Plant carbon storage estimation method and system based on satellite remote sensing and Internet of Things technology

Abstract

The present application provides a plant carbon stock estimation method and system based on satellite remote sensing and Internet of Things technology, which relates to the field of forest resource management. The method first obtains TM remote sensing images and DEM images of a certain area, then extracts features from the acquired TM remote sensing images and DEM images of a certain area to obtain remote sensing factor features and geological information features respectively, and then fuses the above features to obtain biomass-related features. Finally, the obtained biomass-related features are decoded by a decoder to generate the plant carbon stock value of the area.

Description

Technical Field

[0001] The present application relates to the field of forest resource management, and more specifically, to a plant carbon storage estimation method and system based on satellite remote sensing and Internet of Things technology. Background Art

[0002] Plants play a vital role as one of the most important biomass carbon reservoirs on Earth. They absorb large amounts of carbon dioxide through photosynthesis, convert it into organic matter, and release oxygen. This process is essential for the stability of the Earth's climate and the health of its ecosystems.

[0003] By estimating plant carbon storage, we can better understand the contribution of plants in the carbon cycle. This is very important for the study of global carbon balance, the prediction of climate change trends, and the formulation of policies and measures to address climate change. Estimating plant carbon storage can also help evaluate the differences in carbon storage in different regions and different types of vegetation, reveal the dynamic changes in plant growth and ecosystems, and provide a scientific basis for ecosystem management and protection. The results of plant carbon storage estimation are of great significance to the study of global carbon cycle and climate change. They can help us understand the efficiency of plants in absorbing and storing carbon dioxide, as well as their regulation of carbon dioxide concentration in the atmosphere.

[0004] Therefore, a plant carbon storage estimation scheme based on satellite remote sensing and Internet of Things technology is needed. Summary of the invention

[0005] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a plant carbon storage estimation method and system based on satellite remote sensing and Internet of Things technology, which first obtains TM remote sensing images and DEM images of a certain area, then extracts features from the acquired TM remote sensing images and DEM images of a certain area to obtain remote sensing factor features and geological information features respectively, and then fuses the above features to obtain biomass-related features, and finally decodes the obtained biomass-related features through a decoder to generate the plant carbon storage value of the area.

[0006] According to one aspect of the present application, a method for estimating plant carbon storage based on satellite remote sensing and Internet of Things technology is provided, which comprises:

[0007] Obtain TM remote sensing images and DEM images of a certain area;

[0008] Performing biomass feature coding on the TM remote sensing image and the DEM image of the certain area to obtain a biomass information feature map;

[0009] Based on the biomass information characteristic map, the plant carbon storage value of the certain area is obtained.

[0010] In the above-mentioned plant carbon stock estimation method based on satellite remote sensing and Internet of Things technology, the biomass feature encoding of the TM remote sensing image and the DEM image of the certain area to obtain a biomass information feature map includes: extracting remote sensing factor features of the TM remote sensing image of the certain area to obtain a remote sensing factor feature map; extracting geoscientific factor features of the DEM image of the certain area to obtain a geoscientific information feature map; and fusing the remote sensing factor feature map and the geoscientific information feature map to obtain the biomass information feature map.

[0011] In the above-mentioned plant carbon stock estimation method based on satellite remote sensing and Internet of Things technology, the remote sensing factor feature extraction of the TM remote sensing image of the certain area to obtain the remote sensing factor feature map includes: declouding the TM remote sensing image of the certain area to obtain a declouded TM remote sensing image; extracting the band reflectance of different bands from the declouded TM remote sensing image; performing different operations on the band reflectances of the different bands to respectively obtain multiple vegetation indices, multiple principal component analysis band linear transformation values, and multiple tasseled cap transformation band linear transformation values; arranging the multiple vegetation indices, the multiple principal component analysis band linear transformation values, and the multiple tasseled cap transformation band linear transformation values ​​in two dimensions to obtain a remote sensing factor data matrix; and passing the remote sensing factor data matrix through a convolutional neural network model as a feature extractor to obtain the remote sensing factor feature map.

[0012] In the above-mentioned plant carbon storage estimation method based on satellite remote sensing and Internet of Things technology, the remote sensing factor data matrix is ​​passed through a convolutional neural network model as a feature extractor to obtain the remote sensing factor feature map, including: each layer of the convolutional neural network model as a feature extractor performs the following on the input data during the forward transfer process of the layer: convolution processing on the input data to obtain a convolution feature map; mean pooling processing based on a local feature matrix on the convolution feature map to obtain a pooling feature map; nonlinear activation on the pooling feature map to obtain an activation feature map; wherein the output of the last layer of the convolutional neural network model is the remote sensing factor feature map, and the input of the first layer of the convolutional neural network model is the remote sensing factor data matrix.

[0013] In the above-mentioned plant carbon stock estimation method based on satellite remote sensing and Internet of Things technology, the extraction of geo-factor features from the DEM image of the certain area to obtain a geo-information feature map includes: calculating the DEM image of the certain area to obtain a slope aspect grayscale map and a slope grayscale map respectively; passing the slope aspect grayscale map and the slope grayscale map through multiple convolution layers respectively to obtain a slope aspect grayscale feature map and a slope grayscale feature map; and cascading the slope aspect grayscale feature map and the slope grayscale feature map to obtain the geo-information feature map.

