Forest stand dominant tree species remote sensing recognition system and method based on texture feature fractal dimension

Abstract

The invention relates to the technical field of forest tree species identification, and discloses a forest stand dominant tree species remote sensing identification system and method based on textural feature fractal dimensions, and the system comprises an acquisition module which is configured to acquire multispectral data of a forest research area by using a high-resolution remote sensing image; the data processing module is configured to preprocess the multispectral data and generate a remote sensing image of the research area. The feature processing module is configured to extract the image texture features of the remote sensing image of the research area based on the gray level co-occurrence matrix, and determine the fractal dimension of the remote sensing image according to the image texture features. The feature processing module is further configured to screen out preferred classification features according to the evaluation result of the feature combination. The modeling unit is configured to establish a classification model based on the preferred classification features. The identification module is configured to verify the classification model, and the verified classification model generates a dominant tree species classification chart. The forest stand dominant tree species can be accurately identified and classified through cooperation of all the modules.

Description

Technical Field

[0001] The present invention relates to the technical field of forest tree species identification, and in particular to a remote sensing identification system and method for forest stand dominant tree species based on texture feature fractal dimension. Background Art

[0002] With the continuous development of remote sensing technology, the use of remote sensing images for forest resource surveys and ecological monitoring has become an important research field. Traditional forest tree species identification methods mainly rely on manual surveys and ground sampling, but this method is time-consuming and labor-intensive and cannot cover large areas, and has certain limitations. As an efficient and wide-area method of obtaining ground information, remote sensing images have high spatial resolution and multispectral characteristics. They can obtain various ecological information of forest areas on a large scale and quickly, providing effective support for forest resource assessment and tree species classification.

[0003] At present, although remote sensing images have powerful information collection capabilities, their processing and analysis processes still face many challenges. The spectral characteristics of different tree species are highly similar, and are affected by environmental factors (such as seasonal changes, climatic conditions, etc.), which can easily lead to inaccurate tree species classification results. In addition, fractal dimension, as an important indicator to describe the complexity of image texture, has been widely used in remote sensing image analysis. By extracting the fractal dimension of the image, the spatial feature differences of different tree species in remote sensing images can be effectively quantified, thereby providing a more accurate basis for classification. Although some studies have attempted to use texture features and fractal dimensions for tree species identification, how to accurately extract and screen effective texture features in high-resolution remote sensing images, and how to establish a more efficient classification model, are still research difficulties in current technology.

[0004] Therefore, there is an urgent need to invent a remote sensing image recognition technology to solve the problem of poor accuracy in identifying tree species through remote sensing images in the prior art. Summary of the invention

[0005] In view of this, the present invention proposes a remote sensing identification system and method for dominant tree species in a forest stand based on texture feature fractal dimension, aiming to solve the problem of poor accuracy in tree species identification through remote sensing images in current technology, and at the same time improve the accuracy of tree species identification by accurately extracting effective texture features and fractal dimensions and establishing an efficient classification model.

[0006] The present invention proposes a remote sensing identification system for forest stand dominant tree species based on texture feature fractal dimension, comprising:

[0007] an acquisition module configured to acquire multispectral data of a forest study area using high-resolution remote sensing images;

[0008] A data processing module, electrically connected to the acquisition module, the data processing module is configured to pre-process the multispectral data and generate a remote sensing image of a research area;

[0009] The feature processing module is configured to extract the image texture features of the remote sensing image of the study area based on the gray level co-occurrence matrix, and determine the fractal dimension of the remote sensing image according to the image texture features; the feature processing module is also configured to screen out the preferred classification features according to the evaluation results of the feature combination;

[0010] a modeling unit, electrically connected to the feature processing module, wherein the modeling unit is configured to establish a classification model based on the preferred classification feature;

[0011] An identification module is electrically connected to the modeling unit, and the identification module is configured to verify the classification model, and the verified classification model generates a dominant tree species classification diagram.

[0012] Furthermore, when the feature processing module extracts the image texture features of the remote sensing image of the study area based on the gray level co-occurrence matrix, it includes:

[0013] The feature processing module is further configured to obtain the spatial relationship between the pixels in the remote sensing image after preprocessing, and generate a gray level co-occurrence matrix according to the gray level co-occurrence frequency of each pixel;

[0014] The feature processing module is further configured to obtain the probability distribution of occurrence of each gray level in the gray level co-occurrence matrix based on normalization processing;

[0015] The feature processing module is further configured to obtain the brightness mean of the image texture, the entropy of the image texture, the contrast of the image texture, the grayscale value variance of the image texture, the angular second-order moment of the image texture, the dissimilarity of the image texture, and the linear correlation of the grayscale between pixels in the image texture according to the probability distribution of the occurrence of each grayscale;

[0016] The feature processing module is also configured to establish the image texture feature based on the brightness mean of the image texture, the entropy of the image texture, the contrast of the image texture, the grayscale value variance of the image texture, the angular second-order moment of the image texture, the dissimilarity of the image texture, and the linear correlation of the grayscale between pixels in the image texture.

[0017] Furthermore, when the data processing module pre-processes the multispectral data, it includes:

[0018] The data processing module is further configured to obtain a radiation brightness value in the multispectral data, and substitute the radiation brightness value into Formula I to obtain a ground object reflectivity ρ, wherein Formula I is as follows:

[0019]

[0020] Wherein, L is the radiation brightness value, L p is the path radiance, E s is the solar radiation energy, θ is the solar zenith angle, and T is the atmospheric transmittance;

[0021] The data processing module is further configured to correct the multispectral data according to the relationship between the ground object reflectance ρ and a preset ground object reflectance ρ preconfigured by the data processing module, wherein:

[0022] If the object reflectance ρ is inconsistent with the preset object reflectance ρ, the data processing module corrects the multispectral data, obtains the response parameters of the remote sensing image acquisition equipment, and substitutes the response parameters into formula II to determine the corrected radiation brightness value L, wherein formula II is as follows:

[0023]

[0024] Among them, DN min is the minimum response parameter of the remote sensing image acquisition equipment, DN max is the maximum response parameter of the remote sensing image acquisition equipment, L min is the minimum radiation brightness of the remote sensing image acquisition equipment, L max is the maximum value of the radiation brightness of the remote sensing image acquisition equipment;

[0025] The data processing module is also configured to obtain the geographic boundary of the study area and define the clipping range based on the vector boundary. The data processing module is also configured to clip the corrected remote sensing image with a vector mask, extract the image of the study area, and define it as the remote sensing image of the study area.

