and optimization model, importing risk assessment information and natural resource data into the evaluation and optimization model, and generating the growth trend of natural resource data. The present invention combines remote sensing technology and drone images, and according to the environmental characteristics of the target area, can quickly evaluate the natural resources in the target area on a large scale, obtain the growth trend of the natural resources in the target area, and give optimization suggestions for the target area according to the growth trend of the natural resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention.

[0010] Figure 1 The flowchart of the method provided by the embodiment of the present invention is shown. DETAILED DESCRIPTION OF THE INVENTION

[0011] In order to make the objectives, technical solutions and advantages of the present invention clearer, the following further describes the present invention in detail with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, and are not used to limit the present invention.

[0012] To solve the above problems, in the embodiments of the present invention, satellite remote sensing is mainly used, supplemented by drone photography, to collect regional image data of the target natural area; the regional image data is preprocessed to obtain standard image data, and feature recognition and classification statistics of natural resources are carried out to obtain natural resource data; environmental monitoring data of the target natural area is obtained, and based on the environmental monitoring data, risk assessment of the natural resource data is carried out to obtain risk assessment information; an evaluation optimization model is constructed, the risk assessment information and the natural resource data are imported into the evaluation optimization model, the growth trend of natural resources is output, and corresponding optimization suggestions for the target area are generated according to the growth trend of natural resources.

[0013] Figure 1 The flowchart of the method provided by the embodiments of the present invention is shown.

[0014] Specifically, a risk assessment method for natural resources, the method specifically includes the following steps: Step S101, mainly using satellite remote sensing and supplemented by drone photography, collect regional image data of the target natural area.

[0015] In the embodiments of the present invention, based on satellite remote sensing technology, remote sensing image data of the target natural area is collected, and remote sensing quality analysis is carried out on the remote sensing image data to determine multiple low-quality sub-regions, and then according to the geographical locations and ranges of the multiple low-quality sub-regions, drone shooting control is carried out to obtain multiple supplementary shooting images, and by comprehensively sorting out the remote sensing image data and the multiple supplementary shooting images, regional image data of the target natural area is generated.

[0016] Among them, in the preferred embodiment provided by the present invention, the mainly using satellite remote sensing and supplemented by drone photography to collect regional image data of the target natural area specifically includes the following steps: Step S1011, based on satellite remote sensing technology, collect remote sensing image data of the target natural area; Step S1012, perform quality analysis on the remote sensing image data to determine multiple low-quality sub-regions; Step S1013, carry out drone shooting control on the multiple low-quality sub-regions to obtain multiple supplementary shooting images; Step S1014, comprehensively combine the remote sensing image data and the multiple supplementary shooting images to generate regional image data of the target natural area.

[0017] Among them, in the preferred embodiment provided by the present invention, the comprehensively combining the remote sensing image data and the multiple supplementary shooting images to generate regional image data of the target natural area specifically includes the following steps: Perform spectral correction and geometric correction operations on the remote sensing image data in sequence to obtain a preprocessed remote sensing image; Perform geometric correction and color adjustment operations on the supplementary captured images in sequence to obtain preprocessed supplementary captured images; Use spectral analysis method to extract the preprocessed remote sensing images to obtain a spectral feature matrix; Use a convolutional neural network to extract multi-scale texture information of the preprocessed remote sensing images to obtain a spatial feature matrix; Analyze the change trend of the preprocessed remote sensing images at different time points to obtain a time feature matrix; Use the Canny edge detection algorithm to extract the supplementary captured images to obtain a geometric feature matrix; Perform semantic segmentation on the preprocessed supplementary captured images using the U-Net model to obtain a semantic feature matrix; Enhance the supplementary captured images using the local contrast enhancement algorithm to obtain local enhancement features; Use the spectral feature matrix to perform global similarity analysis on the remote sensing images and the supplementary captured images to obtain the overlapping regions; Use the boundary information of the spatial feature matrix to match the key points within the overlapping regions to obtain a preliminary alignment result; Use the geometric feature matrix to refine the preliminary alignment result for local correction to obtain a local correction result; Enhance the high-resolution regions in the local correction result using the local enhancement features to obtain a fine alignment result; Select the actually consistent stable regions from the fine alignment result according to the annotation information in the semantic feature matrix as the reference points for time registration and use the time feature matrix for adjustment to obtain the aligned remote sensing images and captured images; Use wavelet transform to fuse the spectral values of the remote sensing images in the aligned remote sensing images and captured images with the pixel details of the captured images to obtain an image containing spectral information and high-resolution texture to generate the regional image data of the target natural region.