[0014] In the above-mentioned plant carbon storage estimation method based on satellite remote sensing and Internet of Things technology, the cascading of the aspect grayscale feature map and the slope grayscale feature map to obtain the geoscientific information feature map includes: fusing the aspect grayscale feature map and the slope grayscale feature map with the following cascade formula to obtain the geoscientific information feature map; wherein the cascade formula is:

[0015] F = Concat[F a ,F b ]

[0016] Among them, F a represents the slope grayscale feature map, F b represents the slope grayscale feature map, F represents the geological information feature map, and Concat[·,·] represents a cascade function.

[0017] In the above-mentioned plant carbon storage estimation method based on satellite remote sensing and Internet of Things technology, the fusion of the remote sensing factor characteristic map and the geoscientific information characteristic map to obtain the biomass information characteristic map includes: calculating the posterior hidden feature attention of the matching nodes between the remote sensing factor characteristic map and the geoscientific information characteristic map by the following formula to obtain the biomass information characteristic map;

[0018] Wherein, the formula is:

[0019]

[0020] Among them, M 1i represents the characteristic matrix of the ith channel of the remote sensing factor characteristic map, M 2i represents the feature matrix of the ith channel of the geoscience information feature map, T represents the transpose of the vector, C represents the number of channels of the feature map, ⊙ represents the positional dot multiplication of the feature matrix, Represents matrix multiplication, M ci The feature matrix representing the i-th channel of the biomass information feature map.

[0021] In the above-mentioned plant carbon stock estimation method based on satellite remote sensing and Internet of Things technology, the plant carbon stock value of the certain area is obtained based on the biomass information characteristic map, including: passing the biomass information characteristic map through a decoder to obtain a decoded value, and the decoded value is used to represent the plant carbon stock value of the certain area.

[0022] According to another aspect of the present application, a plant carbon storage estimation system based on satellite remote sensing and Internet of Things technology is provided, which includes:

[0023] The estimation related data acquisition module is used to obtain TM remote sensing images and DEM images of a certain area;

[0024] An estimation-related data encoding module is used to perform biomass feature encoding on the TM remote sensing image and the DEM image of the certain area to obtain a biomass information feature map;

[0025] The carbon storage estimation result generation module is used to obtain the plant carbon storage value of the certain area based on the biomass information characteristic map.

[0026] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0028] Figure 1 This is a flow chart of a method for estimating plant carbon storage based on satellite remote sensing and Internet of Things technology according to an embodiment of the present application.

[0029] Figure 2 This is an architectural diagram of a plant carbon storage estimation method based on satellite remote sensing and Internet of Things technology according to an embodiment of the present application.

[0030] Figure 3 The present invention is a flow chart of performing biomass feature encoding on TM remote sensing images and DEM images of a certain area to obtain a biomass information feature map in a plant carbon storage estimation method based on satellite remote sensing and Internet of Things technology according to an embodiment of the present application.

[0031] Figure 4 The present invention is a flow chart of extracting remote sensing factor features from a TM remote sensing image of a certain area to obtain a remote sensing factor feature map in a plant carbon storage estimation method based on satellite remote sensing and Internet of Things technology according to an embodiment of the present application.

[0032] Figure 5 The present invention is a flow chart of extracting geo-factor features from a DEM image of a certain area to obtain a geo-information feature map in a plant carbon storage estimation method based on satellite remote sensing and Internet of Things technology according to an embodiment of the present application.

[0033] Figure 6 4 is a system block diagram of a plant carbon storage estimation system based on satellite remote sensing and Internet of Things technology according to an embodiment of the present application. DETAILED DESCRIPTION

[0034] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0035] As mentioned in the above background technology, plants are one of the most important biological carbon reservoirs on the earth. They absorb a large amount of carbon dioxide through photosynthesis and convert it into organic matter while releasing oxygen. Plant carbon storage estimation can help understand the role of plants in the carbon cycle, evaluate their regulatory effects on atmospheric carbon dioxide concentrations, and respond to global climate change. By estimating the carbon storage of plants, we can quantify the contribution of plants to the carbon cycle. This is crucial for understanding the global carbon balance, predicting climate change trends, and formulating policies and measures to address climate change. Plant carbon storage estimation can also help evaluate the differences in carbon storage in different regions and different types of vegetation, reveal the dynamic changes in plant growth and ecosystems, and provide a scientific basis for ecosystem management and protection. Therefore, a plant carbon storage estimation scheme based on satellite remote sensing and Internet of Things technology is expected.

[0036] In recent years, deep learning and neural networks have been widely used in computer vision, natural language processing, text signal processing and other fields. In addition, deep learning and neural networks have also shown a level close to or even beyond that of humans in image classification, object detection, semantic segmentation, text translation and other fields. The development of deep learning and neural networks has provided new solutions and solutions for the estimation of plant carbon storage based on satellite remote sensing and Internet of Things technology.

[0037] Figure 1 This is a flow chart of a method for estimating plant carbon storage based on satellite remote sensing and Internet of Things technology according to an embodiment of the present application. Figure 2 FIG. 1 is a schematic diagram of a plant carbon storage estimation method based on satellite remote sensing and Internet of Things technology according to an embodiment of the present application. Figure 1 and Figure 2 As shown, according to the plant carbon stock estimation method based on satellite remote sensing and Internet of Things technology according to the embodiment of the present application, the method includes: S110, acquiring TM remote sensing images and DEM images of a certain area; S120, performing biomass feature encoding on the TM remote sensing images and DEM images of the certain area to obtain a biomass information feature map; S130, obtaining the plant carbon stock value of the certain area based on the biomass information feature map.

