A forest resource detection method based on satellite image data

Abstract

The application relates to the technical field of forest resource detection, and discloses a forest resource detection method based on satellite image data, which comprises the following steps: acquiring initial satellite image data of a target forest area, pre-processing the initial satellite image data to obtain first satellite image data; performing radiation atmosphere correction on the first satellite image data through a radiation atmosphere correction model to obtain second satellite image data; performing super-resolution reconstruction on the second satellite image data according to a reconstruction enhancement model to obtain third satellite image data; identifying the third satellite image data based on a forest ground object extraction model to obtain forest ground object information; and identifying the third satellite image data and the forest ground object information through a forest resource detection model to obtain a forest resource distribution map of the target forest area. The application improves the precision and accuracy of forest resource detection, enhances the resolution of satellite image data, and realizes accurate identification of forest ground object information through the forest ground object extraction model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of forest resource detection, and in particular to a forest resource detection method based on satellite image data. BACKGROUND

[0002] Forests are the main body of terrestrial ecosystems, playing a leading and supporting role. Protecting forest resources is to create a foundation for ecological civilization. Studying the dynamic changes of forest resources helps to understand the quantity, quality, and structural change mechanism of forest resources, and through scientific and technological measures and reasonable control policies, the structure can be made more reasonable. As the largest ecosystem, forests can regulate the climate, conserve water, prevent soil erosion, prevent wind and sand, prevent drought, prevent waterlogging, reduce noise, and beautify the environment, playing a crucial role in maintaining the biosphere ecosystem.

[0003] Traditional forest resource survey methods usually rely on ground surveys, which have the disadvantages of long time, high cost, limited coverage, and are difficult to meet the needs of rapid changes in forest resource monitoring. With the advancement of remote sensing technology, using satellite images for forest resource monitoring has become an effective means. Remote sensing technology can quickly obtain surface information in a large area, with the advantages of high efficiency and accuracy. Although there are some forest resource detection methods based on remote sensing, there are still problems of insufficient processing of atmospheric influence and insufficient image resolution, which affect the accuracy and reliability of the detection results.

[0004] Existing forest resource detection methods based on remote sensing ignore the influence of water vapor and aerosols in the atmosphere, causing the satellite image to appear hazy and unclear, reducing the contrast and clarity of the image. SUMMARY

[0005] Therefore, the present application provides a forest resource detection method based on satellite image data, which eliminates the radiation and atmospheric effects of satellite image data, enhances the resolution of satellite image data, effectively identifies the target forest area resources and generates a forest resource distribution map, realizes accurate identification of forest land information, improves the accuracy and accuracy of forest resource detection, and solves the problem of hazy and unclear satellite images.

[0006] The technical scheme of the present application is as follows: The present application provides a forest resource detection method based on satellite image data, comprising the following steps:

[0007] S1, obtaining initial satellite image data of a target forest area, preprocessing the initial satellite image data to obtain first satellite image data;

[0008] S2, performing radiation atmosphere correction on the first satellite image data by a radiation atmosphere correction model to obtain second satellite image data;

[0009] S3, performing super-resolution reconstruction on the second satellite image data according to a reconstruction enhancement model to obtain third satellite image data;

[0010] S4, identifying the third satellite image data based on a forest object extraction model to obtain forest object information;

[0011] S5, identifying the third satellite image data and the forest object information by a forest resource detection model to obtain a forest resource distribution map of the target forest region.

[0012] In the above technical solution, preferably, step S2 comprises:

[0013] The radiation atmosphere correction model comprises radiation correction and atmosphere correction;

[0014] The first satellite image data is subjected to radiation correction:

[0015]

[0016] wherein, is the first satellite image data after radiation correction, is the average solar intensity, g min is the minimum solar intensity, g max is the maximum solar intensity, is the maximum solar radiation value, is the minimum solar radiation value;

[0017] The first satellite image data after radiation correction is subjected to atmosphere correction:

[0018]

[0019] wherein, is the second satellite image data, is the first satellite image data after radiation correction, d dr is the distance between the earth and the sun, is the solar radiation flux, a s is the solar zenith angle, is the atmosphere adjustment coefficient, t sq is the water vapor factorization factor, t qr is the aerosol factorization factor, A dx is the terrain correction coefficient.

