Biomass carbon reserve inversion method and device, electronic equipment and storage medium

Abstract

The invention discloses a biomass carbon reserve inversion method and device, electronic equipment and a storage medium, and the method can effectively improve the precision and details of a generative graph through the application of a two-stage generative adversarial network, and can further solve a problem that a single generative adversarial network is difficult to generate detailed data. By combining sparse representation, a generative adversarial network and an unmanned aerial vehicle hyperspectral imaging technology, the efficiency and accuracy of biomass and carbon reserve inversion are remarkably improved, dimension reduction processing is performed on hyperspectral data through a sparse representation technology, a sparse-based dictionary capable of effectively extracting key spectral information is constructed, and a sparse coding algorithm is combined, so that the efficiency and accuracy of biomass and carbon reserve inversion are improved. According to the method, the data redundancy and calculation complexity are greatly reduced, the data processing speed is improved, important feature information in the hyperspectral data is reserved, the subsequent inversion process is more efficient and accurate, and the method can be widely applied to the technical field of data processing.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a biomass carbon stock inversion method, device, electronic equipment and storage medium. Background Art

[0002] With the increasing severity of environmental protection and climate change issues, the global demand for monitoring and assessment of biomass and carbon stocks is increasing. As important indicators for measuring the health of forest and grassland ecosystems, biomass and carbon stocks are of great significance for the study of carbon cycle and climate change. At present, remote sensing monitoring using drones equipped with hyperspectral sensors has become an effective means of obtaining biomass and carbon stock data over large areas. However, traditional hyperspectral data processing methods often face some technical difficulties and challenges when applied to biomass and carbon stock inversion.

[0003] First of all, existing hyperspectral data processing methods usually need to process a large amount of high-dimensional data. The data has high dimensions and much redundant information, resulting in high computational complexity and low data processing efficiency. In this case, even if high-precision biomass and carbon stock data can be obtained, it is often accompanied by extremely high computational costs, which affects the real-time and practicality of the inversion results. In addition, since traditional methods cannot effectively extract and utilize key feature information in hyperspectral data, it is easy to cause the inversion results to be inaccurate, especially when dealing with complex terrain and diverse vegetation types. The adaptability and robustness of the inversion model are poor, and it is impossible to stably provide accurate biomass and carbon stock estimation results under different environmental conditions.

[0004] Although traditional generative adversarial networks can generate distribution maps close to real data when used for biomass and carbon stock inversion, due to the complex training process of their generators and discriminators, they are prone to overfitting or mode collapse, resulting in the generated biomass and carbon stock distribution maps lacking details and unable to accurately reflect the actual biomass and carbon stock distribution characteristics.

[0005] Therefore, the existing technologies have obvious deficiencies in hyperspectral data processing, biomass and carbon stock inversion, and the application of generative adversarial networks, and cannot meet the requirements of modern ecological monitoring and carbon stock assessment for efficiency, accuracy, and robustness. The defects of the existing technologies limit the application of UAV hyperspectral technology in large-scale ecosystem monitoring. Summary of the invention

[0006] The present invention aims to solve the problem of related technical limitations at least to a certain extent. To this end, the present invention provides a biomass carbon stock inversion method, device, electronic device and storage medium, which can achieve biomass carbon stock inversion efficiently and accurately.

[0007] On the one hand, an embodiment of the present invention provides a biomass carbon stock inversion method, comprising: Acquire hyperspectral image data of a sample plot area in the target area; the hyperspectral image data includes multi-band reflectance spectral information of surface vegetation in the sample plot area; the sample plot area includes multiple sample plots of different terrain types in the target area; Preprocess the hyperspectral image data; the preprocessing includes radiation correction, geometric correction and noise removal; A sparse basis dictionary is constructed based on the preprocessed hyperspectral image data, and a sparse coding algorithm is used to extract feature vectors; Based on the feature vector, a first distribution map of biomass and carbon storage is generated using a preset first generative adversarial network; Based on the first distribution map, a second distribution map of biomass and carbon storage is generated using a preset second generative adversarial network; Among them, the first generative adversarial network and the second generative adversarial network are constructed based on adversarial training of the generator and the discriminator; An inversion model is constructed based on the mapping relationship between the second distribution map and the characteristic vector; The inversion model is applied to the whole area of ​​the target region to obtain the third distribution map of biomass and carbon stock in the target region.

[0008] Optionally, obtaining hyperspectral image data of a sample area in a target area includes the following steps: In response to the control command of the target object, the target UAV is controlled to fly to the sample area of ​​the target area based on preset operating parameters; the target UAV is provided with a hyperspectral imaging sensor; Obtain multi-band reflectance spectral information collected by the hyperspectral imaging sensor of the target UAV during its flight in the sample area; According to the flight path and time series of the target UAV, all multi-band reflectance spectral information is spliced ​​and registered to obtain the hyperspectral image data of the sample area.

[0009] Optionally, a sparse basis dictionary is constructed based on the preprocessed hyperspectral image data, and a sparse coding algorithm is used to extract a feature vector, including the following steps: The initial sparse basis dictionary is constructed based on the preprocessed hyperspectral image data; the expression of the initial sparse basis dictionary is: ,in, represents the i-th basis vector, k is the number of basis vectors in the initial sparse basis dictionary; The initial sparse basis dictionary is optimized to obtain an optimized sparse basis dictionary; the expression of the optimized sparse basis dictionary is: ; in, Represents an optimized sparse basis dictionary; represents the preprocessed hyperspectral image data, represents the surface coordinates, Indicates wavelength; Represents the sparse coefficient corresponding to the i-th basis vector; is the regularization parameter; The sparse coding algorithm is used to sparsely represent the preprocessed hyperspectral image data, and the sparse representation coefficient vector of the spectral vector corresponding to each hyperspectral pixel is obtained; the expression of the sparse representation coefficient vector is: ; in, represents the sparse representation coefficient vector; represents the initial sparse basis dictionary; Represents the sparse coefficient corresponding to the hyperspectral pixel; and is the regularization parameter; represents the sparse coefficient corresponding to the jth pixel around the hyperspectral pixel, and m is the number of pixels around the hyperspectral pixel; The initial feature vector is obtained by extracting the sparse representation coefficient vector in combination with the optimized sparse basis dictionary; The initial feature vector is sparsely encoded and dimensionally reduced to obtain the feature vector.

[0010] Optionally, the method further comprises the following steps: Obtain actual distribution maps of biomass and carbon storage in the sample area; A first generator is constructed based on a first mapping function between the feature vector and the actual distribution graph; A first discriminant is constructed based on a first discriminant function that distinguishes a generated result of the first generator from an actual distribution graph; Generate a first training distribution graph through a first generator according to the feature vector; Generate a first discrimination result through a first discriminator according to the first training distribution graph and the actual distribution graph; Based on the first discrimination result, performing first adversarial training on the first generator and the first discriminator according to a preset first loss function; Optimize the parameters of the first generator and the first discriminator according to the result of the first adversarial training; return to execute the step of generating a first training distribution map through the first generator according to the feature vector until the first number of iterations is reached to obtain a first generative adversarial network.

[0011] Optionally, the method further comprises the following steps: Obtain actual distribution maps of biomass and carbon storage in the sample area; A second generator is constructed based on a second mapping function of the first distribution map and the actual distribution map; A second discriminant is constructed based on a second discriminant function that distinguishes the generated result of the second generator from the actual distribution graph; According to the first distribution map, a second training distribution map is generated by a second generator; Generate a second discrimination result through a second discriminator according to the second training distribution map and the actual distribution map; Based on the second discrimination result, performing second adversarial training on the second generator and the second discriminator according to a preset second loss function; Optimize the parameters of the second generator and the second discriminator according to the results of the second adversarial training; return to the step of generating a second training distribution map through the second generator according to the first distribution map until the second number of iterations is reached to obtain a second generative adversarial network.

[0012] Optionally, constructing an inversion model based on a mapping relationship between the second distribution map and the characteristic vector comprises the following steps: Obtain actual distribution maps of biomass and carbon storage in the sample area; According to the mapping function between the eigenvector and the actual distribution map, the least square method is used to optimize the error of the mapping relationship between the second distribution map and the eigenvector, and the inversion model is obtained by fitting.

[0013] Optionally, the method further comprises the following steps: Obtain a true distribution map of biomass and carbon storage in the target area; Based on the third distribution map, a classification analysis of the terrain and vegetation types of the target area is performed to obtain a terrain feature map and a vegetation type map of the target area; According to the terrain feature map and vegetation type map, statistical values ​​under different terrain conditions and vegetation types are obtained; the statistical values ​​include biomass statistical values ​​and carbon storage statistical values; Based on the terrain characteristic map and vegetation type map, combined with the actual distribution map, the statistical values ​​under different terrain conditions and vegetation types in the target area are evaluated and analyzed to obtain an analysis report on the biomass and carbon storage of the target area.