[0038] Specifically, in the technical solution of the present application, first, TM remote sensing images and DEM images of a certain area are obtained. It should not be difficult to understand that TM remote sensing images (Thematic Mapper) are multispectral remote sensing images acquired by satellites, which contain information of different bands. In the estimation of plant carbon reserves, TM remote sensing images can provide reflectance data of vegetation, and by analyzing the reflectance of different bands, the biomass information of plants can be obtained. The reflectance of different bands in remote sensing images is related to the light absorption and reflection characteristics of vegetation, which can reflect the growth status of plants and the intensity of photosynthesis. Therefore, the purpose of acquiring TM remote sensing images is to extract vegetation index and other related features from them for the estimation of plant carbon reserves. DEM images (Digital Elevation Model) are digital models that describe terrain elevation. In the estimation of plant carbon reserves, DEM images are used to calculate terrain information, such as slope and aspect. Terrain has an important influence on plant growth and distribution, and terrain conditions with different slopes and aspects will affect the biomass distribution of plants. By calculating the DEM image, we can obtain the grayscale image of terrain information, extract the characteristics of geological information, and then fuse it with remote sensing data to improve the accuracy of plant carbon storage estimation. The TM remote sensing image of a certain area can be obtained through a professional remote sensing data platform, and the DEM image of a certain area can be obtained through a remote sensing data platform or through a geospatial data cloud website.

[0039] Figure 3 The present invention is a flow chart of a method for estimating plant carbon storage based on satellite remote sensing and Internet of Things technology, in which the TM remote sensing image and DEM image of a certain area are subjected to biomass feature coding to obtain a biomass information feature map. Figure 3 As shown, it includes: S121, extracting remote sensing factor features from the TM remote sensing image of the certain area to obtain a remote sensing factor feature map; S122, extracting geoscientific factor features from the DEM image of the certain area to obtain a geoscientific information feature map; S123, fusing the remote sensing factor feature map and the geoscientific information feature map to obtain the biomass information feature map.

[0040] Figure 4 This is a flow chart of extracting remote sensing factor features from the TM remote sensing image of a certain area to obtain a remote sensing factor feature map in a plant carbon storage estimation method based on satellite remote sensing and Internet of Things technology according to an embodiment of the present application. Figure 4As shown, it includes: S1211, performing cloud removal processing on the TM remote sensing image of the certain area to obtain a cloud removal TM remote sensing image; S1212, extracting the band reflectance of different bands from the cloud removal TM remote sensing image; S1213, performing different operations on the band reflectances of the different bands to respectively obtain multiple vegetation indices, multiple principal component analysis band linear transformation values, and multiple tassel cap transformation band linear transformation values; S1214, arranging the multiple vegetation indices, the multiple principal component analysis band linear transformation values, and the multiple tassel cap transformation band linear transformation values ​​in two dimensions to obtain a remote sensing factor data matrix; S1215, passing the remote sensing factor data matrix through a convolutional neural network model as a feature extractor to obtain the remote sensing factor feature map.

[0041] Cloud cover is one of the common interference factors in remote sensing images. Clouds block the entry of solar radiation, resulting in changes in the surface reflectance spectrum, making the vegetation information in the remote sensing image inaccurate. Cloud cover can also cause shadows, making the lighting conditions on the surface uneven. Therefore, in the technical solution of the present application, the TM remote sensing image of the certain area is declouded to obtain a declouded TM remote sensing image. By removing the cloud cover, an image with less or no cloud shadow cover is obtained to obtain more accurate vegetation information. This can reduce the interference of cloud cover on plant carbon storage estimation and improve the reliability and accuracy of the estimation results.

[0042] In an embodiment of the present application, a possible implementation method for performing cloud removal processing on the TM remote sensing image of the certain area to obtain a cloud-removed TM remote sensing image may be: 1. First, preprocessing the TM remote sensing image, including radiation correction, atmospheric correction and geometric correction, etc. These steps can improve the quality and accuracy of the TM remote sensing image; 2. Using a cloud detection algorithm to detect clouds in the TM remote sensing image. Commonly used cloud detection algorithms include threshold method, texture analysis method, morphological processing method, etc. These algorithms can detect clouds based on the brightness, texture and shape of the cloud layer, and generate a cloud mask; 3. Based on the cloud Detection results, generate a cloud mask, the cloud mask is a binary image, in which the cloud area is marked as 1 and the non-cloud area is marked as 0; 4. According to the cloud mask, the reflectivity of the TM remote sensing image is replaced. For the cloud area, the reflectivity value of the adjacent cloud-free area can be used for replacement. The interpolation method (such as nearest neighbor interpolation, bilinear interpolation, etc.) can be used to estimate the reflectivity value of the cloud-free area and apply it to the cloud area; 5. Based on the reflectivity replacement result, generate a de-clouded TM remote sensing image. In the de-clouded image, the cloud area is replaced with the reflectivity value of the cloud-free area, while the non-cloud area remains unchanged.

[0043] Next, the band reflectance of different bands is extracted from the de-clouded TM remote sensing image. It is not difficult to understand that the spectral information of vegetation exists in the de-clouded TM remote sensing image. The spectral reflectance characteristics of plants are closely related to their physiological conditions and tissue structures. Electromagnetic radiation of different bands has different reflectances on the surface of vegetation, which reflects the absorption and reflection ability of plants to light of different bands. By analyzing the reflectance of different bands, information such as vegetation coverage can be inferred. TM remote sensing images usually contain multiple bands, such as red light, near infrared, short-wave infrared, etc. The reflectance of these bands can be used to calculate vegetation indices, such as the normalized difference vegetation index (NDVI), the enhanced vegetation index (EVI), etc. Vegetation indices are calculated based on the ratio or difference between the reflectances of different bands. They can provide information on vegetation growth and photosynthesis, and are then used to estimate plant carbon storage.