[0018] In the embodiments of the present invention, multi-dimensional feature information such as space, spectrum, time, and semantics is combined to ensure the multi-dimensional consistency of the final image. By fusing multi-modal data, the generated comprehensive image has the characteristics of high resolution, high spectral coverage, and high semantic accuracy. It can monitor the ground objects in the natural region more comprehensively, such as forest coverage rate, river network changes, etc.

[0019] Among them, in the preferred embodiment provided by the present invention, the step of performing semantic segmentation on the preprocessed supplementary captured images using the U-Net model to obtain a semantic feature matrix specifically includes the following steps: Step S101401, input the preprocessed supplementary captured images into the encoder, and perform feature extraction on the preprocessed supplementary captured images through a convolutional layer with a convolutional kernel size of 3×3 to obtain local features; Step S101402: Perform a normalization operation on the local features to obtain a normalized result, and then apply the ReLU activation function to perform a non-linear transformation on the normalized result to obtain the transformed features; Step S101403: Then downsample the transformed features by using a max pooling operation to obtain a downsampled feature map; Step S101404: Repeat steps S101401 to S101403 in an iterative manner to obtain a low-resolution feature map; Step S101405: Input the low-resolution feature map into the decoder, perform feature extraction on the low-resolution feature map by using dilated convolutions with different kernel sizes to obtain features of different scales, and fuse the features of different scales to obtain a multi-scale feature map; Step S101406: Perform a transposed convolution operation on the multi-scale feature map to obtain an upsampled feature map; Step S101407: Fuse the upsampled features with the outputs corresponding to the encoder layers by using skip connections to obtain a connected feature map; Step S101408: Perform convolution operations, batch normalization operations, and non-linear activation operations on the connected feature map in sequence to obtain a reconstructed feature map; Step S101409: Repeat steps S101405 to S101408 in an iterative manner until the spatial size of the original input image is restored to obtain a final feature map; Step S101410: Perform a global pooling operation on the final feature map to obtain global features; Step S101411: Generate attention weights by passing the global features through a set of convolution operations, and multiply the attention weights with the final feature map pixel by pixel to obtain an enhanced feature map; Step S101412: Pass the enhanced feature map through a 1×1 convolutional layer to perform dimensionality reduction to obtain a dimensionality-reduced feature map; Step S101413: Apply the Softmax function to each pixel in the dimensionality-reduced feature map to convert the output of the pixel into a probability distribution for each category, and finally obtain a semantic feature matrix.

[0020] Among them, in the preferred embodiment provided by the present invention, the step of selecting an actually consistent stable region from the fine alignment result according to the annotation information in the semantic feature matrix as a reference point for time registration and using the time feature matrix for adjustment to obtain an aligned remote sensing image and a captured image specifically includes the following steps: Set a stability threshold, compare the category probability value of each pixel point in the semantic feature matrix with the stability threshold, and use the pixel points greater than the stability threshold as reference points; Calculate the absolute difference of pixel pairs in the fine alignment result, and use the pixels whose absolute difference is less than the stability threshold and the class probability value of the pixel points is greater than the stability threshold as stable pixel points, mark the stable pixel points to form a stability mask; Perform Gaussian smoothing on the time feature matrix to obtain a smoothed time feature matrix; Use the local search method to find several stable pixels adjacent to each reference point in the stability mask to obtain a set of matching points; Calculate the mean value of the time feature values of the pixel points in the set of matching points to obtain the time change trend; Adjust the pixel values of the reference points in the remote sensing image and the UAV image according to the time change trend to obtain the aligned remote sensing image and the captured image.

[0021] In the embodiment of the present invention, semantic information, time features, and spatial features are combined to achieve high-precision alignment of images. This method can make full use of various feature information, starting from different dimensions, comprehensively considering the complexity of image data, and ensuring the stability and accuracy of the registration result. When processing multi-source data such as remote sensing images and UAV images, this comprehensiveness makes the registration method more robust and can process images from different times, different devices, and different resolutions. And by extracting reference points through the semantic feature matrix, it is ensured that the selected areas have relatively high class credibility, avoiding noise or errors that may be introduced by low-credibility areas. Using the smoothed trend of the time feature matrix to reduce noise interference further improves the reliability and accuracy of the adjustment. By means of local matching (such as matching key points in the stability mask), the consistency of the local area is ensured, avoiding the accumulation of errors caused by global adjustment.