[0026] Furthermore, when the feature processing module determines the fractal dimension of the remote sensing image according to the image texture feature, it includes:

[0027] The feature processing module is further configured to grid the remote sensing image according to a preset size, and obtain the texture feature value in each grid and the number of boxes containing the target texture feature;

[0028] The feature processing module is further configured to substitute the number of boxes containing target texture features into Formula III to obtain the fractal dimension of the remote sensing image, wherein Formula III is as follows:

[0029]

[0030] Where N(r) represents the number of boxes when the grid side length is r, and r represents the side length of the grid;

[0031] The feature processing module is further configured to compare the fractal dimension with the actual distribution of each tree species area in the remote sensing image, wherein:

[0032] If the fractal dimension is consistent with the actual distribution, the feature processing module determines that the fractal dimension is accurate;

[0033] If the fractal dimension is inconsistent with the actual distribution, the feature processing module determines that the fractal dimension calculation is wrong and adjusts the preset size until the fractal dimension is consistent with the actual distribution.

[0034] Furthermore, the identification module is configured to verify the classification model, and when the verified classification model generates a dominant tree species classification diagram, it includes:

[0035] The recognition module is also configured to classify the test set based on the classification model, obtain the prediction result, and compare it with the actual category label to generate a confusion matrix;

[0036] The recognition module is further configured to obtain the recall rate and precision rate of the classification model based on the confusion matrix, and determine the classification accuracy score of the classification model based on the recall rate and precision rate and the F1 score; and determine the accuracy of the classification model based on the relationship between the classification accuracy score and the preset classification accuracy score pre-configured by the recognition module, wherein:

[0037] If the classification accuracy score is lower than the preset classification accuracy score, the recognition module determines that the accuracy of the classification model is low, and iteratively corrects the classification model until the classification accuracy score is higher than or equal to the preset classification accuracy score;

[0038] If the classification accuracy score is higher than or equal to the preset classification accuracy score, the identification module determines that the classification model has high accuracy and generates the dominant tree species classification map based on the distribution of tree species characteristics in the remote sensing image identified by the classification model.

[0039] Compared with the prior art, the beneficial effect of the present invention is that through the acquisition of high-resolution remote sensing images and multi-spectral data, the system can cover a wide range of forest areas, provide a large amount of spatial and spectral information, and overcome the limitation that traditional ground surveys cannot cover large-scale. Multi-spectral data can provide information of different bands, so that in a complex forest environment, the subtle differences between different tree species can be identified, ensuring the effectiveness and comprehensiveness of remote sensing data. Secondly, the data processing module pre-processes the multi-spectral data, which not only effectively removes noise interference, but also enhances the image quality, making subsequent analysis more accurate. By generating remote sensing images of the study area, the system lays a solid foundation for feature extraction and classification, ensuring the clarity of the image and the accuracy of the processing results. The accuracy of data pre-processing directly affects the subsequent texture feature extraction and fractal dimension calculation. At the same time, the feature processing module can deeply analyze the spatial structure characteristics in the image by extracting the texture features of the image based on the grayscale co-occurrence matrix, and obtain texture information closely related to the distribution of tree species. The grayscale co-occurrence matrix can capture the spatial relationship between the grayscale levels of the image, thereby providing fine texture data for subsequent classification. Moreover, by using fractal dimension as a quantitative index of image complexity, it can accurately reflect the spatial feature differences of different tree species in remote sensing images, which helps to enhance the discrimination of tree species identification. By extracting these advanced texture features and fractal dimensions, subtle differences in tree species can be identified, thereby improving the accuracy of classification. In addition, by evaluating the effect of feature combination, the feature processing module can screen out the optimal features with the best classification ability. The optimal classification features can significantly reduce redundant data and improve the efficiency and accuracy of the classification model. When establishing the classification model, the addition of optimal features enables the model to better capture the key differences between tree species, avoid overfitting, and improve the generalization ability of the model. Finally, the classification model is verified by the recognition module to ensure that the established classification model can run stably in practical applications and generate accurate classification maps of dominant tree species. This process can not only provide accurate tree species distribution information, but also provide a scientific basis for forest resource management based on the classification results. This is of great significance for improving the efficiency of forest monitoring, ecological protection and resource management, and can provide data support for relevant decision-making.

[0040] On the other hand, the present application also provides a method for remote sensing identification of dominant tree species in forest stands based on texture feature fractal dimension, comprising:

[0041] Use high-resolution remote sensing images to obtain multispectral data of the forest study area;

[0042] Preprocessing the multispectral data and generating remote sensing images of the research area;

[0043] Extracting image texture features of the remote sensing image of the study area based on the gray level co-occurrence matrix, determining the fractal dimension of the remote sensing image according to the image texture features; and selecting the preferred classification features according to the evaluation results of the feature combination;

[0044] A classification model is established based on the preferred classification features, the classification model is verified, and the verified classification model generates a classification diagram of dominant tree species.

[0045] Furthermore, when extracting the image texture features of the remote sensing image of the study area based on the gray level co-occurrence matrix, it includes:

[0046] Acquire the spatial relationship between pixels in the remote sensing image after preprocessing, and generate a gray level co-occurrence matrix according to the gray level co-occurrence frequency of each pixel;

[0047] Obtaining the probability distribution of each gray level in the gray level co-occurrence matrix based on normalization processing;

[0048] According to the probability distribution of the occurrence of each grayscale, the brightness mean of the image texture, the entropy of the image texture, the contrast of the image texture, the grayscale value variance of the image texture, the angular second-order moment of the image texture, the dissimilarity of the image texture, and the linear correlation of the grayscale between pixels in the image texture are obtained;

[0049] The image texture feature is established according to the brightness mean of the image texture, the entropy of the image texture, the contrast of the image texture, the grayscale value variance of the image texture, the angular second-order moment of the image texture, the dissimilarity of the image texture and the linear correlation of the grayscale between pixels in the image texture.