[0044] In an embodiment of the present application, an implementation method for extracting band reflectances of different bands from the de-clouded TM remote sensing image may be: 1. Selecting a region of interest (ROI), that is, selecting an area representing vegetation coverage in the de-clouded remote sensing image, which may be performed by an automatic classification algorithm; 2. For each band, calculating the average reflectance of the pixels in the ROI. The reflectance is typically expressed as a value between 0 and 1, which may be calculated by dividing the digital value by the radiation correction coefficient of the sensor.

[0045] In order to extract different and convenient vegetation information to more accurately estimate the carbon storage of plants, the band reflectance of the different bands is calculated differently in the technical solution of the present application to obtain multiple vegetation indices, multiple principal component analysis band linear transformation values, and multiple tasseled cap transformation band linear transformation values. It should be well understood that the vegetation index is an indicator that reflects the growth status and photosynthesis capacity of vegetation by calculating the ratio or difference between the reflectance of different bands. Common vegetation indices include the normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), ratio vegetation index (SR), difference vegetation index (DVI), etc. These indices use the difference in reflectance between red and near-infrared bands to reflect the chlorophyll content, growth status and photosynthesis capacity of vegetation, thereby providing the information required for plant carbon storage estimation. In remote sensing data, each band reflects certain ground object information to varying degrees, but due to the large number of bands, the complexity of data processing and analysis will be increased when applied. In order to reduce the correlation coefficient between bands, remove a large amount of redundant information between bands, better extract vegetation information from bands, and enhance the information function in the image spectrum, it is necessary to change the principal components of the image. Principal component analysis (PCA) is a commonly used multispectral image processing method. It generates a set of new principal component bands by linearly combining the original bands. These principal component bands are linear combinations of the original bands and have different energy distributions and information expression capabilities. In the estimation of plant carbon storage, PCA can be used to extract the main change direction in the vegetation spectrum and capture key information on plant growth and photosynthesis. Tasseled Cap Transformation is a method for converting multispectral data into Tasseled Cap Space. Tasseled Cap Space is a set of band linear transformation values ​​related to vegetation growth and land cover. These band linear transformation values ​​represent information such as vegetation growth status, soil moisture, and land cover. By calculating the linear transformation values ​​of the Tasseled Cap Transformation bands, vegetation and land characteristics can be extracted and used for plant carbon storage estimation.

[0046] In an embodiment of the present application, different operations are performed on the band reflectances of the different bands to obtain multiple vegetation indices, multiple principal component analysis band linear transformation values, and multiple tassel cap transformation band linear transformation values. One possible implementation method can be: 1. Calculate the vegetation index: Different vegetation indices have different vegetation index calculation formulas. For the required vegetation index, each pixel is calculated, and the vegetation index value is calculated using the corresponding formula. The result is normalized to limit the value to an appropriate range; 2. Principal component analysis: Select the band that needs to be PCA, and standardize the selected band so that its mean is 0. , the variance is 1, and a covariance matrix or a correlation matrix is ​​constructed. The covariance matrix is ​​decomposed by eigenvalue to obtain eigenvalues ​​and eigenvectors. According to the size of the eigenvalues, the first N principal components are selected (N is the required number of principal components), and the original bands are linearly combined with the selected principal components to obtain the linear transformation values ​​of the principal component analysis bands; 3. Tasseled cap transformation: Select the bands that need to be tasseled cap transformed, and define tasseled cap transformation coefficients for each band. These coefficients can be selected or calibrated according to the study area and data set, and each pixel is linearly transformed. The original band is multiplied by the tasseled cap transformation coefficient to obtain the linear transformation value of the tasseled cap transformed band.

[0047] In order to combine different remote sensing indicators and provide more comprehensive remote sensing information, the multiple vegetation indices, the multiple principal component analysis band linear transformation values, and the multiple tasseled cap transformation band linear transformation values ​​are arranged in two dimensions to obtain a remote sensing factor data matrix. By arranging multiple remote sensing indicators into a matrix, the data of different indicators can be integrated together to form a unified data structure. This can simplify the data processing process and make data access and operation more efficient. The remote sensing factor data matrix is ​​a two-dimensional matrix, in which each row represents a remote sensing indicator and each column represents a specific value of a remote sensing indicator.

[0048] Next, the remote sensing factor data matrix is ​​passed through a convolutional neural network model as a feature extractor to obtain a remote sensing factor feature map. Those skilled in the art should know that convolutional neural networks excel in feature extraction, wherein the convolution layer performs a convolution operation on the input data through a convolution kernel to extract local features of the input data, introduces a nonlinear activation function, and increases the expressive power of the model, while the pooling layer reduces the dimension of the features through downsampling operations to extract more abstract features. Through the convolutional neural network model as a feature extractor, key features in remote sensing factor data can be learned and extracted to further improve the accuracy of plant carbon storage estimation.

[0049] In an embodiment of the present application, a possible implementation method of passing the remote sensing factor data matrix through a convolutional neural network model as a feature extractor to obtain a remote sensing factor feature map may be as follows: each layer of the convolutional neural network model as a feature extractor performs the following on the input data during the forward transfer process of the layer: convolution processing on the input data to obtain a convolution feature map; mean pooling processing based on a local feature matrix on the convolution feature map to obtain a pooled feature map; nonlinear activation on the pooled feature map to obtain an activated feature map; wherein the output of the last layer of the convolutional neural network model as a feature extractor is the remote sensing factor feature map, and the input of the first layer of the convolutional neural network model as a feature extractor is the remote sensing factor data matrix.