[0020] In the above technical solution, preferably, step S3 comprises:

[0021] decomposing the second satellite image data into a plurality of low-resolution frames, and extracting a feature vector of each low-resolution frame;

[0022] aligning the feature vectors of different times of the low-resolution frames through deformable convolution;

[0023] extracting a scale adaptive feature vector of the low-resolution frame based on a plurality of scale factors;

[0024] performing super-resolution reconstruction on the low-resolution frame according to the plurality of scale factors and the scale adaptive feature vector to obtain third satellite image data.

[0025] In the above technical solution, preferably, the calculation formula of the third satellite image data is:

[0026]

[0027] wherein, is a low-resolution frame at time t, is is a corresponding feature vector, f FE (·) is a feature extraction function, is is a feature vector obtained after time alignment, is a feature vector corresponding to a low-resolution frame at time i, w i is a weight at time i, p is an index position coordinate, and Δp i is an index position coordinate offset at time i, and N is the number of support frames on the left and right sides at time t, is a scale adaptive feature vector at scale n, f SFEM (·) is a scale adaptive extraction function, r1 n and r2 n are scaling factors in length and width at scale n, respectively, is is a corresponding low-resolution frame, is third satellite image data, Conv 1×1 and Conv 3×3 are 1x1 convolution operation and 3x3 convolution operation, respectively, and ReLU(·) is a ReLU activation function, f GFCUM (·) is a global implicit function.

[0028] In the above technical solution, preferably, step S4 comprises:

[0029] S41, obtaining historical forest ground object information data, and dividing the historical forest ground object information data into a training set and a test set;

[0030] S42, an initial forest feature extraction model is constructed based on a U-NET model, the initial forest feature extraction model is iteratively trained by a training set, a loss function and hyperparameters are configured, until an iteration stopping condition is reached, a trained initial forest feature extraction model is obtained, the iteration stopping condition includes reaching a preset iteration number and loss function convergence;

[0031] S43, the trained initial forest feature extraction model is evaluated by a test set, the model is evaluated based on an mAP value, when the model evaluation meets the requirements, a forest feature extraction model is obtained, if the model evaluation does not meet the requirements, the model is iteratively trained after adjusting the hyperparameters;

[0032] S44, the third satellite image data is identified based on the forest feature extraction model, and forest feature information of a target forest area is obtained.

[0033] Based on the above technical solutions, preferably, the calculation formula of the forest feature extraction model is:

[0034]

[0035] wherein, is the forest feature information of the target forest area, is the score of the kth forest feature, K is the number of forest feature types, W k is the weight of the kth forest feature, b k is the bias term of the kth forest feature.

[0036] Based on the above technical solutions, preferably, the calculation formula of the loss function is:

[0037]

[0038] wherein, Z is the model input, E(Z) is the loss function value of Z, Ω is a set of pixel points of Z, Y is a pixel point in Ω, w(Y) is the weight of Y, and p(Y) is the predicted probability distribution of the model input.

[0039] Based on the above technical solutions, preferably, step S5 includes:

[0040] The forest resource detection model includes a first regression model and a second regression model;

[0041] The first regression model is used to identify the third satellite image data, divide the target forest area into a plurality of target forest sub-areas of the same size, and obtain forest resource type division results of the plurality of target forest sub-areas;

[0042] The second regression model is used for identifying forest feature information, and a forest resource type area content of a plurality of target forest sub-areas is obtained.