[0050] Furthermore, when the multispectral data is preprocessed, it includes:

[0051] Obtain the radiation brightness value in the multispectral data, and substitute the radiation brightness value into formula I to obtain the ground object reflectivity ρ, wherein formula I is as follows:

[0052]

[0053] Wherein, L is the radiation brightness value, L p is the path radiance, E s is the solar radiation energy, θ is the solar zenith angle, and T is the atmospheric transmittance;

[0054] The multispectral data is corrected according to the relationship between the ground object reflectance ρ and the preset ground object reflectance ρ preconfigured by the data processing module, wherein:

[0055] If the reflectivity ρ of the ground object is inconsistent with the preset reflectivity ρ of the ground object, the multispectral data is corrected to obtain the response parameters of the remote sensing image acquisition equipment, and the response parameters are substituted into Formula II to determine the corrected radiation brightness value L, wherein Formula II is as follows:

[0056]

[0057] Among them, DN min is the minimum response parameter of the remote sensing image acquisition equipment, DN max is the maximum response parameter of the remote sensing image acquisition equipment, L min is the minimum radiation brightness of the remote sensing image acquisition equipment, L max is the maximum value of the radiation brightness of the remote sensing image acquisition equipment;

[0058] The geographic boundary of the study area is obtained, and a clipping range is defined based on the vector boundary. The data processing module is also configured to clip the corrected remote sensing image with a vector mask, extract the image of the study area, and define it as the remote sensing image of the study area.

[0059] Furthermore, when determining the fractal dimension of the remote sensing image according to the image texture feature, it includes:

[0060] Gridding the remote sensing image according to a preset size, and obtaining texture feature values ​​in each grid and the number of boxes containing target texture features;

[0061] Substitute the number of boxes containing target texture features into Formula III to obtain the fractal dimension of the remote sensing image, wherein Formula III is as follows:

[0062]

[0063] Where N(r) represents the number of boxes when the grid side length is r, and r represents the side length of the grid;

[0064] A comparison is performed between the fractal dimension and the actual distribution of each tree species area in the remote sensing image, wherein:

[0065] If the fractal dimension is consistent with the actual distribution, it is determined that the fractal dimension is accurate;

[0066] If the fractal dimension is inconsistent with the actual distribution, it is determined that the fractal dimension is calculated incorrectly, and the preset size is adjusted until the fractal dimension is consistent with the actual distribution.

[0067] Furthermore, the classification model is verified, and when the verified classification model generates a classification diagram of dominant tree species, it includes:

[0068] Classify the test set based on the classification model, obtain the prediction results, and compare them with the actual category labels to generate a confusion matrix;

[0069] Obtaining the recall rate and precision rate of the classification model based on the confusion matrix, and determining the classification accuracy score of the classification model based on the recall rate and precision rate and the F1 score;

[0070] The accuracy of the classification model is determined according to the relationship between the classification accuracy score and a preset classification accuracy score pre-configured by the recognition module, wherein:

[0071] If the classification accuracy score is lower than the preset classification accuracy score, it is determined that the accuracy of the classification model is low, and the classification model is iteratively corrected until the classification accuracy score is higher than or equal to the preset classification accuracy score;

[0072] If the classification accuracy score is higher than or equal to the preset classification accuracy score, it is determined that the classification model has high accuracy, and the dominant tree species classification map is generated based on the distribution of tree species characteristics in the remote sensing image identified by the classification model.

[0073] It can be understood that the remote sensing identification system and method for forest stand dominant tree species based on texture feature fractal dimension in the above embodiments of the present invention have the same beneficial effects and will not be described in detail. BRIEF DESCRIPTION OF THE DRAWINGS

[0074] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0075] Figure 1 A functional block diagram of a remote sensing identification system for forest stand dominant tree species based on texture feature fractal dimension provided by an embodiment of the present invention;

[0076] Figure 2 This is an effect diagram of extracting image texture features in an embodiment of the present invention;

[0077] Figure 3 A flowchart of a method for remote sensing identification of dominant tree species in a forest stand based on texture feature fractal dimension is provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0078] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art. It should be noted that, in the absence of conflict, the embodiments of the present invention and the features described in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0079] like Figure 1-Figure 2 As shown, in some embodiments of the present application, this embodiment provides a remote sensing identification system for dominant tree species in a forest stand based on texture feature fractal dimension, including: an acquisition module, a data processing module, a feature processing module, a modeling unit and an identification module.

[0080] Specifically, the acquisition module is configured to use high-resolution remote sensing images to acquire multispectral data of the forest study area; the data processing module is electrically connected to the acquisition module, and the data processing module is configured to pre-process the multispectral data and generate remote sensing images of the study area; the feature processing module is configured to extract the image texture features of the remote sensing images of the study area based on the grayscale co-occurrence matrix, and determine the fractal dimension of the remote sensing images based on the image texture features; the feature processing module is also configured to screen out preferred classification features based on the evaluation results of the feature combination; the modeling unit is electrically connected to the feature processing module, and the modeling unit is configured to establish a classification model based on the preferred classification features; the identification module is electrically connected to the modeling unit, and the identification module is configured to verify the classification model, and the verified classification model generates a classification map of dominant tree species.

[0081] It can be understood that the acquisition module is responsible for collecting remote sensing data of the forest study area, and the data processing module preprocesses these data to generate high-quality remote sensing images. The feature processing module extracts the texture features of the image based on the grayscale co-occurrence matrix, and quantifies the spatial complexity of the image through the fractal dimension, thereby effectively distinguishing the spatial characteristics of different tree species. The module also selects the preferred classification features based on the evaluation results of the feature combination, reduces redundant information, and enhances the accuracy of the classification model. The modeling unit establishes a classification model based on these preferred features, and verifies the model through the recognition module, and finally generates an accurate classification map of dominant tree species, thereby providing accurate data support for forest resource management and ecological monitoring.