[0050] Figure 5 The present invention is a flow chart of extracting geo-factor features from the DEM image of a certain area to obtain a geo-information feature map in a plant carbon storage estimation method based on satellite remote sensing and Internet of Things technology according to an embodiment of the present application. Figure 5 As shown, it includes: S1221, calculating the DEM image of the certain area to obtain a slope aspect grayscale map and a slope grayscale map respectively; S1222, passing the slope aspect grayscale map and the slope grayscale map through multiple convolution layers to obtain a slope aspect grayscale feature map and a slope grayscale feature map respectively; S1223, cascading the slope aspect grayscale feature map and the slope grayscale feature map to obtain the geoscience information feature map.

[0051] In order to obtain terrain information and provide the characteristics of the plant growth environment, the DEM image of the area is calculated in the technical solution of the present application to obtain the aspect grayscale map and the slope grayscale map respectively. In the estimation of plant carbon storage, the elevation information of the DEM image is very important for understanding the plant growth environment. Slope refers to the slope or inclination of the surface at a certain point. By calculating the DEM image, the slope value of each pixel can be obtained to form a slope grayscale map. The slope grayscale map can show the slope distribution of the surface and reflect the steepness of the terrain. The slope has a great influence on the plant growth environment. Terrains with different slopes will affect the distribution of water, soil erosion and the growth of plant roots. Aspect refers to the direction or orientation of the surface at a certain point. By calculating the DEM image, the aspect value of each pixel can be obtained to form an aspect grayscale map. The aspect grayscale map can show the aspect distribution of the surface, that is, the change of the surface orientation. Slope direction also has an important impact on the growth environment of plants. Topography with different slope directions will affect factors such as sunlight, temperature distribution, and water evaporation. These geoscientific information are of great significance to plant growth. The purpose of calculating DEM images is to obtain these key information to improve the accuracy of estimation.

[0052] In an embodiment of the present application, a possible implementation method for calculating the DEM image of the certain area to obtain the aspect grayscale map and the slope grayscale map respectively may be: 1. Preprocessing the acquired DEM data, including removing outliers, filling blank values, smoothing, etc. These preprocessing steps help to improve the accuracy of subsequent calculations; 2. Calculating the slope: using a computer algorithm to calculate the slope value of each pixel point. The slope can be obtained by calculating the elevation change rate of each pixel point in the DEM data. The commonly used calculation formula is to use the central difference method or the slope algorithm to calculate the slope of each pixel point; 3. Calculating the slope direction: using a computer algorithm to calculate the slope value of each pixel point. Calculate the aspect value of each pixel. The aspect can be obtained by calculating the direction of each pixel in the DEM data. The commonly used calculation formula is to use the central difference method or trigonometric function method to calculate the aspect of each pixel; 4. Normalize the calculated slope and aspect values ​​and convert them into grayscale images. Normalization can map the slope and aspect values ​​to a suitable grayscale range for visualization and subsequent analysis; 5. Generate a slope grayscale map and an aspect grayscale map: map the normalized slope value to the pixel value of the grayscale image to generate a slope grayscale map; map the normalized aspect value to the pixel value of the grayscale image to generate an aspect grayscale map.

[0053] In order to extract higher-level feature information from the aspect grayscale feature map and the slope grayscale feature map, the aspect grayscale map and the slope grayscale map are respectively passed through multiple convolution layers to obtain the aspect grayscale feature map and the slope grayscale feature map. The convolution layer is a neural network layer commonly used in deep learning. It performs convolution operations on the input data through convolution kernels (also called filters) to extract local features of the data. In this case, we can regard the aspect grayscale map and the slope grayscale map as input data and extract their features through multiple convolution layers. Through multiple convolution operations, feature information such as edges, textures, etc. in the aspect grayscale map and the slope grayscale map can be gradually extracted. These feature maps can better represent the characteristics of the plant growth environment, thereby providing richer information for plant carbon storage estimation.

[0054] In an embodiment of the present application, the aspect grayscale map and the slope grayscale map are respectively passed through multiple convolution layers to obtain an aspect grayscale feature map and a slope grayscale feature map. One possible implementation method may be: 1. The aspect grayscale map or the slope grayscale map is passed as input data to the first convolution layer. Note that the feature extraction of the aspect grayscale map and the slope grayscale map is performed separately; 2. In the first convolution layer, a convolution kernel is used to perform a convolution operation on the input data. The convolution operation can extract local features of the input data; 3. Activation function: After the convolution operation, the convolution result is nonlinearly mapped by applying an activation function. Commonly used activation functions include ReLU (Rectified Linear Regression Unit (RLU)). Unit) function, etc., which can increase the nonlinear expression ability of the network; 4. Pooling operation: After the convolution layer, pooling operation is usually performed, and maximum pooling or average pooling can be selected. The pooling operation can reduce the spatial size of the feature map while retaining the main feature information, improving computational efficiency and reducing the risk of overfitting; 5. Repeat steps 2-4 to gradually extract more advanced features through multiple layers of convolutional layers. The number and size of convolution kernels in each convolutional layer can be adjusted according to the specific task and data characteristics.