[0022] Furthermore, the risk assessment method for natural resources further includes the following steps: Step S102, perform data preprocessing on the regional image data to obtain standard image data, and perform feature recognition and classification statistics of natural resources to obtain natural resource data.

[0023] In the embodiment of the present invention, redundant, incorrect, and irrelevant useless data in the regional image data are identified, and the useless data in the regional image data are removed to obtain effective image data. Then, the effective image data are enhanced, denoised, and standardized to obtain standard image data, so that the data from different sources in the standard image data have consistent formats and units. Furthermore, feature recognition such as texture, color, and shape is performed on the standard image data to obtain feature recognition data, and then classification statistics of natural resources such as forests, grasslands, wetlands, and / or minerals are performed on the feature recognition data to obtain the target natural area and obtain natural resource data.

[0024] Among them, in the preferred embodiment provided by the present invention, the data preprocessing is performed on the regional image data to obtain standard image data, and the feature recognition and classification statistics of natural resources are performed, and the acquisition of natural resource data specifically includes the following steps: Step S1021, removing useless data from the regional image data to obtain valid image data; Step S1022, performing enhancement, denoising and standardization processing on the effective image data to obtain standard image data; Step S1023, performing feature recognition on the standard image data to obtain feature recognition data; Step S1024, classify and count the feature recognition data to obtain natural resource data.

[0025] Among them, in the preferred embodiment provided by the present invention, the performing feature recognition on the standard image data and obtaining feature recognition data specifically comprises the following steps: Perform dimensionality reduction operation on standard image data to obtain the feature matrix after dimensionality reduction; The feature matrix after dimension reduction is subjected to multi-level feature extraction operation by using convolutional neural network to obtain low-level features, intermediate features and high-level semantic features, and the low-level features, intermediate features and high-level semantic features are subjected to multi-scale fusion operation to obtain the first spatial features; The feature matrix after dimension reduction is subjected to feature extraction operation using a recursive neural network to capture the dynamic change pattern in the time series and obtain the first time feature; The first spatial feature and the first temporal feature are fused to generate a comprehensive feature matrix to obtain multi-level key features of the image; Obtaining a context description of standard image data, and encoding the context description into a context vector; Generate query vector, key vector and value vector from multi-level key features; The context vector is weighted and fused with the query vector and key vector by element-by-element multiplication to generate an enhanced query vector and an enhanced key vector. The dot product of the enhanced query vector and the enhanced key vector is calculated to obtain the attention score matrix, and the attention score matrix is ​​weighted on the value vector to obtain the feature recognition data.

[0026] In an embodiment of the present invention, a context awareness mechanism is introduced. By encoding context information into a global vector, the global vector can include information such as geographical location (e.g., longitude and latitude), historical environmental characteristics (e.g., cumulative rainfall, average annual temperature), etc. Then, the context vector is jointly calculated with the query vector and the key vector, so that the weight of the feature not only depends on the local features of the input image, but is also affected by context information such as geography, climate, and history. Furthermore, the present invention can improve the sensitivity to the dynamic changes of natural resources, thereby capturing features more accurately. Therefore, the context-aware dynamic attention of the present invention dynamically adjusts the attention weight by introducing external context information, making it more adaptable to different scenarios.

[0027] Furthermore, the risk assessment method for the natural resources further includes the following steps: Step S103: Obtain the environmental monitoring data of the target natural area. Based on the environmental monitoring data, conduct a risk assessment on the natural resources data to obtain risk assessment information.

[0028] In an embodiment of the present invention, environmental monitoring is carried out on the target natural area to obtain the environmental monitoring data of the target natural area. Based on the environmental monitoring data, an impact analysis is conducted on the natural resources data to determine the impact type, impact time, impact scope, and impact degree. Then, according to the impact type, impact time, impact scope, and impact degree, a risk assessment is conducted on the natural resources data to obtain risk assessment information, such as risk assessment of drought, risk assessment of typhoon, etc.