[0082] It can be seen that the multispectral data of the forest study area are obtained through high-resolution remote sensing images. These data include information of different bands and can provide rich spectral characteristics of the forest area. This information is particularly important for distinguishing different tree species, because different tree species usually have similar spectral characteristics in remote sensing images, and the combination of multispectral data can help the system capture more dimensional features, thereby improving the accuracy of classification. Next, the data processing module is electrically connected to the acquisition module and is responsible for preprocessing the acquired multispectral data. The preprocessing process includes steps such as image denoising, radiation correction and geometric correction, which aim to eliminate environmental interference, adjust image quality, and make subsequent texture feature extraction more accurate. By generating processed remote sensing images, a clearer and more accurate image basis can be provided to ensure that subsequent analysis work can reflect the real forest ecological environment. The feature processing module extracts the texture features of the image through the gray level co-occurrence matrix (GLCM). The gray level co-occurrence matrix can effectively capture the spatial relationship between pixels in the image and the statistical characteristics between gray levels, thereby revealing the structural differences within the forest area. Texture features can reflect information such as the spatial distribution and morphological characteristics of tree species, and these information are very different in remote sensing images of different tree species. Therefore, texture features play an important role in tree species identification. Furthermore, the feature processing module uses the fractal dimension, a measure of image complexity, to quantify the complexity of image texture, thereby helping the system to better identify subtle differences between tree species. The calculation of fractal dimension is achieved by analyzing the multi-scale changes of details in the image, which can reflect the spatial law of tree species growth to a certain extent. Specifically, in the feature processing stage, the feature processing module will also select the preferred classification features based on the evaluation results of the feature combination. The purpose of this process is to improve the accuracy and computational efficiency of the classification model by analyzing the impact of different feature combinations on the classification results and removing redundant and irrelevant features. Through feature screening, the system can only retain features that contribute significantly to tree species identification, reduce the dimension of the feature space, reduce the complexity of the model, and avoid the risk of overfitting. The screening of preferred classification features greatly improves the accuracy and reliability of tree species classification. Then, the modeling unit constructs a classification model based on the preferred classification features. This model uses effective features extracted from remote sensing images to train and learn the characteristic distribution of different tree species, and establishes classification rules based on these features. The establishment of the classification model not only depends on the quality of the features, but also combines the optimization of the machine learning algorithm to ensure that the model can accurately classify tree species in unknown data. In order to ensure the robustness and practicality of the model, the modeling unit also verifies and adjusts the training data to optimize the classification effect. Finally, the recognition module will verify the trained classification model. The verified classification model can generate the final dominant tree species classification map based on the input remote sensing image.These classification maps can accurately display the distribution of various tree species in the study area, providing important data support for forest resource management, ecological monitoring, species protection and other aspects.

[0083] Specifically, when the feature processing module extracts the image texture features of the remote sensing image of the study area based on the grayscale co-occurrence matrix, it includes: the feature processing module is also configured to obtain the spatial relationship between each pixel in the remote sensing image after preprocessing, and generate a grayscale co-occurrence matrix according to the grayscale co-occurrence frequency of each pixel; the feature processing module is also configured to obtain the probability distribution of the occurrence of each grayscale in the grayscale co-occurrence matrix based on normalization processing; the feature processing module is also configured to obtain the brightness mean of the image texture, the entropy of the image texture, the contrast of the image texture, the grayscale value variance of the image texture, the angular second moment of the image texture, the dissimilarity of the image texture, and the linear correlation of the grayscale between each pixel in the image texture according to the probability distribution of the occurrence of each grayscale; the feature processing module is also configured to establish image texture features according to the brightness mean of the image texture, the entropy of the image texture, the contrast of the image texture, the grayscale value variance of the image texture, the angular second moment of the image texture, the dissimilarity of the image texture, and the linear correlation of the grayscale between each pixel in the image texture.

[0084] It can be understood that the feature processing module first obtains the spatial relationship between each pixel in the preprocessed remote sensing image, and generates a grayscale co-occurrence matrix by calculating the grayscale co-occurrence frequency of each pixel. Then, the grayscale co-occurrence matrix is ​​normalized to obtain the occurrence probability distribution of each grayscale value. Based on this probability distribution, the system further calculates multiple features of the image texture, including brightness mean, entropy, contrast, grayscale value variance, angular second moment, dissimilarity, and linear correlation of grayscale between pixels. These features can fully reflect the spatial characteristics and complexity of the image texture. These texture features will be used as important input data and further used in the construction of tree species classification models to help improve the accuracy and robustness of tree species identification.

[0085] It can be seen that the feature processing module is a key step in the feature extraction process by obtaining the spatial relationship between each pixel in the preprocessed remote sensing image. At this stage, the feature processing module analyzes each pixel in the image to calculate the grayscale co-occurrence frequency between it and the surrounding pixels. The grayscale co-occurrence matrix is ​​a two-dimensional statistical matrix that records the frequency of the mutual relationship between each pair of pixel grayscale values ​​in the image. In this way, the feature processing module can capture the texture features of the remote sensing image in a low-dimensional space and convert them into quantitative data that can be used for tree species identification. Next, the feature processing module will normalize the generated grayscale co-occurrence matrix to more effectively represent the probability of occurrence of different grayscale values. This step can eliminate the interference caused by factors such as different lighting and image quality in the original grayscale matrix, ensuring that the texture features more accurately reflect the structural information of the image in the classification model. By calculating the probability distribution of each grayscale value, the feature processing module provides more stable and accurate texture feature data for subsequent classification tasks. Then, based on the grayscale occurrence probability distribution calculated in the grayscale co-occurrence matrix, the feature processing module further extracts multiple statistical features of the image texture. Specifically, these features include brightness mean, image texture entropy, contrast, grayscale variance, angular second-order moment, dissimilarity, and linear correlation of grayscale between pixels. These features can reflect the spatial structure and information complexity of remote sensing images from multiple angles and scales, making texture a more comprehensive representation of image content. By extracting and combining these features, the feature processing module can provide richer and more efficient texture feature information for the tree species recognition model, which plays an important role in improving recognition accuracy and classification effect. Finally, by synthesizing the extracted texture features, the feature processing module establishes the final image texture feature data. This data set will serve as the input of the tree species classification model to provide necessary feature support for model training and verification. In this way, the feature processing module can not only improve the accuracy of tree species recognition, but also provide effective data support for other ecological monitoring and resource management tasks based on remote sensing images. In actual operation, the feature processing module processes remote sensing images efficiently to ensure that the texture feature data can truly and accurately reflect the tree species information in the forest area, laying a solid foundation for subsequent classification and recognition work.