[0055] Then, the aspect grayscale feature map and the slope grayscale feature map are cascaded to obtain a geoscience information feature map. Aspect and slope are important indicators of terrain characteristics, and have an important impact on plant growth environment and carbon storage. The aspect describes the direction of the surface, and can reflect factors such as solar radiation and water distribution to which plants are subject. The slope describes the steepness of the surface, and can reflect factors such as soil erosion and water retention to which plants are subject. Therefore, cascading the aspect grayscale feature map and the slope grayscale feature map can comprehensively utilize the information of these two indicators to obtain more comprehensive geoscience features. Through the cascading operation, the aspect grayscale feature map and the slope grayscale feature map can be connected in the channel dimension to form a higher-dimensional geoscience information feature map. Such a feature map can contain information on both the aspect and the slope, thereby more comprehensively describing the terrain characteristics.

[0056] In an embodiment of the present application, a possible implementation method of cascading the aspect grayscale feature map and the slope grayscale feature map to obtain a geoscientific information feature map may be: fusing the aspect grayscale feature map and the slope grayscale feature map using the following cascade formula to obtain the geoscientific information feature map; wherein the cascade formula is:

[0057] F = Concat[F a ,F b ]

[0058] Among them, F a represents the slope grayscale feature map, F brepresents the slope grayscale feature map, F represents the geological information feature map, and Concat[·,·] represents a cascade function.

[0059] The remote sensing factor feature map is a feature image learned and extracted from remote sensing data by a convolutional neural network model, which reflects the vegetation characteristics in the remote sensing data; while the geoscience information feature map is a feature image obtained by processing and analyzing the DEM image, which reflects the impact of geoscience factors on vegetation growth. Remote sensing data and geoscience data provide different information respectively. In order to integrate the information of remote sensing data and geoscience data, the remote sensing factor feature map and the geoscience information feature map are fused in the technical solution of the present application to obtain the biomass information feature map.

[0060] In particular, in the technical solution of the present application, it is considered that the remote sensing factor feature map and the geo-information feature map are obtained through different processing steps and methods. The remote sensing factor feature map is obtained by extracting vegetation indices, principal component analysis band linear transformation values ​​and tasseled cap transformation band linear transformation values ​​of different bands, and arranging them in two dimensions. The geo-information feature map is obtained by calculating the slope grayscale map and the slope grayscale map of the DEM image, and cascading through multiple convolution layers. Since the remote sensing factor feature map and the geo-information feature map are obtained through different processing methods, there may be certain differences in their feature distribution. The remote sensing factor feature map mainly reflects information such as vegetation index and band linear transformation values ​​in remote sensing data, while the geo-information feature map mainly focuses on the slope aspect and slope information of the terrain. Therefore, there is a certain degree of alignment offset in the feature distribution between the remote sensing factor feature map and the geo-information feature map. This means that their features may be misplaced or offset in space, that is, the features at the same position may appear inconsistent in position in different feature maps. Due to the alignment offset of feature distribution, directly weighted fusion of the remote sensing factor feature map and the geoscience information feature map may lead to confusion or loss of information. Based on this, in the technical solution of the present application, the posterior latent feature attention of the matching nodes between the remote sensing factor feature map and the geoscience information feature map is calculated to obtain the biomass information feature map, so as to eliminate or reduce the alignment offset between the feature maps, thereby improving the representation ability and classification performance of the feature maps.

[0061] Specifically, the posterior hidden feature attention of the matching nodes between the remote sensing factor feature map and the geoscience information feature map is calculated by the following formula to obtain the biomass information feature map; wherein the formula is:

[0062]

[0063] Among them, M 1i represents the characteristic matrix of the ith channel of the remote sensing factor characteristic map, M 2irepresents the feature matrix of the ith channel of the geoscience information feature map, T represents the transpose of the vector, C represents the number of channels of the feature map, ⊙ represents the positional dot multiplication of the feature matrix, Represents matrix multiplication, M ci The feature matrix representing the i-th channel of the biomass information feature map.

[0064] It should be understood that since there is an alignment offset in the feature distribution between the remote sensing factor feature map and the geoscience information feature map, that is, the feature distribution and feature shape of the two feature maps in the high-dimensional feature space are inconsistent, this leads to difficulties in information matching and alignment between the feature maps, and it is impossible to effectively perform feature fusion and classification. In response to this technical problem, this paper proposes a method based on the posterior latent feature attention association of matching nodes. This method can calculate the posterior latent feature attention association of matching nodes between the remote sensing factor feature map and the geoscience information feature map to obtain a biomass information feature map, that is, through a method based on an attention mechanism, each node in each feature map can obtain a posterior latent feature weight according to its correlation or dependency with each node in another feature map, thereby achieving information enhancement and dimensionality reduction of the feature map. Specifically, the method uses a relative position-based method to enable each node in each feature map to obtain a relative position embedding according to its relative position in the high-dimensional space, thereby achieving information enhancement and dimensionality reduction of the feature map. At the same time, through a feature modulation-based method, each node in each feature map can obtain a feature modulation according to its posterior latent feature attention with the matching node in another feature map, thereby realizing information enhancement and dimensionality reduction of the feature map. In this way, it is ensured that the class probability distribution of the biomass information feature map is consistent with the class probability distribution of the associated remote sensing factor feature map and the geoscientific information feature map to a certain extent, so as to effectively eliminate or reduce the alignment offset between feature maps, thereby improving the representation ability and classification performance of the feature maps.

[0065] Finally, the biomass information characteristic graph is passed through a decoder to obtain a decoded value, and the decoded value is used to represent the plant carbon storage value of the certain area. The decoder decodes the biomass information characteristic graph into the plant carbon storage value of a certain area, so that the abstract characteristic information can be converted into a specific plant carbon storage value. In this way, a quantitative estimation of plant carbon storage is achieved, which can provide important information on the carbon cycle of plant ecosystems.