[0014] On the other hand, an embodiment of the present invention provides a biomass carbon stock inversion device, comprising: The first module is used to obtain hyperspectral image data of a sample plot area in the target area; the hyperspectral image data includes multi-band reflectance spectral information of surface vegetation in the sample plot area; the sample plot area includes multiple sample plots of different terrain types in the target area; The second module is used to preprocess the hyperspectral image data; the preprocessing includes radiation correction, geometric correction and noise removal; The third module is used to construct a sparse basis dictionary based on the preprocessed hyperspectral image data and extract the feature vector using a sparse coding algorithm; A fourth module is used to generate a first distribution map of biomass and carbon storage based on the feature vector using a preset first generative adversarial network; A fifth module is used to generate a second distribution map of biomass and carbon storage based on the first distribution map by using a preset second generative adversarial network; Among them, the first generative adversarial network and the second generative adversarial network are constructed based on adversarial training of the generator and the discriminator; A sixth module is used to construct an inversion model based on the mapping relationship between the second distribution map and the characteristic vector; The seventh module is used to apply the inversion model to the entire area of ​​the target area to obtain a third distribution map of biomass and carbon storage in the target area.

[0015] Optionally, the device further comprises: The eighth module is used to obtain the actual distribution map of biomass and carbon storage in the sample area; A ninth module is used to construct a first generator based on a first mapping function between the feature vector and the actual distribution graph; A tenth module is used to construct a first discriminator based on a first discriminant function that distinguishes a generated result of the first generator from an actual distribution graph; An eleventh module is used to generate a first training distribution graph through a first generator according to the feature vector; A twelfth module is used to generate a first discrimination result through a first discriminator according to the first training distribution map and the actual distribution map; A thirteenth module is used to perform a first adversarial training on the first generator and the first discriminator according to a preset first loss function based on the first discrimination result; The fourteenth module is used to optimize the parameters of the first generator and the first discriminator according to the results of the first adversarial training; return to execute the eleventh module until the first iteration number is reached to obtain the first generative adversarial network.

[0016] Optionally, the device further comprises: Module 15 is used to obtain the actual distribution map of biomass and carbon storage in the sample area; A sixteenth module is used to construct a second generator based on a second mapping function between the first distribution map and the actual distribution map; A seventeenth module is used to construct a second discriminator based on a second discriminant function that distinguishes the generated result of the second generator from the actual distribution graph; An eighteenth module is used to generate a second training distribution map through a second generator according to the first distribution map; A nineteenth module is used to generate a second discrimination result through a second discriminator according to the second training distribution map and the actual distribution map; A twentieth module, configured to perform second adversarial training on the second generator and the second discriminator according to a preset second loss function based on the second discrimination result; The twenty-first module is used to optimize the parameters of the second generator and the second discriminator according to the results of the second adversarial training; return to execute the eighteenth module until the second iteration number is reached to obtain the second generative adversarial network.

[0017] On the other hand, an embodiment of the present invention provides an electronic device, including: a processor and a memory; the memory is used to store programs; the processor executes the program to implement the above-mentioned biomass carbon stock inversion method.

[0018] On the other hand, an embodiment of the present invention provides a computer storage medium, in which a program executable by a processor is stored. When the program executable by the processor is executed by the processor, it is used to implement the above-mentioned biomass carbon stock inversion method.

[0019] The embodiment of the present invention obtains hyperspectral image data of a sample plot area in a target area; the hyperspectral image data includes multi-band reflectance spectral information of surface vegetation in the sample plot area; the sample plot area includes multiple sample plots of different terrain types in the target area; the hyperspectral image data is preprocessed; the preprocessing includes radiation correction, geometric correction and noise removal; a sparse basis dictionary is constructed based on the preprocessed hyperspectral image data, and a sparse coding algorithm is used to extract a feature vector; based on the feature vector, a first distribution map of biomass and carbon storage is generated using a preset first generative adversarial network; based on the first distribution map, a second distribution map of biomass and carbon storage is generated using a preset second generative adversarial network; wherein the first generative adversarial network and the second generative adversarial network are constructed based on adversarial training of a generator and a discriminator; an inversion model is constructed based on the mapping relationship between the second distribution map and the feature vector; the inversion model is applied to the entire area of ​​the target area to obtain a third distribution map of biomass and carbon storage in the target area. The present invention can effectively improve the accuracy and details of the generated image by applying a two-level generative adversarial network, and can further solve the problem that a single generative adversarial network is difficult to generate detailed data. In addition, the present invention uses sparse representation technology to effectively reduce the redundancy of hyperspectral data and extract key features, thereby reducing computational costs, and can provide stable and accurate inversion results even under complex environmental conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings are used to provide a further understanding of the technical solution of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the technical solution of the present invention and do not constitute a limitation on the technical solution of the present invention.

[0021] Figure 1 A schematic diagram of an implementation environment for biomass carbon stock inversion provided by an embodiment of the present invention; Figure 2 A schematic diagram of a process for inverting biomass carbon stock provided by an embodiment of the present invention; Figure 3 A schematic diagram of the expansion process of step S100 provided in an embodiment of the present invention; Figure 4 A schematic diagram of the expanded flow of step S300 provided in an embodiment of the present invention; Figure 5 A schematic diagram of the architecture principle of a primary generative adversarial network provided in an embodiment of the present invention; Figure 6 A schematic diagram of the architecture principle of an advanced generative adversarial network provided by an embodiment of the present invention; Figure 7 A schematic diagram of the expanded flow of step S600 provided in an embodiment of the present invention; Figure 8 A schematic diagram of the overall process principle of the biomass carbon stock inversion method provided by an embodiment of the present invention; Fig. 9 A schematic diagram of the structure of a biomass carbon stock inversion device provided by an embodiment of the present invention; Fig.10 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0022] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0023] It should be noted that, although the functional modules are divided in the system schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the system or the order in the flowchart. The terms "first / S100", "second / S200", etc. in the specification, claims and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0024] Reference to "embodiments" in the present invention means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described in the present invention may be combined with other embodiments.

[0025] It is understandable that the biomass carbon stock inversion method provided in the embodiment of the present invention can be applied to any computer device with data processing and computing capabilities, and this computer device can be various terminals or servers. When the computer device in the embodiment is a server, the server is an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms. Optionally, the terminal is a smart phone, a tablet computer, a laptop computer, a desktop computer, etc., but is not limited to this.

[0026] To facilitate the understanding of the technical solution of the present invention, the specific terms of the technical features that may appear in the embodiments of the present invention are first explained: Sparse representation: Sparse representation is a technical method for signal processing and data analysis, which simplifies data by representing signals or data as a combination of a set of sparse, linearly independent basis vectors.

[0027] In short, sparse representation describes complex data sets by using only a few important features (rather than all information), thereby achieving dimensionality reduction and efficient calculation, and is widely used in image processing, pattern recognition and other fields. However, it is less used in biomass carbon stock inversion.

[0028] Hyperspectral region: Hyperspectral region refers to multi-band spectral data obtained using hyperspectral imaging technology, where each band records the reflection or absorption information of an object at different wavelengths. Unlike traditional color images (red, green, and blue bands), hyperspectral imaging can capture data from dozens or even hundreds of bands, so it can provide more detailed spectral characteristics of materials. Hyperspectral data can accurately identify and distinguish different material components, such as vegetation, soil, or water bodies, and is a key technology for biomass and carbon storage analysis.

[0029] like Figure 1 FIG. 1 is a schematic diagram of an implementation environment provided by an embodiment of the invention. Figure 1 The implementation environment includes at least one terminal 102 and a server 101. The terminal 102 and the server 101 can be connected to a network wirelessly or wired to complete data transmission and exchange.

[0030] Server 101 can be an independent physical server, or a server cluster or distributed system composed of multiple physical servers. It can also be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), as well as big data and artificial intelligence platforms.

[0031] In addition, the server 101 can also be a node server in the blockchain network. Blockchain is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, and encryption algorithm.

[0032] The terminal 102 may be a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited thereto. The terminal 102 and the server 101 may be directly or indirectly connected via wired or wireless communication, which is not limited in the embodiment of the present invention.

[0033] Based on the example Figure 1 In the implementation environment shown, an embodiment of the present invention provides a biomass carbon stock inversion method. The following is explained using the example of the biomass carbon stock inversion method being applied in the server 101. It can be understood that the biomass carbon stock inversion method can also be applied in the terminal 102.