[0086] Specifically, when the data processing module preprocesses the multispectral data, the data processing module is further configured to obtain the radiation brightness value in the multispectral data, and substitute the radiation brightness value into formula I to obtain the reflectivity ρ of the ground object, wherein formula I is as follows:

[0087] Among them, L is the radiation brightness value, L p is the path radiance, E sis the solar radiation energy, θ is the solar zenith angle, and T is the atmospheric transmittance; the data processing module is also configured to correct the multispectral data according to the relationship between the ground object reflectance ρ and the preset ground object reflectance ρ pre-configured by the data processing module, wherein: if the ground object reflectance ρ is inconsistent with the preset ground object reflectance ρ, the data processing module corrects the multispectral data, obtains the response parameters of the remote sensing image acquisition equipment, and substitutes the response parameters into formula II to determine the corrected radiation brightness value L, wherein formula II is as follows:

[0088]

[0089] Among them, DN min is the minimum response parameter of the remote sensing image acquisition equipment, DN max is the maximum response parameter of the remote sensing image acquisition equipment, L min is the minimum radiation brightness of the remote sensing image acquisition equipment, L max is the maximum radiation brightness of the remote sensing image acquisition device; the data processing module is also configured to obtain the geographic boundary of the study area and define the clipping range based on the vector boundary. The data processing module is also configured to clip the corrected remote sensing image with a vector mask, extract the image of the study area, and define it as the remote sensing image of the study area.

[0090] It can be seen that the data processing module obtains the radiance value in the multispectral data and converts it into the reflectance ρ of the ground object through formula I, thereby eliminating the influence of factors such as the atmosphere and the sun angle and obtaining more accurate ground object information. Then, the module corrects the multispectral data according to the difference between the ground object reflectance ρ and the preset ground object reflectance ρ. If the two are inconsistent, the module further adjusts the data and calculates the corrected radiance value through formula II to ensure that the response parameters of the remote sensing image acquisition equipment are effectively corrected. Finally, the data processing module uses the geographic boundary information of the study area to crop the corrected remote sensing image based on the vector mask, retaining only the image data of the study area, thereby providing an accurate image basis for subsequent texture feature extraction and tree species classification. Through these steps, the accuracy of the data and the specificity of the region are ensured, providing high-quality input data for remote sensing analysis.

[0091] It can be understood that the data processing module obtains the radiance value in the multispectral data, substitutes it into formula I, and calculates the reflectance ρ of the ground object. The core purpose of this step is to perform atmospheric correction on the original radiance value to eliminate the influence of external factors such as the atmosphere and solar radiation angle. In this way, the spectral information reflected by the surface of the ground object can be accurately restored, so that the remote sensing image can more realistically reflect the spectral characteristics of the ground object, providing a reliable data basis for subsequent analysis. Next, the data processing module compares the calculated ground object reflectance ρ with the preset standard ground object reflectance ρ to determine whether there is an error in the multispectral data. If it is found that there is an inconsistency between the ground object reflectance ρ and the preset value, the data processing module will start the correction program to ensure the accuracy of the data. At this time, by substituting formula II, the data processing module calculates the corrected radiance value L to ensure that the response parameters of the remote sensing image acquisition equipment are accurately corrected. The main purpose of this link is to eliminate the errors caused by inconsistent equipment response, ensure the authenticity and consistency of the image data, and make it more suitable for subsequent analysis and model training. After the acquisition and correction of the remote sensing image, the data processing module will also perform image cropping and region extraction. Using the geographic boundary information of the study area, the data processing module clips the calibrated remote sensing image through a vector mask to extract the image data within the study area. This process can not only eliminate the influence of irrelevant areas, but also accurately focus on the study area, improving the efficiency and accuracy of subsequent analysis. The clipped remote sensing image will serve as the basic data for the feature processing module to further extract texture features and establish a classification model.

[0092] Specifically, when the feature processing module determines the fractal dimension of the remote sensing image according to the image texture feature, it includes: the feature processing module is also configured to grid the remote sensing image according to a preset size, and obtain the texture feature value in each grid and the number of boxes containing the target texture feature; the feature processing module is also configured to substitute the number of boxes containing the target texture feature into formula III to obtain the fractal dimension of the remote sensing image, wherein formula III is as follows: Among them, N(r) represents the number of boxes when the grid side length is r, and r represents the side length of the grid; the feature processing module is also configured to compare the fractal dimension with the actual distribution of each tree species area in the remote sensing image, wherein: if the fractal dimension is consistent with the actual distribution, the feature processing module determines that the fractal dimension is accurate; if the fractal dimension is inconsistent with the actual distribution, the feature processing module determines that the fractal dimension is calculated incorrectly and adjusts the preset size until the fractal dimension is consistent with the actual distribution.