[0066] In an embodiment of the present application, the biomass information characteristic map is passed through a decoder to obtain a decoded value, and the decoded value is used to represent the plant carbon storage value of the certain area. One possible implementation method may be: using the decoder to decode and regress the biomass information characteristic map using the following decoding formula to obtain the decoded value; wherein the decoding formula is:

[0067]

[0068] Wherein, X is the biomass information characteristic graph, Y is the decoded value, W is the weight matrix, Represents matrix multiplication.

[0069] In summary, a method for estimating plant carbon stocks based on satellite remote sensing and Internet of Things technology according to an embodiment of the present application is explained, which first obtains TM remote sensing images and DEM images of a certain area, then performs feature extraction on the acquired TM remote sensing images and DEM images of a certain area to obtain remote sensing factor features and geological information features, respectively, and then integrates the above features to obtain biomass-related features, and finally decodes the obtained biomass-related features through a decoder to generate the plant carbon stock value of the area.

[0070] Figure 6 : is a system block diagram of a plant carbon storage estimation system based on satellite remote sensing and Internet of Things technology according to an embodiment of the present application. Figure 6 As shown, according to the plant carbon stock estimation system 100 based on satellite remote sensing and Internet of Things technology according to the embodiment of the present application, it includes: an estimation-related data acquisition module 110, which is used to acquire TM remote sensing images and DEM images of a certain area; an estimation-related data encoding module 120, which is used to perform biomass feature encoding on the TM remote sensing images and DEM images of the certain area to obtain a biomass information feature map; a carbon stock estimation result generation module 130, which is used to obtain the plant carbon stock value of the certain area based on the biomass information feature map.

[0071] Here, those skilled in the art will appreciate that the specific functions and operations of the various units and modules in the plant carbon storage estimation system 100 based on satellite remote sensing and Internet of Things technology have been described in detail above. Figures 1 to 5 It has been introduced in detail in the description of the plant carbon stock estimation method based on satellite remote sensing and Internet of Things technology, and therefore, its repeated description will be omitted.

[0072] In summary, a plant carbon stock estimation system 100 based on satellite remote sensing and Internet of Things technology according to an embodiment of the present application is explained, which first acquires TM remote sensing images and DEM images of a certain area, then performs feature extraction on the acquired TM remote sensing images and DEM images of a certain area to obtain remote sensing factor features and geological information features, respectively, and then fuses the above features to obtain biomass-related features, and finally decodes the obtained biomass-related features through a decoder to generate the plant carbon stock value of the area.

[0073] As described above, the plant carbon stock estimation system 100 based on satellite remote sensing and Internet of Things technology according to the embodiment of the present application can be implemented in various wireless terminals, such as a server for plant carbon stock estimation based on satellite remote sensing and Internet of Things technology. In one example, the plant carbon stock estimation system 100 based on satellite remote sensing and Internet of Things technology according to the embodiment of the present application can be integrated into the wireless terminal as a software module and / or hardware module. For example, the plant carbon stock estimation system 100 based on satellite remote sensing and Internet of Things technology can be a software module in the operating system of the wireless terminal, or can be an application developed for the wireless terminal; of course, the plant carbon stock estimation system 100 based on satellite remote sensing and Internet of Things technology can also be one of the many hardware modules of the wireless terminal.

[0074] Alternatively, in another example, the plant carbon stock estimation system 100 based on satellite remote sensing and Internet of Things technology and the wireless terminal may also be separate devices, and the plant carbon stock estimation system 100 based on satellite remote sensing and Internet of Things technology may be connected to the wireless terminal via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0075] In the several embodiments provided in this application, it should be understood that the disclosed method, system or device can be implemented in other ways. For example, the system embodiment described above is only illustrative, for example, the division of the modules is only a logical function division, and there may be other division methods in actual implementation.

[0076] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical units, that is, they may be located in one place or 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.

[0077] In addition, each functional module in each embodiment of the present application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of hardware plus software functional modules.

[0078] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and that the present application can be implemented in other specific forms without departing from the spirit or essential features of the present application.

[0079] Therefore, no matter from which point of view, the embodiments should be regarded as illustrative and non-restrictive, and the scope of the present application is limited by the appended claims rather than the above description, so it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present application. Any attached figure mark in the claims should not be regarded as limiting the claims involved.

[0080] In addition, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the system claim can also be implemented by one unit or device through software or hardware. The second and other words are used to indicate names, but not to indicate any particular order.

[0081] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present application and are not intended to limit it. Although the present application has been described in detail with reference to the preferred embodiments, a person of ordinary skill in the art should understand that the technical solution of the present application may be modified or replaced by equivalents without departing from the spirit of the technical solution of the present application.

Claims

1. A method for estimating plant carbon storage based on satellite remote sensing and Internet of Things technology, characterized in that: include: Obtain TM remote sensing images and DEM images of a certain area; Performing biomass feature coding on the TM remote sensing image and the DEM image of the certain area to obtain a biomass information feature map; Based on the biomass information characteristic map, obtaining the plant carbon storage value of the certain area; The biomass feature coding is performed on the TM remote sensing image and the DEM image of the certain area to obtain a biomass information feature map, including: Extracting remote sensing factor features from the TM remote sensing image of the certain area to obtain a remote sensing factor feature map; Extracting geo-factor features from the DEM image of the certain area to obtain a geo-information feature map; Fusion of the remote sensing factor characteristic map and the geoscience information characteristic map to obtain the biomass information characteristic map; Wherein, fusing the remote sensing factor feature map and the geoscience information feature map to obtain the biomass information feature map comprises: calculating the posterior latent feature attention of the matching nodes between the remote sensing factor feature map and the geoscience information feature map by the following formula to obtain the biomass information feature map; Wherein, the formula is: in, The remote sensing factor characteristic map is represented by The feature matrix of channels, The first The feature matrix of channels, represents the transpose of a vector, Indicates the number of channels of the feature map, represents the positional dot product of the feature matrix, represents matrix multiplication, The biomass information characteristic graph is represented by The feature matrix of each channel.