[0029] Among them, in the preferred embodiment provided by the present invention, the steps of obtaining the environmental monitoring data of the target natural area, conducting a risk assessment on the natural resources data based on the environmental monitoring data, and obtaining risk assessment information specifically include the following steps: Step S1031: Obtain the environmental monitoring data of the target natural area; Step S1032: Based on the environmental monitoring data, conduct an impact analysis on the natural resources data to determine the impact type, impact time, impact scope, and impact degree; Step S1033: According to the impact type, the impact time, the impact scope, and the impact degree, conduct a risk assessment on the natural resources data to obtain risk assessment information.

[0030] Furthermore, the risk assessment method for the natural resources further includes the following steps: Step S104: Construct an evaluation optimization model, import the risk assessment information and the natural resources data into the evaluation optimization model, output the growth trend of the natural resources, and generate an optimization suggestion for the target area according to the growth trend of the natural resources.

[0031] In an embodiment of the present invention, by importing risk assessment information and natural resource data into an evaluation and optimization model, the natural resource data is automatically analyzed through the evaluation and optimization model to obtain the growth trend of natural resources, and then, based on the growth trend of natural resources, optimization suggestions for the target natural area are generated.

[0032] Among them, in a preferred embodiment provided by the present invention, the steps of importing the risk assessment information and natural resource data into the evaluation and optimization model, outputting the growth trend of natural resources, and generating optimization suggestions for the target area according to the growth trend of natural resources specifically include the following steps: Step S1051: Import the natural resource data into the evaluation and optimization model; Step S1052: Obtain the growth trend of natural resources through the evaluation and optimization model; Step S1053: Generate optimization suggestions for the target natural area according to the growth trend of natural resources.

[0033] In another preferred embodiment provided by the present invention, a risk assessment system for natural resources is proposed. The system includes: A relevant data acquisition unit 101, which is mainly used for satellite remote sensing and supplemented by drone photography to collect regional image data of the target natural area.

[0034] A natural resource identification unit 102, which is used to perform data preprocessing on the regional image data to obtain standard image data, and perform feature identification and classification statistics of natural resources to obtain natural resource data.

[0035] An environmental risk assessment unit 103, which is used to obtain environmental monitoring data of the target natural area, and based on the environmental monitoring data, perform risk assessment on the natural resource data to obtain risk assessment information.

[0036] A model construction and optimization unit 104, which is used to construct an evaluation and optimization model, import the risk assessment information and natural resource data into the evaluation and optimization model, output the growth trend of natural resources, and generate optimization suggestions for the target area according to the growth trend of natural resources.

[0037] It should be understood that although the steps in the flowcharts of the embodiments of the present invention are shown in sequence according to the indications of the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear indication in this article, there is no strict order limit for the execution of these steps, and these steps can be executed in other orders. Moreover, at least a part of the steps in each embodiment may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or sub-steps or stages of other steps.

[0038] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0039] The technical features of the above-described embodiments can 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, it should be considered as the scope described in this specification.

[0040] The above-described embodiments merely represent several implementation manners of the present invention. The description thereof is relatively specific and detailed, but it should not be construed as a limitation to the scope of the patent for the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several variations and improvements can still be made, and these all fall within the protection scope of the present invention. Therefore, the protection scope of the patent for the present invention shall be subject to the appended claims.

[0041] The above is only the preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims

1. A natural resource risk assessment method, characterized in that: The method specifically comprises the following steps: Using satellite remote sensing as the main method and drone photography as the auxiliary method, regional image data of the target natural area is collected; Preprocessing the regional image data to obtain standard image data, and performing feature recognition and classification statistics of natural resources to obtain natural resource data; Acquire environmental monitoring data of a target natural area, and based on the environmental monitoring data, perform risk assessment on the natural resource data to obtain risk assessment information; Constructing an assessment optimization model, importing the risk assessment information and natural resource data into the assessment optimization model, outputting the growth trend of natural resources, and generating target area optimization suggestions based on the growth trend of natural resources; The method of collecting regional image data of the target natural area mainly by satellite remote sensing and supplemented by drone photography specifically includes the following steps: Based on satellite remote sensing technology, collect remote sensing image data of target natural areas; Performing quality analysis on the remote sensing image data to determine a plurality of low-quality sub-regions; Controlling the drone to shoot the multiple low-quality sub-areas to obtain multiple supplementary shooting images; The remote sensing image data and the plurality of supplementary photographed images are integrated to generate regional image data of the target natural area.