[0093] It can be understood that the feature processing module grids the remote sensing image, divides the image into multiple small grids according to the preset size, and calculates the texture feature value in each grid and the number of boxes containing the target texture feature. Then, by substituting into Formula III, the fractal dimension of the image is calculated using the relationship between the grid side length r and the number of boxes N(r). This process quantifies the texture complexity of the image through the fractal dimension, thereby revealing the spatial structural characteristics of the tree species in the image. In addition, the feature processing module will also verify the fractal dimension. By comparing the fractal dimension with the actual distribution of each tree species area in the remote sensing image, if the calculated fractal dimension is consistent with the actual distribution, the fractal dimension is considered to be accurate; if not, it is considered that there is an error in the calculation result, and the preset size is adjusted until the fractal dimension matches the actual distribution. The core of this step is to ensure that the description of the texture characteristics of the remote sensing image through the fractal dimension has a high accuracy, thereby improving the accuracy of tree species identification.

[0094] It can be seen that the feature processing module divides the remote sensing image into several small areas by gridding the image. The texture feature value in each area and the number of boxes containing the target texture feature will become the basis for the subsequent fractal dimension calculation. The main purpose of gridding is to convert a large-scale remote sensing image into small units that are easy to analyze, so as to capture and quantify the texture complexity in different areas. By analyzing these small units, the detail level of the image can be grasped more accurately, especially the slight differences between tree species in the forest area can be identified. This method can overcome the complexity that may be encountered when directly analyzing the entire image, making the extraction of texture features more efficient and accurate. With the completion of the gridding process, the feature processing module further calculates the fractal dimension of the image through formula III. Fractal dimension is an important indicator to measure the complexity of image texture, reflecting the distribution law and hierarchical structure of each texture element in the image. In remote sensing images, different tree species may show different texture characteristics in space, and fractal dimension can effectively capture the differences in these spatial characteristics. By counting the texture feature values ​​in each grid and the number of boxes containing the target texture features, the feature processing module can calculate the fractal dimension of the image, thereby providing a quantitative way to describe the spatial structure of the image. Specifically, the relationship between the number of boxes N(r) and the grid side length r in Formula III can reflect the self-similarity of the image and maintain consistency at different scales, which makes the fractal dimension have a unique advantage in analyzing complex textures. In order to ensure the accuracy of the fractal dimension calculation, the feature processing module also sets a comparison mechanism, that is, by comparing the calculated fractal dimension with the actual tree species distribution, the reliability of the fractal dimension is verified. If the calculated fractal dimension is consistent with the actual distribution, the calculation result is considered to be correct; if it is inconsistent, it indicates that there is an error in the calculation of the fractal dimension and the preset size needs to be adjusted. Through this adjustment process, the system can self-optimize and continuously improve the calculation accuracy until the fractal dimension matches the actual distribution of the tree species area. This process is similar to a feedback mechanism that can dynamically adjust the image analysis parameters to improve the accuracy of the results. This adaptive mechanism based on fractal dimension adjustment not only improves the accuracy of tree species identification, but also enhances the system's adaptability to different environmental conditions and changes in image quality. In addition, this method based on fractal dimension and grid processing has strong scalability and can be applied in a variety of remote sensing image analysis. With the continuous development of remote sensing technology, the resolution of acquired images is getting higher and higher, and the complexity of images is also increasing. In this case, traditional texture analysis methods may face problems such as low processing efficiency and insufficient accuracy. Through grid processing and fractal dimension calculation, the micro-texture differences in the image can be effectively captured at a higher resolution, avoiding the analysis difficulties caused by the high complexity of the image.Therefore, this method can not only be applied to remote sensing identification of dominant tree species in forest stands, but can also be extended to other types of remote sensing image analysis, such as agricultural monitoring, urban green space analysis and other fields, and has broad application prospects.

[0095] Specifically, the recognition module is configured to verify the classification model, and when the verified classification model generates a classification map of dominant tree species, it includes: the recognition module is also configured to classify the test set based on the classification model, obtain the prediction result, and compare it with the actual category label to generate a confusion matrix; the recognition module is also configured to obtain the recall rate and precision rate of the classification model based on the confusion matrix, and determine the classification accuracy score of the classification model based on the recall rate and the precision rate and the F1 score; and determine the accuracy of the classification model based on the relationship between the classification accuracy score and the preset classification accuracy score pre-configured by the recognition module, wherein: if the classification accuracy score is lower than the preset classification accuracy score, the recognition module determines that the accuracy of the classification model is low, and iteratively corrects the classification model until the classification accuracy score is higher than or equal to the preset classification accuracy score; if the classification accuracy score is higher than or equal to the preset classification accuracy score, the recognition module determines that the accuracy of the classification model is high, and generates a classification map of dominant tree species based on the distribution of tree species characteristics in the remote sensing image identified by the classification model.

[0096] It can be seen that the recognition module obtains the prediction results by inputting the test set into the classification model for classification, and compares the prediction results with the actual category labels to generate a confusion matrix. Through the confusion matrix, the recognition module can further calculate the recall and precision of the classification model, and then comprehensively evaluate the performance of the classification model through the F1 score. The calculation of the classification accuracy score is based on these indicators. If the score is lower than the preset value, the model is iteratively corrected to improve the classification accuracy. Finally, the optimized classification model is used to generate the dominant tree species classification map. This process ensures the accuracy of the classification model and achieves continuous optimization by dynamically adjusting the model parameters, thereby ensuring that the generated tree species classification map has a high degree of accuracy.