[0034] Reference Figure 2 , Figure 2 The flowchart of the biomass carbon stock inversion method applied to the server provided in the embodiment of the present invention, the execution subject of the biomass carbon stock inversion method can be any of the aforementioned computer devices (including servers or terminals). Figure 2 , the method comprises the following steps: S100, obtaining hyperspectral image data of a sample area in a target area; The hyperspectral image data includes multi-band reflectance spectral information of surface vegetation in the sample plot area; the sample plot area includes multiple sample plots of different terrain types in the target area; It should be noted that in some embodiments, Figure 3As shown, step S100 may include the following steps: S101, in response to the control instruction of the target object, controlling the target UAV to fly to the sample area of ​​the target area based on preset operating parameters; a hyperspectral imaging sensor is arranged on the target UAV; S102, obtaining multi-band reflection spectral information collected by the hyperspectral imaging sensor during the flight of the target UAV in the sample area; S103, according to the flight path and time series of the target UAV, all multi-band reflection spectral information is spliced ​​and registered to obtain hyperspectral image data of the sample area.

[0035] Exemplarily, in some specific implementations, the embodiments of the present invention first utilize a drone equipped with a hyperspectral imaging sensor to fly over a target area to obtain hyperspectral image data covering the target area, wherein the hyperspectral image data contains multi-band reflectance spectral information of surface vegetation; wherein multiple sample plots may be set up on different terrain types in the target area, for example, 50 sample plots, each with an area of ​​approximately 1 hectare, serving as collection points for training samples and verification samples; the drone is equipped with a hyperspectral imaging sensor, and according to a preset flight path, it performs multiple flight missions under different altitudes and terrain conditions, and collects hyperspectral image data covering the entire study area.

[0036] Optionally, in some preferred implementations, step S100 may be implemented as follows: S11, selecting a UAV flight path and altitude suitable for the target area, and determining the UAV flight speed v and flight altitude h, wherein the flight speed v should be adjusted according to the frame rate f and ground resolution r of the hyperspectral imaging sensor; S12. Calculate the width W and length L of the ground surface that can be covered in a single flight based on the flight altitude h of the drone and the field of view FOV of the sensor: ; ; in, and are the field of view angles of the hyperspectral imaging sensor in the horizontal and vertical directions, respectively; S13. During the flight of the UAV, the multi-band reflectance spectrum data (multi-band reflectance spectrum information) collected by the hyperspectral imaging sensor is obtained in real time, and the spectral reflectance of each image pixel is recorded. ,in Indicated in wavelength The reflectivity at the wavelength range covers multiple bands from visible light to near infrared , so that the hyperspectral data reflects the spectral characteristics of surface vegetation; S14. According to the flight path and time sequence of the UAV, the continuously acquired hyperspectral image data are spliced ​​and registered to generate complete hyperspectral image data covering the target area. , where x and y are the surface coordinates, is the wavelength; S15. In the acquired hyperspectral image data, the area and boundary of the surface vegetation are calibrated, and the effective area data containing multi-band reflectance spectrum information is extracted. ,in, Represents spectral data with biomass and carbon stock inversion significance in the target area.

[0037] S200, preprocessing the hyperspectral image data; Among them, preprocessing includes radiation correction, geometric correction and noise removal; it can be achieved by using corresponding processing algorithms or data processing software. The specific processing method is not the focus of the process logic of the present invention and will not be repeated here.

[0038] S300, constructing a sparse basis dictionary based on the preprocessed hyperspectral image data, and extracting a feature vector using a sparse coding algorithm; It should be noted that in some embodiments, Figure 4 As shown, step S300 may include the following steps: S301, constructing an initial sparse basis dictionary based on the preprocessed hyperspectral image data; the expression of the initial sparse basis dictionary is: ,in, represents the i-th basis vector, k is the number of basis vectors in the initial sparse basis dictionary; S302, optimizing the initial sparse basis dictionary to obtain an optimized sparse basis dictionary; the expression of the optimized sparse basis dictionary is: ; in, Represents an optimized sparse basis dictionary; represents the preprocessed hyperspectral image data, represents the surface coordinates, Indicates wavelength; Represents the sparse coefficient corresponding to the i-th basis vector; is the regularization parameter; S303, using a sparse coding algorithm to perform sparse representation on the preprocessed hyperspectral image data, and obtaining a sparse representation coefficient vector of the spectral vector corresponding to each hyperspectral pixel; the expression of the sparse representation coefficient vector is: ; in, represents the sparse representation coefficient vector; represents the initial sparse basis dictionary; Represents the sparse coefficient corresponding to the hyperspectral pixel; and is the regularization parameter; represents the sparse coefficient corresponding to the jth pixel around the hyperspectral pixel, and m is the number of pixels around the hyperspectral pixel; S304, extracting an initial feature vector based on the sparse representation coefficient vector and the optimized sparse basis dictionary; S305 , performing sparse coding and dimensionality reduction processing on the initial feature vector to obtain a feature vector.

[0039] For example, in some specific implementations, a sparse basis dictionary may be constructed and a sparse coding algorithm may be used to extract the most representative feature vectors in the hyperspectral image data, perform dimensionality reduction processing, and retain key spectral information in the hyperspectral image data.

[0040] Optionally, in some preferred implementations, step S300 may be implemented as follows: S31, based on the hyperspectral image data of the target area , construct a sparse basis dictionary ,in, represents the i-th basis vector, k is the number of basis vectors in the dictionary, The build is optimized based on: ; in, is the optimized sparse basis dictionary, is the corresponding sparse coefficient, is a regularization parameter used to control the norm of the dictionary vector to ensure that the constructed dictionary can effectively represent the key spectral information in the hyperspectral data and reduce noise interference; S32, using sparse coding algorithm For each hyperspectral pixel spectral vector x, we can obtain each spectral vector The sparse representation coefficient vector : ; in, and is the regularization parameter, is the sparse coefficient of the surrounding pixels, which is used to consider spatial consistency, so that the generated sparse coefficient not only reflects the local spectral characteristics, but also takes into account the spectral consistency between regions, which is beneficial to the accuracy of the hyperspectral data collected by UAVs in the inversion of carbon storage in local areas; S33, according to the sparse representation coefficient vector Extracting Hyperspectral Image Data The most representative eigenvector in : ; in, is the smoothing parameter, x represents the spatial coordinates of the hyperspectral image data, that is, the position of the pixel in the image, that is, the horizontal coordinate, and y is the vertical coordinate of this point. Furthermore, x' means the position of the neighboring pixels, that is, the adjacent coordinates used for smoothing, and y' is the same. D* means the optimized sparse basis dictionary, is a weight function used to measure the similarity between adjacent pixels to ensure that when extracting feature vectors, not only the local key spectral information is retained, but also the data in different regions can be smoothly transitioned, thus improving the representativeness and stability of the feature vectors in the inversion process; S34, yes Sparse coding and dimensionality reduction are performed on all hyperspectral pixels in the target area to generate sparse representation hyperspectral image data of the target area. ,in is the dimension-reduced feature vector of the j-th pixel, and m is the number of pixels in the hyperspectral image.

[0041] S400, generating a first distribution map of biomass and carbon storage based on the feature vector using a preset first generative adversarial network; In some optional embodiments, the method may further include the following steps: obtaining an actual distribution map of biomass and carbon storage in the sample plot area; constructing a first generator based on a first mapping function between a feature vector and the actual distribution map; constructing a first discriminant based on a first discriminant function that distinguishes a generation result of the first generator from the actual distribution map; generating a first training distribution map through the first generator according to the feature vector; generating a first discrimination result through the first discriminator according to the first training distribution map and the actual distribution map; based on the first discrimination result, performing a first adversarial training on the first generator and the first discriminator according to a preset first loss function; optimizing parameters of the first generator and the first discriminator according to the result of the first adversarial training; returning to execute the step of generating a first training distribution map through the first generator according to the feature vector until the first number of iterations is reached to obtain a first generative adversarial network.