[0043] Based on the above technical scheme, preferably, the calculation formula of the first regression model is:

[0044]

[0045] The calculation formula of the second regression model is:

[0046]

[0047] wherein, is a forest resource type prediction value of the target forest sub-area, m is the number of decision trees of the first regression model, is a forest resource type prediction value of the target forest sub-area of the jth decision tree of the first regression model, P is a forest resource type division result of the target forest sub-area, P1, P2, P v are the target forest sub-areas of the forest resource type 1, the forest resource type 2 and the forest resource type v respectively, Y1, Y2, Y v-1 , Y v are the 1st, 2nd, (v-1)th and vth forest resource type thresholds respectively, f(X u ) is a prediction value of the uth forest resource of the target forest sub-area, alpha d is a weight of the dth decision tree of the second regression model, h d (X u ) is a prediction value of the uth forest resource of the target forest sub-area of the dth decision tree of the second regression model, D is the number of decision trees of the second regression model, L(theta) is a loss function of the second regression model, y u is an actual value of the uth forest resource of the target forest sub-area, Q is the number of forest resource types of the target forest sub-area, and Omega(theta) is a regularization term.

[0048] Based on the above technical scheme, preferably, the step S5 further comprises:

[0049] A scaled drawing of the target forest area is drawn, the scaled drawing is marked based on the forest resource type division results of the plurality of target forest sub-areas of the target forest area and the prediction values of the forest resources of various types, and a forest resource distribution map of the target forest area is obtained.

[0050] The forest resource detection method based on satellite image data has the following beneficial effects compared with the prior art:

[0051] (1) By adopting data preprocessing, radiation atmosphere correction, super-resolution reconstruction, forest feature information extraction and forest resource detection, the radiation influence and atmospheric influence of satellite image data are eliminated, the resolution of satellite image data is enhanced, the target forest area resources are effectively identified and the forest resource distribution map is generated, the accurate identification of forest feature information is realized, and the precision and accuracy of forest resource detection are improved; (2) By deformable convolution and scale adaptive feature vector extraction, forest feature characteristics at different scales can be effectively captured, so that high-quality super-resolution reconstruction is realized, and the improvement of the reconstructed satellite image data in detail and accuracy is ensured, providing a more reliable data basis for forest resource detection; (3) By combining the first regression model and the second regression model, the forest resources are accurately classified and the regional content ratio is evaluated, and the regression model based on decision tree is used to process the complex relationship between forest resources and satellite image data, so that the classification accuracy and the generalization ability of the model are improved. BRIEF DESCRIPTION OF DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0053] Figure 1 A flow chart of a forest resource detection method based on satellite image data according to the present application;

[0054] Figure 2 A structure diagram of the reconstruction enhancement model according to the present application. DETAILED DESCRIPTION

[0055] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0056] Please refer to Figure 1 The present embodiment provides a forest resource detection method based on satellite image data, comprising the following steps:

[0057] S1, obtaining initial satellite image data of a target forest area, preprocessing the initial satellite image data to obtain first satellite image data;

[0058] S2, performing radiative atmospheric correction on the first satellite image data by a radiative atmospheric correction model to obtain second satellite image data;

[0059] S3, performing super-resolution reconstruction on the second satellite image data according to a reconstruction enhancement model to obtain third satellite image data;

[0060] S4, identifying the third satellite image data based on a forest feature extraction model to obtain forest feature information;

[0061] S5, identifying the third satellite image data and the forest feature information by a forest resource detection model to obtain a forest resource distribution map of the target forest area.

[0062] Specifically, the forest resource detection method based on satellite image data of the embodiment eliminates the radiation influence and atmospheric influence of the satellite image data, enhances the resolution of the satellite image data, effectively identifies the resources of the target forest area and generates a forest resource distribution map, realizes accurate identification of forest feature information, and improves the precision and accuracy of forest resource detection.

[0063] The preprocessing includes standardization, denoising and filtering processing on the initial satellite image data;

[0064] The calculation formula of the standardization is:

[0065]

[0066] wherein z is the standardized initial satellite image data, is the value of the initial satellite image data, is the mean value of the value of the initial satellite image data, and sigma is the standard deviation of all data values of the initial satellite image data corresponding to the data source.