2. A natural resource risk assessment method according to claim 1, characterized in that: The step of integrating the remote sensing image data and the plurality of supplementary captured images to generate regional image data of the target natural area specifically comprises the following steps: Perform spectral correction and geometric correction operations on the remote sensing image data in sequence to obtain a preprocessed remote sensing image; performing geometric correction and color adjustment operations on the supplementary captured image in sequence to obtain a pre-processed supplementary captured image; The spectral analysis method is used to extract the pre-processed remote sensing image to obtain the spectral feature matrix; Convolutional neural network is used to extract multi-scale texture information of pre-processed remote sensing images and obtain spatial feature matrix; Analyze the changing trends of preprocessed remote sensing images at different time points to obtain the time feature matrix; The Canny edge detection algorithm is used to extract the supplementary shooting images to obtain the geometric feature matrix; The preprocessed supplementary images are semantically segmented using the U-Net model to obtain a semantic feature matrix; The supplementary captured image is enhanced using a local contrast enhancement algorithm to obtain local enhancement features; The spectral feature matrix is ​​used to perform global similarity analysis on remote sensing images and supplementary shooting images to obtain overlapping areas; Using the boundary information of the spatial feature matrix, the key points in the overlapping area are matched to obtain the preliminary alignment result; The preliminary alignment result is refined by using the geometric feature matrix to perform local correction and obtain a local correction result; The high-resolution area in the local correction result is enhanced using the local enhancement feature to obtain a fine alignment result; According to the annotation information in the semantic feature matrix as the reference point for temporal registration, the stable area that is actually consistent is selected from the fine alignment result and adjusted using the temporal feature matrix to obtain the aligned remote sensing image and the photographed image; Wavelet transform is used to fuse the spectral values ​​of the aligned remote sensing images and the pixel details of the captured images to obtain an image containing spectral information and high-resolution texture to generate regional image data of the target natural area.

3. A natural resource risk assessment method according to claim 2, characterized in that: The method of performing semantic segmentation on the preprocessed supplementary captured images using a U-Net model to obtain a semantic feature matrix specifically includes the following steps: Step S101401, input the pre-processed supplementary captured image into the encoder, perform feature extraction on the pre-processed supplementary captured image through a convolution layer with a convolution kernel size of 3×3 to obtain local features; Step S101402, performing a normalization operation on the local features to obtain a normalized result, and then applying a ReLU activation function to perform a nonlinear transformation on the normalized result to obtain a transformed feature; Step S101403, the transformed features are then reduced using a maximum pooling operation to obtain a downsampled feature map; Step S101404, repeating steps S101401 to S101403 in an iterative manner to obtain a low-resolution feature map; Step S101405, inputting the low-resolution feature map into the decoder, performing feature extraction on the low-resolution feature map using dilated convolutions with convolution kernels of different sizes, obtaining features of different scales, and fusing the features of different scales to obtain a multi-scale feature map; Step S101406, performing a deconvolution operation on the multi-scale feature map to obtain an upsampled feature map; Step S101407, fusing the upsampled features with the output corresponding to the number of encoder layers using a skip connection to obtain a connected feature map; Step S101408, performing convolution operation, batch normalization operation and nonlinear activation operation on the connected feature map in sequence to obtain a reconstructed feature map; Step S101409, repeating steps S101405 to S101408 in an iterative manner until the spatial size of the original input image is restored to obtain a final feature map; Step S101410, performing a global pooling operation on the final feature map to obtain a global feature; Step S101411, generating attention weights through a set of convolution operations on the global features, and multiplying the attention weights by the final feature map pixel by pixel to obtain an enhanced feature map; Step S101412, passing the enhanced feature map through a 1×1 convolution layer to reduce the dimension, thereby obtaining a feature map after dimension reduction; Step S101413, applying the Softmax function to each pixel in the feature map after dimensionality reduction, converting the pixel output into the probability distribution of each category, and finally obtaining the semantic feature matrix.