2. The method for estimating plant carbon storage based on satellite remote sensing and Internet of Things technology according to claim 1, characterized in that: Extracting remote sensing factor features from the TM remote sensing image of the certain area to obtain a remote sensing factor feature map includes: Performing cloud removal processing on the TM remote sensing image of the certain area to obtain a cloud-removed TM remote sensing image; Extracting the band reflectance of different bands from the de-clouded TM remote sensing image; Performing different operations on the band reflectances of the different bands to obtain a plurality of vegetation indices, a plurality of principal component analysis band linear transformation values, and a plurality of tasseled cap transformation band linear transformation values; Arranging the plurality of vegetation indices, the plurality of principal component analysis band linear transformation values, and the plurality of tasseled cap transformation band linear transformation values ​​in two dimensions to obtain a remote sensing factor data matrix; The remote sensing factor data matrix is ​​passed through a convolutional neural network model as a feature extractor to obtain the remote sensing factor feature map.

3. The method for estimating plant carbon storage based on satellite remote sensing and Internet of Things technology according to claim 2, characterized in that: The remote sensing factor data matrix is ​​passed through a convolutional neural network model as a feature extractor to obtain the remote sensing factor feature map, including: Each layer of the convolutional neural network model as a feature extractor performs the following operations on the input data during the forward pass of the layer: Performing convolution processing on the input data to obtain a convolution feature map; Performing mean pooling processing based on a local feature matrix on the convolution feature map to obtain a pooled feature map; Performing nonlinear activation on the pooled feature map to obtain an activated feature map; Among them, the output of the last layer of the convolutional neural network model is the remote sensing factor feature map, and the input of the first layer of the convolutional neural network model is the remote sensing factor data matrix.

4. The method for estimating plant carbon storage based on satellite remote sensing and Internet of Things technology according to claim 3 is characterized in that: Extracting geo-factor features from the DEM image of the region to obtain a geo-information feature map includes: Calculating the DEM image of the certain area to obtain a slope aspect grayscale map and a slope grayscale map respectively; Passing the slope aspect grayscale map and the slope grayscale map through multiple convolution layers to obtain a slope aspect grayscale feature map and a slope grayscale feature map respectively; The aspect grayscale feature map and the slope grayscale feature map are cascaded to obtain the geoscience information feature map.

5. The method for estimating plant carbon storage based on satellite remote sensing and Internet of Things technology according to claim 4, characterized in that: Cascading the aspect grayscale feature map and the slope grayscale feature map to obtain the geoscience information feature map, comprising: fusing the aspect grayscale feature map and the slope grayscale feature map using the following cascade formula to obtain the geoscience information feature map; Wherein, the cascade formula is: in, represents the slope grayscale feature map, represents the slope grayscale feature map, represents the characteristic map of the geoscience information, Represents a cascade function.

6. The method for estimating plant carbon storage based on satellite remote sensing and Internet of Things technology according to claim 5, characterized in that: Based on the biomass information characteristic map, the plant carbon storage value of the certain area is obtained, including: passing the biomass information characteristic map through a decoder to obtain a decoded value, and the decoded value is used to represent the plant carbon storage value of the certain area.

7. A plant carbon storage estimation system based on satellite remote sensing and Internet of Things technology, characterized in that: include: The estimation related data acquisition module is used to obtain TM remote sensing images and DEM images of a certain area; An estimation-related data encoding module is used to perform biomass feature encoding on the TM remote sensing image and the DEM image of the certain area to obtain a biomass information feature map; A carbon storage estimation result generation module is used to obtain the plant carbon storage value of the certain area based on the biomass information characteristic map; The biomass feature coding is performed on the TM remote sensing image and the DEM image of the certain area to obtain a biomass information feature map, including: Extracting remote sensing factor features from the TM remote sensing image of the certain area to obtain a remote sensing factor feature map; Extracting geo-factor features from the DEM image of the certain area to obtain a geo-information feature map; Fusion of the remote sensing factor characteristic map and the geoscience information characteristic map to obtain the biomass information characteristic map; Wherein, fusing the remote sensing factor feature map and the geoscience information feature map to obtain the biomass information feature map comprises: calculating the posterior latent feature attention of the matching nodes between the remote sensing factor feature map and the geoscience information feature map by the following formula to obtain the biomass information feature map; Wherein, the formula is: Wherein, the formula is: in, The remote sensing factor characteristic map is represented by The feature matrix of channels, The first The feature matrix of channels, represents the transpose of a vector, Indicates the number of channels of the feature map, represents the positional dot product of the feature matrix, represents matrix multiplication, The biomass information characteristic graph is represented by The feature matrix of each channel.

8. The plant carbon storage estimation system based on satellite remote sensing and Internet of Things technology according to claim 7 is characterized in that: The carbon storage estimation result generating module is used to: pass the biomass information characteristic map through a decoder to obtain a decoded value, and the decoded value is used to represent the plant carbon storage value of the certain area.

Images