[0067] Specifically, the standardization processing of the embodiment can eliminate the dimensional difference and numerical distribution difference in the initial satellite image data, so that the data of different data sources or channels are comparable, which is conducive to subsequent data fusion and analysis processing. This preprocessing method can improve the precision and stability of the steps of radiative atmospheric correction and super-resolution reconstruction, and lays a good data foundation for forest feature identification and resource detection.

[0068] Step S2 includes:

[0069] The radiative atmospheric correction model includes radiation correction and atmospheric correction;

[0070] The radiation correction on the first satellite image data includes:

[0071]

[0072] wherein, is the first satellite image data after radiation correction, is the average solar radiation, g min is the minimum solar radiation, g max is the maximum solar radiation, is the maximum solar radiation value, is the minimum solar radiation value;

[0073] Atmospheric correction is performed on the first satellite image data after radiation correction:

[0074]

[0075] wherein, is the second satellite image data, is the first satellite image data after radiation correction, d dr is the distance from the earth to the sun, is the solar radiation flux, a s is the solar zenith angle, is the atmospheric adjustment coefficient, t sq is the water vapor factor, t qr is the aerosol factor, A dx is the terrain correction coefficient.

[0076] Specifically, by radiation correction, the image data error caused by the change of solar intensity is eliminated, so that the image data after radiation correction more truly reflects the actual situation of the target forest area.

[0077] Based on the atmospheric correction step, the interference of atmospheric components (such as water vapor and aerosol) on satellite image data is removed, ensuring the clarity and accuracy of the image data.

[0078] After radiation and atmospheric correction, the image data has higher comparability under different time and conditions, which is convenient for time series analysis and change detection, and supports dynamic monitoring of forest resources.

[0079] Through radiation correction and atmospheric correction, the second satellite image data generated provides high-quality basic data for super-resolution reconstruction and forest feature extraction, improving the accuracy of forest resource detection.

[0080] Step S3 includes:

[0081] The second satellite image data is decomposed into a plurality of low-resolution frames, and the feature vector of each low-resolution frame is extracted;

[0082] The feature vectors of different times of the low-resolution frames are aligned by deformable convolution.

[0083] extracting scale adaptive feature vectors of the low-resolution frames based on the multiple scale factors;

[0084] super-resolution reconstructing the low-resolution frames based on the multiple scale factors and the scale adaptive feature vectors to obtain the third satellite image data.

[0085] Specifically, the embodiment fully utilizes the time sequence information by decomposing the second satellite image data into multiple low-resolution frames and extracting feature vectors of each frame, thereby improving the temporal consistency of the reconstruction result.

[0086] The feature vectors of different times are aligned through deformable convolution, which overcomes the geometric distortion between low-resolution frames caused by factors such as shooting angle and time, and ensures the accurate alignment of feature information.

[0087] Extracting scale adaptive feature vectors of the low-resolution frames based on the multiple scale factors can better capture detailed information at different scales and improve the detail preservation ability of the reconstruction result.

[0088] The super-resolution reconstruction is comprehensively performed by utilizing the multiple scale factors and the scale adaptive feature vectors, which fully utilizes the advantages of time sequence information, geometric alignment and multi-scale features, and finally obtains high-quality third satellite image data, thereby enhancing the effect of forest feature recognition and resource detection.

[0089] The calculation formula of the third satellite image data is:

[0090]

[0091] wherein, is a low-resolution frame at time t, is a corresponding feature vector, f FE is a feature extraction function, is a feature vector obtained after time alignment, is a feature vector corresponding to a low-resolution frame at time i, w i is a weight at time i, p is an index position coordinate, and Δp i is an index position coordinate offset at time i, and N is the number of support frames on the left and right sides at time t, is a scale adaptive feature vector at scale n, f SFEM is a scale adaptive extraction function, r1 n and r2 n are scaling factors in length and width at scale n, respectively, is a corresponding low-resolution frame, For third satellite imagery data, Conv 1×1 and Conv 3×3 These represent 1×1 and 3×3 convolution operations, respectively, with ReLU(·) as the ReLU activation function, f GFCUM (·) represents a global implicit function.