4. A natural resource risk assessment method according to claim 3, characterized in that: The method uses the annotation information in the semantic feature matrix as a reference point for temporal registration, selects an actually consistent stability region from the fine alignment result, and uses the temporal feature matrix to adjust the region to obtain an aligned remote sensing image and a captured image, specifically comprising the following steps: Set a stability threshold, compare the category probability value of each pixel in the semantic feature matrix with the stability threshold, and use the pixel points greater than the stability threshold as reference points; Calculate the absolute difference of the pixel pairs in the fine alignment result, and take the pixels whose absolute difference is less than the stability threshold and whose category probability value is greater than the stability threshold as stable pixels, mark the stable pixels, and form a stability mask; Gaussian smoothing is applied to the time feature matrix to obtain a smoothed time feature matrix; A local search method is used to find several stable pixels close to each reference point in the stability mask to obtain a set of matching points. Calculate the mean of the time feature values ​​of the pixels in the matching point set to obtain the time change trend; The pixel values ​​of the reference points in the remote sensing image and the UAV image are adjusted according to the temporal variation trend to obtain aligned remote sensing images and photographed images.

5. A natural resource risk assessment method according to claim 4, characterized in that: The data preprocessing of the regional image data to obtain standard image data, and the feature recognition and classification statistics of natural resources are performed to obtain natural resource data specifically include the following steps: Eliminating useless data in the image data of the region to obtain valid image data; Perform enhancement, denoising and standardization on the effective image data to obtain standard image data; Performing feature recognition on the standard image data to obtain feature recognition data; The feature recognition data are classified and counted to obtain natural resource data.

6. A natural resource risk assessment method according to claim 5, characterized in that: The performing feature recognition on the standard image data to obtain feature recognition data specifically comprises the following steps: Perform dimensionality reduction operation on standard image data to obtain the feature matrix after dimensionality reduction; The feature matrix after dimension reduction is subjected to multi-level feature extraction operation by using convolutional neural network to obtain low-level features, intermediate features and high-level semantic features, and the low-level features, intermediate features and high-level semantic features are subjected to multi-scale fusion operation to obtain the first spatial features; The feature matrix after dimension reduction is subjected to feature extraction operation using a recursive neural network to capture the dynamic change pattern in the time series and obtain the first time feature; The first spatial feature and the first temporal feature are fused to generate a comprehensive feature matrix to obtain multi-level key features of the image; Obtaining a context description of standard image data, and encoding the context description into a context vector; Generate query vector, key vector and value vector from multi-level key features; The context vector is weighted and fused with the query vector and key vector by element-by-element multiplication to generate an enhanced query vector and an enhanced key vector. The dot product of the enhanced query vector and the enhanced key vector is calculated to obtain the attention score matrix, and the attention score matrix is ​​weighted on the value vector to obtain the feature recognition data.

7. A natural resource risk assessment method according to claim 6, characterized in that: The step of obtaining environmental monitoring data of a target natural area, and performing risk assessment on the natural resource data based on the environmental monitoring data, and obtaining risk assessment information specifically comprises the following steps: Obtain environmental monitoring data for target natural areas; Based on the environmental monitoring data, an impact analysis is performed on the natural resource data to determine the impact type, impact time, impact scope and impact degree; According to the impact type, the impact time, the impact scope and the impact degree, a risk assessment is performed on the natural resource data to obtain risk assessment information.

8. A natural resource risk assessment method according to claim 7, characterized in that: The step of importing the risk assessment information and natural resource data into the assessment optimization model, outputting the growth trend of natural resources, and generating target area optimization suggestions according to the growth trend of natural resources specifically includes the following steps: Importing the natural resource data into the evaluation optimization model; Obtaining the growth trend of natural resources through the evaluation and optimization model; Based on the growth trend of the natural resources, an optimization suggestion for the target natural area is generated.

9. A natural resource risk assessment system, the system being applied to the natural resource risk assessment method according to any one of claims 1 to 8, characterized in that: The system includes a relevant data collection unit, a natural resource identification unit, an environmental risk assessment unit and a model building optimization unit, wherein: Related data collection units are used to collect regional image data of target natural areas mainly through satellite remote sensing and supplemented by drone photography; A natural resource identification unit is used to perform data preprocessing on the regional image data to obtain standard image data, and perform feature identification and classification statistics of natural resources to obtain natural resource data; An environmental risk assessment unit, used to obtain environmental monitoring data of a target natural area, and based on the environmental monitoring data, perform risk assessment on the natural resource data to obtain risk assessment information; The model building optimization unit is used to build an assessment optimization model, import the risk assessment information and natural resource data into the assessment optimization model, output the growth trend of natural resources, and generate target area optimization suggestions based on the growth trend of natural resources.

Images