[0042] Exemplarily, in some specific embodiments, a primary generative adversarial network (i.e., a first generative adversarial network) may be constructed, wherein a primary generator of the primary generative adversarial network is used to generate a biomass and carbon stock distribution map corresponding to the hyperspectral image data after dimensionality reduction, and a primary discriminator is used to distinguish the generated biomass and carbon stock distribution map from the real biomass and carbon stock data, and the training of the primary generative adversarial network is performed through an adversarial process between the primary generator and the primary discriminator; Optionally, in some preferred embodiments, as Figure 5 As shown, constructing a primary generative adversarial network can be implemented as follows: S41. Based on the dimension-reduced hyperspectral image data Y, a primary generative adversarial network is constructed. The primary generative adversarial network includes a primary generator and the primary discriminator ; S42, Primary Generator Receive the reduced feature vector As input, generate corresponding biomass and carbon stock distribution maps , the goal of the primary generator is to learn the mapping function : , so that the generated biomass and carbon stock distribution maps are close to the real data; S43, primary discriminator The input is the generated biomass and carbon stock distribution map and real biomass and carbon storage data The goal of the primary discriminator output is to distinguish the generated biomass and carbon storage distribution maps from the real data, that is, to learn a discriminant function : , where 1 represents real data and 0 represents generated data; S44, through the primary generator and the primary discriminator The adversarial training process between them optimizes the parameters of the primary generator and the discriminator. The loss function of the primary generator is defined as: ; in, and Represents the real biomass and carbon storage data respectively And the feature vector after dimensionality reduction The expectation on the primary discriminator The goal is to maximize the accuracy of discriminating real data, and the primary generator The goal is to minimize the ability of the primary discriminator to distinguish the generated data; S45. Optimize the primary generator of the primary generative adversarial network through multiple rounds of iterative training and the primary discriminator The parameters of the generated biomass and carbon storage distribution map Visually and statistically close to real data .

[0043] S500, based on the first distribution map, using a preset second generative adversarial network to generate a second distribution map of biomass and carbon storage; Among them, the first generative adversarial network and the second generative adversarial network are constructed based on adversarial training of the generator and the discriminator; In some optional embodiments, the method may further include the following steps: obtaining an actual distribution map of biomass and carbon storage in the sample plot area; constructing a second generator based on a second mapping function between the first distribution map and the actual distribution map; constructing a second discriminant based on a second discriminant function that distinguishes the generation result of the second generator from the actual distribution map; generating a second training distribution map through the second generator according to the first distribution map; generating a second discrimination result through the second discriminator according to the second training distribution map and the actual distribution map; based on the second discrimination result, performing a second adversarial training on the second generator and the second discriminator according to a preset second loss function; optimizing the parameters of the second generator and the second discriminator according to the result of the second adversarial training; returning to execute the step of generating a second training distribution map through the second generator according to the first distribution map until the second number of iterations is reached to obtain a second generative adversarial network.

[0044] Exemplarily, in some specific implementations, the embodiments of the present invention further construct an advanced generative adversarial network (i.e., a second generative adversarial network), and the advanced generator of the advanced generative adversarial network further optimizes the generated results based on the biomass and carbon storage distribution maps generated by the primary generative adversarial network, so that the biomass and carbon storage distribution maps are more detailed and accurate, and the advanced discriminator distinguishes the results generated by the primary generative adversarial network from the real data. The training of the advanced generative adversarial network is carried out through an adversarial process between the advanced generator and the advanced discriminator, so that the biomass and carbon storage distribution maps generated by the advanced generator are more consistent with the distribution characteristics of the real data; Optionally, in some preferred embodiments, as Figure 6 As shown, building an advanced generative adversarial network can be achieved as follows: S51. Biomass and carbon storage distribution map generated based on primary generative adversarial network , build advanced generative adversarial networks, advanced generative adversarial networks include advanced generators and high-level discriminators ; S52, Advanced Generator Receive primary generation of biomass and carbon stock distribution map As input, generate optimized biomass and carbon stock distribution maps , the goal of the advanced generator is to learn more complex mapping functions , optimize the biomass and carbon stock distribution maps at the detailed level, making them more refined and accurate; S53, Advanced Discriminator The input is the biomass and carbon stock distribution map generated by the advanced generator and real biomass and carbon storage data , the discriminator outputs the goal of distinguishing the biomass and carbon storage distribution maps generated by advanced processes from the real data, that is, learning the discriminant function , where 1 represents real data and 0 represents generated data; S54, through the advanced generator and high-level discriminators The adversarial training process between them optimizes the parameters of the high-level generator and the high-level discriminator. The loss function of the high-level generator is defined as: ; in, and Represents the real biomass and carbon storage data respectively And the distribution map of primary biomass and carbon storage Expectation on high-level discriminator The goal is to maximize the discrimination of real data, the advanced generator The goal is to minimize the ability of the high-level discriminator to distinguish high-level generated data; S55. Optimize the advanced generator of the advanced generative adversarial network through multiple rounds of iterative training and high-level discriminators The parameters of the advanced generator make the biomass and carbon stock distribution map generated by It is more consistent with the distribution characteristics of real data at the detailed level.

[0045] S600, constructing an inversion model based on the mapping relationship between the second distribution map and the characteristic vector; It should be noted that in some embodiments, Figure 7 As shown, step S600 may include the following steps: S601, obtaining the actual distribution map of biomass and carbon storage in the sample area; S602, according to the mapping function between the characteristic vector and the actual distribution map, using the least squares method to optimize the error of the mapping relationship between the second distribution map and the characteristic vector, and fitting to obtain the inversion model.

[0046] Exemplarily, in some specific embodiments, an inversion model can be constructed based on the biomass and carbon storage distribution map (i.e., the second distribution map) generated by the trained advanced generative adversarial network, and the inversion model estimates the biomass and carbon storage of the target area by using the mapping relationship between the generated carbon storage biomass and carbon storage distribution map and the hyperspectral feature vector of the target area; Optionally, in some preferred implementations, constructing the inversion model may be specifically implemented as follows: S61. Biomass and carbon storage distribution map generated by advanced generative adversarial networks , build the inversion model , the inversion model is used to generate the distribution map The hyperspectral feature vector of the target area Establish a mapping relationship between them, and then estimate the biomass and carbon storage in the target area; S62, Inversion Model The construction process includes defining the mapping function ,in, Represent the real biomass and carbon storage data of the target area by fitting the training samples; S63. Biomass and carbon storage distribution map generated by least squares minimization Real data from the target area The error between them, the loss function is defined as: ; Where N is the number of training samples, represents the Euclidean norm, represents the sparsity regularization term of the model parameters, is a regularization parameter used to control the complexity of the model and prevent overfitting.

[0047] S700: Apply the inversion model to the entire target area to obtain a third distribution map of biomass and carbon storage in the target area.

[0048] For example, in some specific implementations, the inversion model can be applied to the hyperspectral image data of the target area to estimate the biomass and carbon stock distribution in the target area.

[0049] Optionally, in some preferred embodiments, estimating the biomass and carbon stock distribution in the target area can be achieved as follows: S71, the optimized inversion model Hyperspectral image data applied to target area , according to the hyperspectral feature vector of each pixel , calculate biomass and carbon stock estimates for pixels : ; in, is a weighting function based on pixel position and wavelength, which is used to adjust the contribution of different spectral bands to biomass and carbon stock estimation. is the hyperspectral feature vector and sparse basis dictionary The similarity kernel function is used to capture the nonlinear relationship between local spectral features and carbon storage. is the sparse coding coefficient, which indicates the contribution of different basis vectors to a specific pixel. is the dictionary basis vector, and are the minimum and maximum wavelength ranges of the hyperspectral data, respectively; S72, traverse all pixels in the target area, and use the inversion model Generate complete biomass and carbon stock distribution maps : ; in, is the weight matrix between pixel position and pixel number, which is used to integrate the estimation results of multiple pixels. is a smoothing function of adjacent pixels, used for spatial smoothing of biomass and carbon storage. is the standard deviation of the Gaussian kernel function, which controls the degree of smoothness, and Respectively represent the coordinate range of the target area in the x and y directions.

[0050] In some optional embodiments, the method may further include the following steps: obtaining a true distribution map of biomass and carbon storage in the target area; based on the third distribution map, performing classification analysis on the terrain and vegetation types of the target area to obtain a terrain characteristic map and a vegetation type map of the target area; according to the terrain characteristic map and the vegetation type map, obtaining statistical values ​​under different terrain conditions and vegetation types; the statistical values ​​include biomass statistical values ​​and carbon storage statistical values; according to the terrain characteristic map and the vegetation type map, combined with the true distribution map, evaluating and analyzing the statistical values ​​under different terrain conditions and vegetation types in the target area to obtain an analysis report on the biomass and carbon storage of the target area.

[0051] For example, in some specific implementations, an analysis report can be provided based on the biomass and carbon storage distribution map generated by the inversion model, including the distribution characteristics under different terrains and vegetation types and the evaluation of the inversion results. In some preferred implementations, the following can be achieved: S91, based on the generated biomass and carbon stock distribution map , classify and analyze the terrain and vegetation types of the target area, and generate a terrain feature map of the target area and vegetation type map ,in, Indicates the distribution characteristics of biomass and carbon storage under different terrain conditions. Indicates the distribution characteristics of biomass and carbon storage under different vegetation types; S92, according to the terrain feature map and vegetation type map , calculate biomass and carbon storage statistics under different terrain conditions and vegetation types and ,in, represents the sum of biomass and carbon storage under specific terrain conditions, Represents the sum of biomass and carbon storage under a specific vegetation type; S93, combined with terrain feature map and vegetation type map , evaluate the carbon storage inversion results of different terrain and vegetation types in the target area, and define the evaluation function : ; in, is the actual biomass and carbon storage data in the target area, and are weight functions of terrain and vegetation types, respectively, which are used to adjust the impact of different terrain and vegetation types on the evaluation results. Represent the relative error distribution of the inversion results and evaluate the accuracy of the inversion results under different conditions; S94, based on evaluation function , generate a biomass and carbon stock analysis report for the target area, which includes the distribution characteristics of biomass and carbon stocks under different terrains and vegetation types, evaluation of inversion results, relative error analysis, and recommendations for ecological monitoring and carbon stock management.