[0092] Specifically, such as Figure 2 The diagram shown is a structural diagram of the reconstruction enhancement model in this embodiment. This embodiment utilizes time series information, extracts the feature vector of each low-resolution frame, and aligns it with time, making full use of the correlation between multiple frames of data.

[0093] By introducing scale-adaptive feature vectors, we can better capture detailed information at different scales, improve the detail preservation ability of the reconstruction results, and enhance the reconstruction effect of high-frequency information.

[0094] The super-resolution reconstruction layer uses 1×1 and 3×3 convolution operations, combined with ReLU activation function and global implicit function, which can make full use of multi-scale feature information. This embodiment fully considers time series information, scale adaptive features and multi-scale feature fusion, realizes effective super-resolution reconstruction of low-resolution frames, and obtains high-quality third satellite image data.

[0095] Step S4 includes:

[0096] S41, acquire historical forest land cover information data, and divide the historical forest land cover information data into a training set and a test set;

[0097] S42, Construct an initial forest feature extraction model based on the U-NET model, iteratively train the initial forest feature extraction model using the training set, configure the loss function and hyperparameters, until the iteration stopping condition is met, and obtain the trained initial forest feature extraction model. The iteration stopping condition includes reaching a preset number of iterations and the convergence of the loss function.

[0098] S43, evaluate the initial forest cover extraction model after training using the test set, evaluate the model based on the mAP value, and obtain the forest cover extraction model when the model evaluation meets the requirements; if the model evaluation does not meet the requirements, adjust the hyperparameters and retrain the model iteratively.

[0099] S44, Based on the forest feature extraction model, the third satellite image data is identified to obtain forest feature information of the target forest area.

[0100] Specifically, this embodiment utilizes historical forest land cover information data, dividing it into training and testing sets, providing a solid data foundation for model training and evaluation, and ensuring the model's generalization ability.

[0101] The embodiment adopts a U-NET model as an initial forest feature extraction model, the model has good semantic segmentation capability and can effectively identify various forest feature information in the third satellite image data. Through iterative training, a suitable loss function and hyperparameters are configured, and finally a forest feature extraction model with excellent performance is obtained.

[0102] The trained model is evaluated by the test set, and the mAP value is used as the evaluation index to objectively measure the recognition performance of the model. When the model evaluation meets the requirements, the final forest feature extraction model is obtained. If the evaluation does not meet the requirements, the hyperparameters can be adjusted to further optimize the model.

[0103] The calculation method of the mAP value is as follows:

[0104]

[0105] Wherein, b is the number of forest feature categories, R(c) is the current position recall rate, R(c+1) is the next position recall rate, and precision(c,c+1) is the accuracy of the current position and the next position;

[0106] The following parameters are defined,

[0107] TP: the number of correctly classified positive samples; actual positive samples, also classified as positive samples by the model;

[0108] FP: the number of incorrectly classified positive samples; actual negative samples, but classified as positive samples by the model;

[0109] TN: the number of correctly classified negative samples; actual negative samples, also classified as negative samples by the model;

[0110] FN: the number of incorrectly classified negative samples; actual positive samples, but classified as negative samples by the model;

[0111] Accuracy Precision=TP / (TP+FP), recall Recall=TP / (TP+FN);

[0112] Taking Precision and Recall of each type as the horizontal and vertical axes respectively, the AP value is actually equivalent to the area surrounded by Precision and Recall. Since the curve area is very jittery, an interpolation method is used to smooth the curve area. If the current recall value is R(c), the interpolation is the maximum Precision between the next position recall R(c+1). Wherein c represents the current position. The mAP value is the average of all types of AP values, and b represents the number of forest feature categories.