[0097] It can be understood that the recognition module first classifies the test set based on the established classification model and compares the predicted results with the actual category labels. Through this comparison, a confusion matrix is ​​generated, which contains the specific distribution information of the classification results. The confusion matrix can intuitively reflect the prediction accuracy of the classification model in different categories, and thus provide a basis for model optimization. By comparing the actual labels with the predicted results, it is possible to identify which tree species are misclassified, thereby providing a direction for subsequent optimization. Subsequently, the recognition module calculates the recall and precision of the classification model. The recall rate measures the ability of the model to correctly identify the target category, while the precision rate reflects the proportion of the category that is actually the target category when the model predicts it to be a certain category. Combining these two indicators, the recognition module further calculates the F1 score, which is an indicator that comprehensively measures the performance of the classification model. The F1 score can effectively balance the recall rate and precision rate, avoid the deviation that may be caused by a single indicator, and make the classification model more balanced in various types of data. If the calculated classification accuracy score is lower than the preset threshold, the recognition module will determine that the accuracy of the classification model is not high and will trigger an iterative correction process. By analyzing the reasons for misclassification, the recognition module will adjust the parameters of the classification model, retrain, and re-verify the classification ability of the model. This process is a continuous optimization process that aims to continuously improve the accuracy of the classification model until the preset classification accuracy score is reached. The iterative process can continuously adjust the classification algorithm or add new features according to different test results to make up for the shortcomings of the existing model in identifying certain tree species. When the classification accuracy score reaches the preset standard, the recognition module will confirm that the accuracy of the classification model meets the standard and use the model for actual tree species classification. By analyzing remote sensing images through this model, the recognition module can accurately identify the distribution of tree species characteristics in different areas of the image. The spatial distribution information of these tree species characteristics will be converted into a dominant tree species classification map, providing important data support for forest management and ecological research. The dominant tree species classification map can help researchers identify the distribution of the most dominant tree species in forest areas and provide a decision-making basis for resource management and ecological protection.

[0098] In the above embodiment, through the acquisition of high-resolution remote sensing images and multi-spectral data, the system can cover a wide range of forest areas, provide a large amount of spatial and spectral information, and overcome the limitation that traditional ground surveys cannot cover on a large scale. Multi-spectral data can provide information of different bands, so that in a complex forest environment, the subtle differences between different tree species can be identified, ensuring the effectiveness and comprehensiveness of remote sensing data. Secondly, the data processing module pre-processes the multi-spectral data, which not only effectively removes noise interference, but also enhances the image quality, making subsequent analysis more accurate. By generating remote sensing images of the study area, the system lays a solid foundation for feature extraction and classification, ensuring the clarity of the image and the accuracy of the processing results. The accuracy of data pre-processing directly affects the subsequent texture feature extraction and fractal dimension calculation. At the same time, the feature processing module can deeply analyze the spatial structure characteristics in the image by extracting the texture features of the image based on the grayscale co-occurrence matrix, and obtain texture information closely related to the distribution of tree species. The grayscale co-occurrence matrix can capture the spatial relationship between the grayscale levels of the image, thereby providing fine texture data for subsequent classification. Moreover, by using fractal dimension as a quantitative index of image complexity, it can accurately reflect the spatial feature differences of different tree species in remote sensing images, which helps to enhance the discrimination of tree species identification. By extracting these advanced texture features and fractal dimensions, subtle differences in tree species can be identified, thereby improving the accuracy of classification. In addition, by evaluating the effect of feature combination, the feature processing module can screen out the optimal features with the best classification ability. The optimal classification features can significantly reduce redundant data and improve the efficiency and accuracy of the classification model. When establishing the classification model, the addition of optimal features enables the model to better capture the key differences between tree species, avoid overfitting, and improve the generalization ability of the model. Finally, the classification model is verified by the recognition module to ensure that the established classification model can run stably in practical applications and generate accurate classification maps of dominant tree species. This process can not only provide accurate tree species distribution information, but also provide a scientific basis for forest resource management based on the classification results. This is of great significance for improving the efficiency of forest monitoring, ecological protection and resource management, and can provide data support for relevant decision-making.

[0099] In another preferred embodiment based on the above embodiment, Figure 3 This embodiment provides a method for remote sensing identification of dominant tree species in a forest stand based on texture feature fractal dimension, including:

[0100] Step S100: Acquire multispectral data of the forest study area using high-resolution remote sensing images.

[0101] Step S200: pre-process the multispectral data and generate a remote sensing image of the study area.

[0102] Specifically, the preprocessing of multispectral data includes: obtaining the radiation brightness value in the multispectral data, and substituting the radiation brightness value into formula I to obtain the reflectivity ρ of the ground object, where formula I is as follows:

[0103]

[0104] Among them, L is the radiation brightness value, L p is the path radiance, E s is the solar radiation energy, θ is the solar zenith angle, and T is the atmospheric transmittance. According to the relationship between the ground object reflectance ρ and the preset ground object reflectance ρ pre-configured by the data processing module, the multispectral data is corrected, where: if the ground object reflectance ρ is inconsistent with the preset ground object reflectance ρ, the multispectral data is corrected, the response parameters of the remote sensing image acquisition equipment are obtained, and the response parameters are substituted into formula II to determine the corrected radiation brightness value L, where formula II is as follows:

[0105]

[0106] Among them, DN min is the minimum response parameter of the remote sensing image acquisition equipment, DN max is the maximum response parameter of the remote sensing image acquisition equipment, L min is the minimum radiation brightness of the remote sensing image acquisition equipment, L max is the maximum value of the radiation brightness of the remote sensing image acquisition device. The geographic boundary of the study area is obtained, and the clipping range is defined based on the vector boundary. The data processing module is also configured to clip the corrected remote sensing image with a vector mask, extract the image of the study area, and define it as the remote sensing image of the study area.

[0107] Step S300: extracting image texture features of the remote sensing image of the study area based on the gray level co-occurrence matrix, determining the fractal dimension of the remote sensing image based on the image texture features, and selecting the preferred classification features based on the evaluation results of the feature combination.

[0108] Specifically, when extracting the image texture features of the remote sensing image of the study area based on the gray level co-occurrence matrix, it includes: obtaining the spatial relationship between each pixel in the remote sensing image after preprocessing, and generating a gray level co-occurrence matrix based on the gray level co-occurrence frequency of each pixel. Based on the normalization process, the probability distribution of each gray level in the gray level co-occurrence matrix is ​​obtained. According to the probability distribution of each gray level, the brightness mean of the image texture, the entropy of the image texture, the contrast of the image texture, the gray level value variance of the image texture, the angular second-order moment of the image texture, the dissimilarity of the image texture, and the linear correlation of the gray levels between each pixel in the image texture are obtained. Image texture features are established based on the brightness mean of the image texture, the entropy of the image texture, the contrast of the image texture, the gray level value variance of the image texture, the angular second-order moment of the image texture, the dissimilarity of the image texture, and the linear correlation of the gray levels between each pixel in the image texture.