[0052] In order to explain the principle of the technical solution of the present invention in detail, the overall process of the present invention is described below in combination with some specific embodiments. It is easy to understand that the following is an explanation of the technical principle of the present invention and cannot be regarded as a limitation of the present invention.

[0053] First of all, it should be noted that although the existing technical solutions can provide data support for large-scale ecosystems by monitoring biomass and carbon storage through drones equipped with hyperspectral sensors, they face multiple challenges when processing hyperspectral data. First, the high dimensionality and redundancy of the data increase the processing complexity, resulting in low computational efficiency and difficulty in achieving real-time performance. Secondly, traditional data processing methods cannot effectively extract key feature information, especially in environments with complex terrain and diverse vegetation types. The inversion model has poor adaptability and cannot provide stable and accurate results. In addition, although the application of generative adversarial networks can generate distribution maps that are close to reality to a certain extent, its training process is complicated and prone to overfitting, resulting in insufficient details in the results, which limits its application in carbon storage estimation.

[0054] The purpose of the present invention is to address these deficiencies in the prior art and propose a UAV hyperspectral biomass carbon stock inversion method based on sparse representation to improve data processing efficiency and enhance the robustness and accuracy of the model. Through sparse representation technology, the redundancy of hyperspectral data is effectively reduced, key features are extracted, thereby reducing computing costs, and stable and accurate inversion results can be provided under complex environmental conditions. In this way, not only can the practicality of UAV hyperspectral technology in ecosystem monitoring be improved, but also the requirements of ecological monitoring for efficiency, accuracy and robustness can be met, and the popularization and application of biomass and carbon stock monitoring can be promoted.

[0055] like Figure 8 As shown, the present invention provides a method for inverting biomass carbon storage in a UAV hyperspectral region based on sparse representation, which may include the following steps: S1. Using a drone equipped with a hyperspectral imaging sensor to fly over a target area, obtaining hyperspectral image data covering the target area, wherein the hyperspectral image data includes multi-band reflectance spectral information of surface vegetation; S2, preprocessing the hyperspectral image data, including radiation correction, geometric correction and noise removal; S3, by constructing a sparse basis dictionary, using a sparse coding algorithm to extract the most representative feature vectors in the hyperspectral image data, and performing dimensionality reduction processing, while retaining the key spectral information in the hyperspectral image data; S4. constructing a primary generative adversarial network, wherein the primary generator of the primary generative adversarial network is used to generate a biomass and carbon storage distribution map corresponding to the hyperspectral image data after dimensionality reduction, and the primary discriminator is used to distinguish the generated biomass and carbon storage distribution map from the real biomass and carbon storage data. The training of the primary generative adversarial network is carried out through an adversarial process between the primary generator and the primary discriminator; S5. Construct an advanced generative adversarial network. The advanced generator of the advanced generative adversarial network further optimizes the generated results based on the biomass and carbon storage distribution maps generated by the primary generative adversarial network, so that the biomass and carbon storage distribution maps are more detailed and accurate. The advanced discriminator distinguishes the results generated by the primary generative adversarial network from the real data. The training of the advanced generative adversarial network is carried out through the adversarial process between the advanced generator and the advanced discriminator, so that the biomass and carbon storage distribution maps generated by the advanced generator are more consistent with the distribution characteristics of the real data. S6. Optimize the generator and discriminator of the primary generative adversarial network and the advanced generative adversarial network through joint training. The primary generative adversarial network and the advanced generative adversarial network work together to optimize the generated biomass and carbon storage distribution map through multiple rounds of iterative training. S7. Based on the biomass and carbon storage distribution maps generated by the trained advanced generative adversarial network, an inversion model is constructed. The inversion model estimates the biomass and carbon storage of the target area by using the mapping relationship between the generated carbon storage biomass and carbon storage distribution maps and the hyperspectral feature vectors of the target area; S8, applying the inversion model to the hyperspectral image data of the target area to estimate the biomass and carbon stock distribution in the target area; S9. Provide analysis reports based on the biomass and carbon stock distribution maps generated by the inversion model, including distribution characteristics under different terrain and vegetation types and evaluation of inversion results.

[0056] In this implementation, S1 includes the following sub-steps: S11, selecting a UAV flight path and altitude suitable for the target area, and determining the UAV flight speed v and flight altitude h, wherein the flight speed v should be adjusted according to the frame rate f and ground resolution r of the hyperspectral imaging sensor; S12. Calculate the width W and length L of the ground surface that can be covered in a single flight based on the flight altitude h of the drone and the field of view FOV of the sensor: ; ; in, and are the field of view angles of the hyperspectral imaging sensor in the horizontal and vertical directions, respectively; S13. During the flight of the UAV, the multi-band reflectance spectrum data (multi-band reflectance spectrum information) collected by the hyperspectral imaging sensor is obtained in real time, and the spectral reflectance of each image pixel is recorded. ,in Indicated in wavelength The reflectivity at the wavelength range covers multiple bands from visible light to near infrared , so that the hyperspectral data reflects the spectral characteristics of surface vegetation; S14. According to the flight path and time sequence of the UAV, the continuously acquired hyperspectral image data are spliced ​​and registered to generate complete hyperspectral image data covering the target area. , where x and y are the surface coordinates, is the wavelength; S15. In the acquired hyperspectral image data, the area and boundary of the surface vegetation are calibrated, and the effective area data containing multi-band reflectance spectrum information is extracted. ,in, Represents spectral data with biomass and carbon stock inversion significance in the target area.

[0057] In this implementation, S3 includes the following sub-steps: S31, based on the hyperspectral image data of the target area , construct a sparse basis dictionary ,in, represents the i-th basis vector, k is the number of basis vectors in the dictionary, The build is optimized based on: ; in, is the optimized sparse basis dictionary, is the corresponding sparse coefficient, is a regularization parameter used to control the norm of the dictionary vector to ensure that the constructed dictionary can effectively represent the key spectral information in the hyperspectral data and reduce noise interference; S32, using sparse coding algorithm For each hyperspectral pixel spectral vector x, we can obtain each spectral vector The sparse representation coefficient vector : ; in, and is the regularization parameter, is the sparse coefficient of the surrounding pixels, which is used to consider spatial consistency, so that the generated sparse coefficient not only reflects the local spectral characteristics, but also takes into account the spectral consistency between regions, which is beneficial to the accuracy of the hyperspectral data collected by UAVs in the inversion of carbon storage in local areas; S33, according to the sparse representation coefficient vector Extracting Hyperspectral Image Data The most representative eigenvector in : ; in, is the smoothing parameter, x represents the spatial coordinates of the hyperspectral image data, that is, the position of the pixel in the image, that is, the horizontal coordinate, and y is the vertical coordinate of this point. Furthermore, x' means the position of the neighboring pixels, that is, the adjacent coordinates used for smoothing, and then Similarly, D* means the optimized sparse basis dictionary. is a weight function used to measure the similarity between adjacent pixels to ensure that when extracting feature vectors, not only the local key spectral information is retained, but also the data in different regions can be smoothly transitioned, thus improving the representativeness and stability of the feature vectors in the inversion process; S34, yes Sparse coding and dimensionality reduction are performed on all hyperspectral pixels in the target area to generate sparse representation hyperspectral image data of the target area. ,in is the dimension-reduced feature vector of the j-th pixel, and m is the number of pixels in the hyperspectral image.