[0113] Based on the optimized forest feature extraction model, the third satellite image data can be accurately identified to obtain various forest feature information of the target forest area, thereby providing data support for forest resource detection.

[0114] The calculation formula of the forest feature extraction model is:

[0115]

[0116] wherein, is the forest feature information of the target forest area, is the score of the kth forest feature, K is the number of forest feature types, W k is the weight of the kth forest feature, b k is the bias term of the kth forest feature.

[0117] The calculation formula of the loss function is:

[0118]

[0119] wherein, Z is the model input, E(Z) is the loss function value of Z, Ω is the pixel point set of Z, Y is a pixel point in Ω, w(Y) is the weight of Y, and p(Y) is the predicted probability distribution of the model input.

[0120] Specifically, the forest feature extraction model of the embodiment adopts a score-based method for feature recognition, wherein the score of each category is determined by the weight and the bias term, which can effectively capture the feature differences of different categories of forest features and improve the recognition accuracy.

[0121] The design of the loss function considers the pixel point set of the model input Z, as well as the weight and predicted probability distribution of each pixel point Y. This pixel-level loss function can better optimize the recognition ability of the model in detail features.

[0122] The forest feature extraction model obtained by training and optimization can accurately identify various forest features in the third satellite image data, thereby providing reliable feature information data for forest resource detection.

[0123] The forest feature extraction model fully considers the feature differences of forest features, adopts a score-based recognition method, and combines pixel-level loss function optimization to improve the accuracy of forest feature recognition through an improved machine learning model.

[0124] Step S5 includes:

[0125] The forest resource detection model includes a first regression model and a second regression model.

[0126] The first regression model is used for identifying the third satellite image data, and dividing the target forest region into a plurality of target forest sub-regions of the same size, to obtain a forest resource type division result of the plurality of target forest sub-regions.

[0127] The calculation formula of the first regression model is:

[0128]

[0129] The second regression model is used for identifying the forest object information, to obtain a forest resource type area content of the plurality of target forest sub-regions.

[0130] The calculation formula of the second regression model is:

[0131]

[0132] wherein, is a forest resource type prediction value of the target forest sub-region, and m is a decision tree number of the first regression model, is a forest resource type prediction value of the target forest sub-region by the jth decision tree of the first regression model, P is a forest resource type division result of the target forest sub-region, P1, P2, P v are target forest sub-regions of the forest resource type 1, the forest resource type 2, and the forest resource type v, respectively, Y1, Y2, Y v-1 , Y v are the 1st, 2nd, (v-1)th, and vth forest resource type thresholds, respectively, f(X u ) is a prediction value of the u-th forest resource of the target forest sub-region, and a d is a weight of the dth decision tree of the second regression model, h d (X u ) is a prediction value of the u-th forest resource of the target forest sub-region by the dth decision tree of the second regression model, D is a decision tree number of the second regression model, L(θ) is a loss function of the second regression model, y u is an actual value of the u-th forest resource of the target forest sub-region, Q is a forest resource type number of the target forest sub-region, and Ω(θ) is a regularization term.

[0133] Specifically, the first regression model of the embodiment is used for identifying the third satellite image data, and dividing the target forest region into a plurality of target forest sub-regions of the same size, to obtain a forest resource type division of each sub-region. This sub-region-based analysis method can better reflect the spatial distribution characteristics of forest resources.

[0134] The second regression model is used to identify the forest object information, and the content proportion of each type of forest resource in each target forest sub-region is obtained. This analysis based on the object information can more accurately evaluate the distribution of different forest resources.

[0135] The calculation formula of the two regression models adopts the form of a decision tree, which can effectively capture the complex nonlinear relationship between forest resources and satellite image data and object information, and improve the accuracy of the detection results.