[0058] In this embodiment, if Figure 5 As shown, S4 includes the following sub-steps: S41. Based on the dimension-reduced hyperspectral image data Y, a primary generative adversarial network is constructed. The primary generative adversarial network includes a primary generator and the primary discriminator ; S42, Primary Generator Receive the reduced feature vector As input, generate corresponding biomass and carbon stock distribution maps , the goal of the primary generator is to learn the mapping function : , so that the generated biomass and carbon stock distribution maps are close to the real data; S43, primary discriminator The input is the generated biomass and carbon stock distribution map and real biomass and carbon storage data The goal of the primary discriminator output is to distinguish the generated biomass and carbon storage distribution maps from the real data, that is, to learn a discriminant function : , where 1 represents real data and 0 represents generated data; S44, through the primary generator and the primary discriminator The adversarial training process between them optimizes the parameters of the primary generator and the discriminator. The loss function of the primary generator is defined as: ; in, and Represents the real biomass and carbon storage data respectively And the feature vector after dimensionality reduction The expectation on the primary discriminator The goal is to maximize the accuracy of discriminating real data, and the primary generator The goal is to minimize the ability of the primary discriminator to distinguish the generated data; S45. Optimize the primary generator of the primary generative adversarial network through multiple rounds of iterative training and the primary discriminator The parameters of the generated biomass and carbon storage distribution map Visually and statistically close to real data .

[0059] In this embodiment, if Figure 6 As shown, S5 includes the following sub-steps: S51. Biomass and carbon storage distribution map generated based on primary generative adversarial network , build advanced generative adversarial networks, advanced generative adversarial networks include advanced generators and high-level discriminators ; S52, Advanced Generator Receive primary generation of biomass and carbon stock distribution map As input, generate optimized biomass and carbon stock distribution maps , the goal of the advanced generator is to learn more complex mapping functions , optimize the biomass and carbon stock distribution maps at the detailed level, making them more refined and accurate; S53, Advanced Discriminator The input is the biomass and carbon stock distribution map generated by the advanced generator and real biomass and carbon storage data , the discriminator outputs the goal of distinguishing the biomass and carbon storage distribution maps generated by advanced processes from the real data, that is, learning the discriminant function , where 1 represents real data and 0 represents generated data; S54, through the advanced generator and high-level discriminators The adversarial training process between them optimizes the parameters of the high-level generator and the high-level discriminator. The loss function of the high-level generator is defined as: ; in, and Represents the real biomass and carbon storage data respectively And the distribution map of primary biomass and carbon storage Expectation on high-level discriminator The goal is to maximize the discrimination of real data, the advanced generator The goal is to minimize the ability of the high-level discriminator to distinguish high-level generated data; S55. Optimize the advanced generator of the advanced generative adversarial network through multiple rounds of iterative training and high-level discriminators The parameters of the advanced generator make the biomass and carbon stock distribution map generated by It is more consistent with the distribution characteristics of real data at the detailed level.

[0060] In this implementation, S7 includes the following sub-steps: S71. Biomass and carbon storage distribution map generated by advanced generative adversarial networks , build the inversion model , the inversion model is used to generate the distribution map The hyperspectral feature vector of the target area Establish a mapping relationship between them, and then estimate the biomass and carbon storage in the target area; S72, Inversion Model The construction process includes defining the mapping function ,in, Represent the real biomass and carbon storage data of the target area by fitting the training samples; S73. Biomass and carbon storage distribution map generated by least squares minimization Real data from the target area The error between them, the loss function is defined as: ; Where N is the number of training samples, represents the Euclidean norm, represents the sparsity regularization term of the model parameters, is a regularization parameter used to control the complexity of the model and prevent overfitting.

[0061] In this implementation, S8 includes the following sub-steps: S81. The optimized inversion model Hyperspectral image data applied to target area , according to the hyperspectral feature vector of each pixel , calculate biomass and carbon stock estimates for pixels : ; in, is a weighting function based on pixel position and wavelength, which is used to adjust the contribution of different spectral bands to biomass and carbon stock estimation. is the hyperspectral feature vector and sparse basis dictionary The similarity kernel function is used to capture the nonlinear relationship between local spectral features and carbon storage. is the sparse coding coefficient, which indicates the contribution of different basis vectors to a specific pixel. is the dictionary basis vector, and are the minimum and maximum wavelength ranges of the hyperspectral data, respectively; S82, traverse all pixels in the target area, and use the inversion model Generate complete biomass and carbon stock distribution maps : ; in, is the weight matrix between pixel position and pixel number, which is used to integrate the estimation results of multiple pixels. is a smoothing function of adjacent pixels, used for spatial smoothing of biomass and carbon storage. is the standard deviation of the Gaussian kernel function, which controls the degree of smoothness, and Respectively represent the coordinate range of the target area in the x and y directions.

[0062] In this implementation, S9 includes the following sub-steps: S91, based on the generated biomass and carbon stock distribution map , classify and analyze the terrain and vegetation types of the target area, and generate a terrain feature map of the target area and vegetation type map ,in, Indicates the distribution characteristics of biomass and carbon storage under different terrain conditions. Indicates the distribution characteristics of biomass and carbon storage under different vegetation types; S92, according to the terrain feature map and vegetation type map , calculate biomass and carbon storage statistics under different terrain conditions and vegetation types and ,in, represents the sum of biomass and carbon storage under specific terrain conditions, Represents the sum of biomass and carbon storage under a specific vegetation type; S93, combined with terrain feature map and vegetation type map , evaluate the carbon storage inversion results of different terrain and vegetation types in the target area, and define the evaluation function : ; in, is the actual biomass and carbon storage data in the target area, and are weight functions of terrain and vegetation types, respectively, which are used to adjust the impact of different terrain and vegetation types on the evaluation results. Represent the relative error distribution of the inversion results and evaluate the accuracy of the inversion results under different conditions; S94, based on evaluation function , generate a biomass and carbon stock analysis report for the target area, which includes the distribution characteristics of biomass and carbon stocks under different terrains and vegetation types, evaluation of inversion results, relative error analysis, and recommendations for ecological monitoring and carbon stock management.

[0063] In some specific application scenarios, the technical solution of the present invention can be implemented as follows: Taking a mountain forest in southwest China as an example, the study area covers an area of ​​about 100 square kilometers with complex terrain, including mountains, hills and plains, and is covered with different types of vegetation, such as evergreen broad-leaved forests, coniferous forests and mixed forests. Due to the diverse vegetation types and complex terrain in the region, traditional biomass and carbon stock inversion methods are difficult to provide accurate estimation results in such a complex environment. The sparse representation-based UAV hyperspectral local area biomass carbon stock inversion method of the present invention has been comprehensively tested and verified in the region to prove its effectiveness and accuracy under complex terrain and diverse vegetation types.

[0064] The research team set up multiple sample plots on different terrain types in the study area, totaling 50 plots. Each plot has an area of ​​about 1 hectare as collection points for training samples and verification samples. The drone is equipped with a hyperspectral imaging sensor and performs multiple flight missions at different altitudes and terrain conditions according to the preset flight path. It collects hyperspectral image data covering the entire study area. The drone flies at an altitude of 300 meters and a speed of 10 meters per second. The spatial resolution of the sensor is 0.5 meters, and the spectral range covers 400 nanometers to 1000 nanometers, with a total of 200 bands.

[0065] After data collection is completed, the method of the present invention is used to process and analyze hyperspectral data. First, the sparse representation technology is used to reduce the dimension of the collected hyperspectral data, a sparse basis dictionary is constructed, and the most representative feature vectors in the hyperspectral data are extracted through the sparse coding algorithm. Subsequently, the primary generative adversarial network is applied to generate preliminary biomass and carbon storage distribution maps, and the generation results are further optimized through the advanced generative adversarial network, so that the biomass and carbon storage distribution maps are more detailed and accurate. Finally, based on the generated biomass and carbon storage distribution maps, an inversion model is constructed, and the inversion model is applied to the entire study area to estimate the biomass and carbon storage distribution under various terrains and vegetation types in the study area.

[0066] To verify the effectiveness of the present invention, the research team conducted ground measurements of biomass and carbon storage in 50 sample plots, and compared the measured data with the inversion results of the method of the present invention and the inversion results of the traditional method. The ground measured data showed that the biomass of different vegetation types in the study area was between 100 tons per hectare and 350 tons per hectare, while the carbon storage was between 50 tons per hectare and 175 tons per hectare. The following Table 1 is a comparison of the inversion results of biomass and carbon storage of the method of the present invention and the traditional method under different vegetation types: Table 1

[0067] It can be seen from the data in Table 1 that the method of the present invention is significantly superior to the traditional method in the accuracy of biomass and carbon stock inversion, especially in complex terrain and diverse vegetation types. The inversion results of the traditional method have large deviations, especially in evergreen broad-leaved forests and mountain vegetation. The deviations between the inversion results of the traditional method and the measured data are 30 tons per hectare and 25 tons per hectare, respectively. The method of the present invention combines sparse representation and generative adversarial network technology to effectively reduce data redundancy, capture the complex relationship between hyperspectral features and biomass and carbon storage, make the inversion results closer to the measured data, and the deviation from the measured data is controlled within 10 tons per hectare, which greatly improves the accuracy of the inversion results.

[0068] In addition, the method of the present invention is also superior to traditional methods in data processing efficiency. The test results of the research team show that the traditional method takes an average of about 12 hours to process 100 square kilometers of hyperspectral data, while the method of the present invention effectively reduces the data dimension through sparse representation technology and optimizes the calculation process by combining generative adversarial network technology, reducing the total time to 8 hours, greatly improving the efficiency of data processing, which is of great significance for application scenarios that require real-time or quasi-real-time monitoring of biomass and carbon reserves.

[0069] Through the verification of this example, the effectiveness and practicability of the present invention in complex terrain and diverse vegetation types have been proved. The biomass and carbon storage distribution map generated by the research team based on the method of the present invention can not only accurately reflect the biomass and carbon storage distribution in the study area, but also provide reliable data support for ecosystem management, carbon sink calculation and environmental protection.

[0070] In summary, the present invention significantly improves the efficiency and accuracy of biomass and carbon stock inversion by combining sparse representation, generative adversarial network and drone hyperspectral imaging technology. The sparse representation technology is used to reduce the dimension of hyperspectral data, and a sparse basis dictionary that can effectively extract key spectral information is constructed. In combination with the sparse coding algorithm, the redundancy and computational complexity of the data are greatly reduced, the speed of data processing is improved, and the important feature information in the hyperspectral data is retained, making the subsequent inversion process more efficient and accurate.

[0071] The present invention designs primary and advanced generative adversarial networks, which gradually optimize the generation quality of biomass and carbon stock distribution maps through a multi-level generative adversarial structure. The primary generative adversarial network generates preliminary biomass and carbon stock distribution maps through adversarial training between the generator and the discriminator. The advanced generative adversarial network further optimizes the generation results based on the primary generative adversarial network, making the distribution map more detailed and accurate. The design of multi-level adversarial generation greatly improves the adaptability of the inversion model to complex terrain and diverse vegetation types, so that the generated biomass and carbon stock distribution map is more in line with the actual situation and has a higher spatial resolution.

[0072] In the construction of the inversion model, the present invention establishes a nonlinear mapping relationship between the biomass and carbon stock distribution maps generated by an advanced generative adversarial network and the hyperspectral feature vectors of the target area, and optimizes the model parameters through the least squares method, effectively capturing the complex relationship between hyperspectral features and biomass and carbon stocks, making the inversion results more accurate and robust, and being able to provide stable and precise estimation results when dealing with diverse terrains and vegetation types.

[0073] On the other hand, Fig. 9 As shown, the embodiment of the present invention provides a biomass carbon stock inversion device 900, which may include: The first module 901 is used to obtain hyperspectral image data of a sample plot area in a target area; the hyperspectral image data includes multi-band reflectance spectral information of surface vegetation in the sample plot area; the sample plot area includes a plurality of sample plots of different terrain types in the target area; The second module 902 is used to preprocess the hyperspectral image data; the preprocessing includes radiation correction, geometric correction and noise removal; The third module 903 is used to construct a sparse basis dictionary based on the preprocessed hyperspectral image data, and extract a feature vector using a sparse coding algorithm; The fourth module 904 is used to generate a first distribution map of biomass and carbon storage based on the feature vector using a preset first generative adversarial network; A fifth module 905 is used to generate a second distribution map of biomass and carbon storage based on the first distribution map by using a preset second generative adversarial network; Among them, the first generative adversarial network and the second generative adversarial network are constructed based on adversarial training of the generator and the discriminator; The sixth module 906 is used to construct an inversion model based on the mapping relationship between the second distribution map and the characteristic vector; The seventh module 907 is used to apply the inversion model to the entire area of ​​the target area to obtain a third distribution map of biomass and carbon storage in the target area.

[0074] In some embodiments, the apparatus may further include: The eighth module is used to obtain the actual distribution map of biomass and carbon storage in the sample area; A ninth module is used to construct a first generator based on a first mapping function between the feature vector and the actual distribution graph; A tenth module is used to construct a first discriminator based on a first discriminant function that distinguishes a generated result of the first generator from an actual distribution graph; An eleventh module is used to generate a first training distribution graph through a first generator according to the feature vector; A twelfth module is used to generate a first discrimination result through a first discriminator according to the first training distribution map and the actual distribution map; A thirteenth module is used to perform a first adversarial training on the first generator and the first discriminator according to a preset first loss function based on the first discrimination result; The fourteenth module is used to optimize the parameters of the first generator and the first discriminator according to the results of the first adversarial training; return to execute the eleventh module until the first iteration number is reached to obtain the first generative adversarial network.

[0075] In some embodiments, the apparatus may further include: Module 15 is used to obtain the actual distribution map of biomass and carbon storage in the sample area; A sixteenth module is used to construct a second generator based on a second mapping function between the first distribution map and the actual distribution map; A seventeenth module is used to construct a second discriminator based on a second discriminant function that distinguishes the generated result of the second generator from the actual distribution graph; An eighteenth module is used to generate a second training distribution map through a second generator according to the first distribution map; A nineteenth module is used to generate a second discrimination result through a second discriminator according to the second training distribution map and the actual distribution map; A twentieth module, configured to perform second adversarial training on the second generator and the second discriminator according to a preset second loss function based on the second discrimination result; The twenty-first module is used to optimize the parameters of the second generator and the second discriminator according to the results of the second adversarial training; return to execute the eighteenth module until the second iteration number is reached to obtain the second generative adversarial network.

[0076] The contents of the method embodiments of the present invention are all applicable to the device embodiments. The functions specifically implemented by the device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0077] On the other hand, an embodiment of the present invention further provides an electronic device, the electronic device comprising a memory and a processor, the memory storing a computer program, and the processor implementing the above-mentioned hydrate stability region bottom boundary prediction method when executing the computer program. The electronic device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0078] It can be understood that the contents of the above method embodiments are all applicable to the present device embodiments, the functions specifically implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0079] like Fig.10 As shown, Fig.10 The hardware structure of an electronic device 1000 of another embodiment is illustrated. The electronic device 1000 includes: The processor 1001 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present invention. The memory 1002 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1002 can store an operating system and other application programs. When the technical solution provided in the embodiment of this specification is implemented by software or firmware, the relevant program code is stored in the memory 1002, and the processor 1001 calls and executes the network node population optimization method of the embodiment of the present invention; Input / output interface 1003, used to implement information input and output; Communication interface 1004, used to realize communication interaction between the device and other devices, which can be realized through wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.); A bus 1005 , which transmits information between various components of the device (e.g., the processor 1001 , the memory 1002 , the input / output interface 1003 , and the communication interface 1004 ); The processor 1001 , the memory 1002 , the input / output interface 1003 and the communication interface 1004 are connected to each other in communication within the device via the bus 1005 .

[0080] The electronic device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0081] The contents of the method embodiments of the present invention are all applicable to the electronic device embodiments. The functions specifically implemented by the electronic device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0082] Another aspect of an embodiment of the present invention further provides a computer-readable storage medium, the storage medium stores a program, and the program is executed by a processor to implement the above method.

[0083] It should be noted that the computer-readable medium shown in the embodiment of the present invention may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (Compact Disc Read to Only Memory, CD to ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, device or device. In the present invention, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer readable signal media may also be any computer readable medium other than computer readable storage media, which may send, propagate, or transmit programs for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wired, etc., or any suitable combination of the above.

[0084] The contents of the method embodiments of the present invention are all applicable to the computer-readable storage medium embodiments. The functions specifically implemented by the computer-readable storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0085] The embodiment of the present invention also discloses a computer program product or a computer program, which includes a computer instruction stored in a computer-readable storage medium. A processor of a computer device can read the computer instruction from the computer-readable storage medium, and the processor executes the computer instruction, so that the computer device executes the above method.

[0086] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present invention. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0087] It should be noted that, although several modules of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to an embodiment of the present invention, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into being embodied by multiple modules or units.

[0088] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD to ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a touch terminal, or a network device, etc.) to execute the method according to the implementation of the present invention.

[0089] In some selectable embodiments, the function / operation mentioned in the block diagram may not occur in the order mentioned in the operation diagram. For example, depending on the function / operation involved, the two boxes shown in succession can actually be executed substantially simultaneously or the boxes can sometimes be executed in reverse order. In addition, the embodiment presented and described in the flow chart of the present invention is provided by way of example, for the purpose of providing a more comprehensive understanding of technology. The disclosed method is not limited to the operation and logic flow presented by the present invention. Selectable embodiments are expected, wherein the order of various operations is changed and the sub-operation of a part described as a larger operation is performed independently.