[0136] The second regression model also introduces a loss function and a regularization term, which can optimize the model parameters during the training process, improve the generalization ability of the model, and enhance the predictability of the actual forest resource distribution.

[0137] Step S5 further comprises:

[0138] A scaled drawing of the target forest region is drawn, and the scaled drawing is labeled based on the forest resource type classification results of the multiple target forest sub-regions of the target forest region and the predicted values of each type of forest resource, to obtain a forest resource distribution map of the target forest region.

[0139] Specifically, in a specific embodiment, first, a scaled drawing of the target forest region is drawn, the geographical boundary information of the target forest region is obtained, and the scaled drawing is drawn according to the actual area of the target forest region;

[0140] Secondly, based on the forest resource type classification results of the multiple target forest sub-regions of the target forest region, the forest resource type classification results of each target forest sub-region obtained by the first regression model in step S5 are labeled on the scaled drawing, and different colors or legends are used to identify different types of forest resource types;

[0141] Thirdly, based on the predicted values of each type of forest resource, the scaled drawing is labeled, and the predicted content proportion of each type of forest resource in each target forest sub-region obtained by the second regression model in step S5 is labeled on the scaled drawing in the form of area proportion, and different colors or legends are used to identify the distribution of different types of forest resources;

[0142] Finally, the forest resource distribution map of the target forest region is obtained, and the labeling results of the above two steps are integrated on the scaled drawing to form a forest resource distribution map that directly reflects the distribution of each type of forest resource in the target forest region.

[0143] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for detecting forest resources based on satellite imagery data, characterized in that, Includes the following steps: S1, acquire initial satellite image data of the target forest area, and preprocess the initial satellite image data to obtain first satellite image data; S2, Perform radiation atmospheric correction on the first satellite image data using a radiation atmospheric correction model to obtain the second satellite image data; Step S2 includes: The radiation atmospheric correction model includes radiation correction and atmospheric correction; Radiometric correction was performed on the first satellite image data: in, This is the first satellite image data after radiometric correction. For average solar intensity, g min For the lowest solar intensity, g max This represents the highest solar radiation intensity. This represents the highest solar radiation value. This is the lowest solar radiation value; Atmospheric correction was performed on the first satellite image data after radiometric correction: in, This is second satellite imagery data. For the first satellite image data after radiometric correction, d dr The distance between the Earth and the Sun. For solar radiation flux, α s The zenith angle of the sun. τ is the atmospheric adjustment coefficient. sq τ is the water vapor mimicry factor. qr As an aerosol mimicry factor, A dx This is the terrain correction factor; S3, perform super-resolution reconstruction on the second satellite image data according to the reconstruction enhancement model to obtain the third satellite image data; Step S3 includes: The second satellite image data is decomposed into multiple low-resolution frames, and the feature vector of each low-resolution frame is extracted. Align feature vectors at different times of low-resolution frames using deformable convolution; Scale-adaptive feature vectors of low-resolution frames are extracted based on multiple scale factors. Super-resolution reconstruction of low-resolution frames is performed based on multiple scale factors and scale-adaptive feature vectors to obtain third satellite image data. S4, Based on the forest feature extraction model, identify the third satellite image data to obtain forest feature information; S5. By using a forest resource detection model to identify third satellite image data and forest land cover information, a forest resource distribution map of the target forest area is obtained.

2. The forest resource detection method based on satellite imagery data as described in claim 1, characterized in that, The formula for calculating the third satellite image data is: in, Let be the low-resolution frame at time t. for The corresponding eigenvector, f FE (·) is the feature extraction function. for The feature vector obtained after time alignment Let w be the feature vector corresponding to the low-resolution frame at time i. i Let p be the weight at time i, p be the index coordinate, and Δp be the weight at time i. i Let be the index position coordinate offset at time i, and N be the number of supporting frames on the left and right sides at time t. f is the scale-adaptive eigenvector at scale n. SFEM (·) is the scale-adaptive extraction function, r1 n and These are scaling factors for length and width at scale n, respectively. for The corresponding low-resolution frame, For third satellite imagery data, Conv 1×1 and Conv 3×3 These represent 1×1 and 3×3 convolution operations, respectively, with ReLU(·) as the ReLU activation function, f GFCUM (·) represents a global implicit function.