[0090] In addition, although the present invention is described in the context of functional modules, it should be understood that, unless otherwise specified, one or more of the functions and / or features can be integrated into a single physical device and / or software module, or one or more functions and / or features can be implemented in separate physical devices or software modules. It is also understood that a detailed discussion of the actual implementation of each module is unnecessary for understanding the present invention. More specifically, in view of the properties, functions and internal relationships of various functional modules in the device disclosed in the present invention, the actual implementation of the module will be understood within the conventional skills of engineers. Therefore, those skilled in the art can implement the present invention set forth in the claims without excessive experimentation using ordinary techniques. It is also understood that the specific concepts disclosed are merely illustrative and are not intended to limit the scope of the present invention, which is determined by the full scope of the appended claims and their equivalents.

[0091] If the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0092] The logic and / or steps represented in the flowchart or otherwise described herein, for example, may be considered as an ordered list of executable instructions for implementing logical functions, and may be embodied in any computer-readable medium for use by an instruction execution device, apparatus or device (such as a computer-based device, a device including a processor, or other device that can fetch instructions from an instruction execution device, apparatus or device and execute the instructions), or in conjunction with such instruction execution device, apparatus or device. For purposes of this specification, "computer-readable medium" may be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution device, apparatus or device, or in conjunction with such instruction execution device, apparatus or device.

[0093] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or processing in another suitable manner as necessary, and then stored in a computer memory.

[0094] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution device. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0095] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0096] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

[0097] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the embodiments. Those skilled in the art can make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of the present invention.

Claims

1. A biomass carbon stock inversion method, characterized in that: The following steps are involved: Acquire hyperspectral image data of a sample plot area in a target area; the hyperspectral image data includes multi-band reflectance spectral information of surface vegetation in the sample plot area; the sample plot area includes a plurality of sample plots of different terrain types in the target area; Preprocessing the hyperspectral image data; the preprocessing includes radiation correction, geometric correction and noise removal; Constructing a sparse basis dictionary based on the preprocessed hyperspectral image data, and extracting a feature vector using a sparse coding algorithm; Based on the feature vector, a first distribution map of biomass and carbon storage is generated using a preset first generative adversarial network; Based on the first distribution map, a second distribution map of biomass and carbon storage is generated using a preset second generative adversarial network; The first generative adversarial network and the second generative adversarial network are constructed based on adversarial training of a generator and a discriminator; An inversion model is constructed based on a mapping relationship between the second distribution map and the characteristic vector; The inversion model is applied to the entire area of ​​the target area to obtain a third distribution map of biomass and carbon storage in the target area.

2. The biomass carbon stock inversion method according to claim 1, characterized in that: The method of obtaining the hyperspectral image data of the sample area in the target area comprises the following steps: In response to a control instruction of a target object, a target UAV is controlled to fly to the sample area of ​​the target area based on preset operating parameters; a hyperspectral imaging sensor is provided on the target UAV; Acquire the multi-band reflectance spectrum information collected by the hyperspectral imaging sensor during the flight of the target UAV in the sample plot area; According to the flight path and time series of the target UAV, all the multi-band reflectance spectral information is spliced ​​and registered to obtain the hyperspectral image data of the sample area.

3. The biomass carbon stock inversion method according to claim 1, characterized in that: The method of constructing a sparse basis dictionary based on the preprocessed hyperspectral image data and extracting a feature vector using a sparse coding algorithm comprises the following steps: An initial sparse basis dictionary is constructed based on the preprocessed hyperspectral image data; the expression of the initial sparse basis dictionary is: ,in, represents the i-th basis vector, k is the number of basis vectors in the initial sparse basis dictionary; The initial sparse basis dictionary is optimized to obtain an optimized sparse basis dictionary; the expression of the optimized sparse basis dictionary is: ； in, represents the optimized sparse basis dictionary; represents the hyperspectral image data after the preprocessing, represents the surface coordinates, Indicates wavelength; Represents the sparse coefficient corresponding to the i-th basis vector; is the regularization parameter; The pre-processed hyperspectral image data is sparsely represented using the sparse coding algorithm to obtain a sparse representation coefficient vector of the spectral vector corresponding to each hyperspectral pixel; the expression of the sparse representation coefficient vector is: ； in, represents the sparse representation coefficient vector; represents the initial sparse base dictionary; Representing the sparse coefficient corresponding to the hyperspectral pixel; and is the regularization parameter; represents the sparse coefficient corresponding to the j-th pixel around the hyperspectral pixel, and m is the number of pixels around the hyperspectral pixel; Extracting an initial feature vector based on the sparse representation coefficient vector in combination with the optimized sparse basis dictionary; The initial feature vector is sparsely encoded and dimensionally reduced to obtain the feature vector.

4. The biomass carbon stock inversion method according to claim 1, characterized in that: The method further comprises the following steps: Obtaining actual distribution maps of biomass and carbon storage in the sample area; A first generator is constructed based on a first mapping function between the feature vector and the actual distribution graph; A first discriminant is constructed based on a first discriminant function that distinguishes the generated result of the first generator from the actual distribution graph; generating a first training distribution graph by the first generator according to the feature vector; Generate a first discrimination result by the first discriminator according to the first training distribution graph and the actual distribution graph; Based on the first discrimination result, performing first adversarial training on the first generator and the first discriminator according to a preset first loss function; Optimize the parameters of the first generator and the first discriminator according to the result of the first adversarial training; return to execute the step of generating a first training distribution map through the first generator according to the feature vector until the first number of iterations is reached, and obtain the first generative adversarial network.

5. The biomass carbon stock inversion method according to claim 1, characterized in that: The method further comprises the following steps: Obtaining actual distribution maps of biomass and carbon storage in the sample area; A second generator is constructed based on a second mapping function between the first distribution map and the actual distribution map; A second discriminant is constructed based on a second discriminant function that distinguishes the generated result of the second generator from the actual distribution graph; According to the first distribution map, generating a second training distribution map by the second generator; Generate a second discrimination result by the second discriminator according to the second training distribution map and the actual distribution map; Based on the second discrimination result, performing second adversarial training on the second generator and the second discriminator according to a preset second loss function; Optimize the parameters of the second generator and the second discriminator according to the result of the second adversarial training; return to execute the step of generating the second training distribution map through the second generator according to the first distribution map until the second number of iterations is reached to obtain the second generative adversarial network.

6. The biomass carbon stock inversion method according to claim 1, characterized in that: The step of constructing an inversion model based on the mapping relationship between the second distribution map and the characteristic vector comprises the following steps: Obtaining actual distribution maps of biomass and carbon storage in the sample area; According to the mapping function between the characteristic vector and the actual distribution map, the least square method is used to perform error optimization on the mapping relationship between the second distribution map and the characteristic vector, and the inversion model is obtained by fitting.

7. The biomass carbon stock inversion method according to claim 1, characterized in that: The method further comprises the following steps: Obtaining a true distribution map of biomass and carbon storage in the target area; Based on the third distribution map, a classification analysis of the terrain and vegetation types of the target area is performed to obtain a terrain feature map and a vegetation type map of the target area; According to the terrain feature map and the vegetation type map, statistical values ​​under different terrain conditions and vegetation types are obtained; the statistical values ​​include biomass statistical values ​​and carbon storage statistical values; According to the terrain feature map and the vegetation type map, combined with the real distribution map, the statistical values ​​under different terrain conditions and vegetation types in the target area are evaluated and analyzed to obtain an analysis report on the biomass and carbon storage of the target area.

8. A biomass carbon stock inversion device, characterized in that: include: The first module is used to obtain hyperspectral image data of the sample area in the target area; The hyperspectral image data includes multi-band reflectance spectral information of surface vegetation in the sample plot area; the sample plot area includes a plurality of sample plots of different terrain types in the target area; The second module is used to preprocess the hyperspectral image data; the preprocessing includes radiation correction, geometric correction and noise removal; The third module is used to construct a sparse basis dictionary based on the preprocessed hyperspectral image data, and extract a feature vector using a sparse coding algorithm; A fourth module is used to generate a first distribution map of biomass and carbon storage based on the feature vector using a preset first generative adversarial network; A fifth module is used to generate a second distribution map of biomass and carbon storage based on the first distribution map by using a preset second generative adversarial network; The first generative adversarial network and the second generative adversarial network are constructed based on adversarial training of a generator and a discriminator; A sixth module is used to construct an inversion model based on the mapping relationship between the second distribution map and the characteristic vector; The seventh module is used to apply the inversion model to the entire area of ​​the target area to obtain a third distribution map of biomass and carbon storage in the target area.

9. An electronic device, characterized in that: including a processor and a memory; The memory is used to store programs; The processor executes the program to implement the method according to any one of claims 1 to 7.

10. A computer storage medium storing a program executable by a processor, characterized in that: The program executable by the processor is used to implement the method according to any one of claims 1 to 7 when executed by the processor.

Images