3. The forest resource detection method based on satellite imagery data as described in claim 2, characterized in that, Step S4 includes: S41, acquire historical forest land cover information data, and divide the historical forest land cover information data into a training set and a test set; S42, Construct an initial forest feature extraction model based on the U-NET model, iteratively train the initial forest feature extraction model using the training set, configure the loss function and hyperparameters, until the iteration stopping condition is met, and obtain the trained initial forest feature extraction model. The iteration stopping condition includes reaching a preset number of iterations and the convergence of the loss function. S43, evaluate the initial forest cover extraction model after training using the test set, evaluate the model based on the mAP value, and obtain the forest cover extraction model when the model evaluation meets the requirements; if the model evaluation does not meet the requirements, adjust the hyperparameters and retrain the model iteratively. S44, Based on the forest feature extraction model, the third satellite image data is identified to obtain forest feature information of the target forest area.

4. The forest resource detection method based on satellite imagery data as described in claim 3, characterized in that, The calculation formula for the forest feature extraction model is as follows: in, It is information on forest features in the target forest area. W represents the extraction score for the k-th type of forest land cover, where K is the number of forest land cover species. k Let b be the weight of the k-th type of forest land cover. k This is the bias term for the k-th type of forest land cover.

5. The forest resource detection method based on satellite imagery data as described in claim 4, characterized in that, The formula for calculating the loss function is: Where Z is the model input, E(Z) is the loss function value of Z, Ω is the set of pixels in Z, Y is a pixel in Ω, w(Y) is the weight of Y, and p(Y) is the predicted probability distribution of the model input.

6. The forest resource detection method based on satellite imagery data as described in claim 5, characterized in that, Step S5 includes: The forest resource detection model includes a first regression model and a second regression model; The first regression model is used to identify the third satellite image data, divide the target forest area into multiple target forest sub-regions of the same size, and obtain the forest resource type classification results of multiple target forest sub-regions; The second regression model is used to identify forest land cover information and obtain the regional content of forest resource types in multiple target forest sub-regions.

7. The forest resource detection method based on satellite imagery data as described in claim 6, characterized in that, The calculation formula for the first regression model is: The second regression model is calculated as follows: in, Let be the predicted forest resource species for the target forest sub-region, and m be the number of decision trees in the first regression model. Let P be the predicted forest resource species of the target forest sub-region by the j-th decision tree in the first regression model, and let P be the forest resource species classification result of the target forest sub-region. P1, P2, P... v These are target forest subregions, Y1, Y2, and Y, representing forest resource type 1, forest resource type 2, and forest resource type v, respectively. v-1 Y v These are the threshold values ​​for the 1st, 2nd, v-1st, and vth forest resource types, respectively, f(X) u α is the predicted value of the u-th type of forest resource in the target forest sub-region. d h represents the weight of the d-th decision tree in the second regression model. d (X u Let y be the predicted value of the d-th decision tree in the second regression model for the u-th type of forest resource in the target forest sub-region, D be the number of decision trees in the second regression model, and L(θ) be the loss function of the second regression model. u Let θ be the actual value of the u-th type of forest resource in the target forest sub-region, Q be the number of forest resource types in the target forest sub-region, and Ω(θ) be the regularization term.

8. The forest resource detection method based on satellite imagery data as described in claim 7, characterized in that, Step S5 also includes: A scaled map of the target forest area is drawn. Based on the classification results of forest resource types in multiple target forest sub-regions of the target forest area and the predicted values ​​of various forest resources, the scaled map is labeled to obtain a forest resource distribution map of the target forest area.

Images