[0109] Specifically, when determining the fractal dimension of a remote sensing image based on image texture features, it includes: gridding the remote sensing image according to a preset size, and obtaining the texture feature value in each grid and the number of boxes containing the target texture feature. Substituting the number of boxes containing the target texture feature into formula III to obtain the fractal dimension of the remote sensing image, where formula III is as follows:

[0110]

[0111] Where N(r) represents the number of boxes when the grid side length is r, and r represents the side length of the grid. A comparison is made between the fractal dimension and the actual distribution of each tree species area in the remote sensing image, where: if the fractal dimension is consistent with the actual distribution, the fractal dimension is determined to be accurate. If the fractal dimension is inconsistent with the actual distribution, the fractal dimension calculation is determined to be wrong, and the preset size is adjusted until the fractal dimension is consistent with the actual distribution.

[0112] Step S400: establishing a classification model based on the preferred classification features, verifying the classification model, and generating a dominant tree species classification diagram using the verified classification model.

[0113] Specifically, the classification model is verified, and when the verified classification model generates a classification map of dominant tree species, it includes: classifying the test set based on the classification model to obtain the prediction results, and comparing them with the actual category labels to generate a confusion matrix. The recall rate and precision rate of the classification model are obtained based on the confusion matrix, and the classification accuracy score of the classification model is determined based on the recall rate and precision rate and the F1 score; the accuracy of the classification model is determined based on the relationship between the classification accuracy score and the preset classification accuracy score pre-configured by the recognition module, wherein: if the classification accuracy score is lower than the preset classification accuracy score, the accuracy of the classification model is determined to be low, and the classification model is iteratively corrected until the classification accuracy score is higher than or equal to the preset classification accuracy score. If the classification accuracy score is higher than or equal to the preset classification accuracy score, the accuracy of the classification model is determined to be high, and a classification map of dominant tree species is generated based on the distribution of tree species characteristics in the remote sensing image identified by the classification model.

[0114] It can be understood that the remote sensing identification system and method for forest stand dominant tree species based on texture feature fractal dimension in the above embodiments of the present invention have the same beneficial effects and will not be described in detail.

[0115] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0116] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0117] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0118] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0119] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A remote sensing identification system for dominant tree species in forest stands based on texture feature fractal dimension, characterized in that: include: an acquisition module configured to acquire multispectral data of a forest study area using high-resolution remote sensing images; A data processing module, electrically connected to the acquisition module, the data processing module is configured to pre-process the multispectral data and generate a remote sensing image of a research area; The feature processing module is configured to extract the image texture features of the remote sensing image of the study area based on the gray level co-occurrence matrix, and determine the fractal dimension of the remote sensing image according to the image texture features; the feature processing module is also configured to screen out the preferred classification features according to the evaluation results of the feature combination; a modeling unit, electrically connected to the feature processing module, wherein the modeling unit is configured to establish a classification model based on the preferred classification feature; An identification module is electrically connected to the modeling unit, and the identification module is configured to verify the classification model, and the verified classification model generates a dominant tree species classification diagram.

2. The remote sensing identification system for forest stand dominant tree species based on texture feature fractal dimension according to claim 1, characterized in that: When the feature processing module extracts the image texture features of the remote sensing image of the study area based on the gray level co-occurrence matrix, it includes: The feature processing module is further configured to obtain the spatial relationship between the pixels in the remote sensing image after preprocessing, and generate a gray level co-occurrence matrix according to the gray level co-occurrence frequency of each pixel; The feature processing module is further configured to obtain the probability distribution of occurrence of each gray level in the gray level co-occurrence matrix based on normalization processing; The feature processing module is further configured to obtain the brightness mean of the image texture, the entropy of the image texture, the contrast of the image texture, the grayscale value variance of the image texture, the angular second-order moment of the image texture, the dissimilarity of the image texture, and the linear correlation of the grayscale between pixels in the image texture according to the probability distribution of the occurrence of each grayscale; The feature processing module is also configured to establish the image texture feature based on the brightness mean of the image texture, the entropy of the image texture, the contrast of the image texture, the grayscale value variance of the image texture, the angular second-order moment of the image texture, the dissimilarity of the image texture, and the linear correlation of the grayscale between pixels in the image texture.

3. The remote sensing identification system for forest stand dominant tree species based on texture feature fractal dimension according to claim 1, characterized in that: When the data processing module pre-processes the multispectral data, it includes: The data processing module is further configured to obtain a radiation brightness value in the multispectral data, and substitute the radiation brightness value into Formula I to obtain a ground object reflectivity ρ, wherein Formula I is as follows: Wherein, L is the radiation brightness value, L p is the path radiance, E s is the solar radiation energy, θ is the solar zenith angle, and T is the atmospheric transmittance; The data processing module is further configured to correct the multispectral data according to the relationship between the ground object reflectance ρ and a preset ground object reflectance ρ preconfigured by the data processing module, wherein: If the object reflectance ρ is inconsistent with the preset object reflectance ρ, the data processing module corrects the multispectral data, obtains the response parameters of the remote sensing image acquisition equipment, and substitutes the response parameters into formula II to determine the corrected radiation brightness value L, wherein formula II is as follows: Among them, DN min is the minimum response parameter of the remote sensing image acquisition equipment, DN max is the maximum response parameter of the remote sensing image acquisition equipment, L min is the minimum radiation brightness of the remote sensing image acquisition equipment, L max is the maximum value of the radiation brightness of the remote sensing image acquisition equipment; The data processing module is also configured to obtain the geographic boundary of the study area and define the clipping range based on the vector boundary. The data processing module is also configured to clip the corrected remote sensing image with a vector mask, extract the image of the study area, and define it as the remote sensing image of the study area.

4. The remote sensing identification system for forest stand dominant tree species based on texture feature fractal dimension according to claim 1, characterized in that: When the feature processing module determines the fractal dimension of the remote sensing image according to the image texture feature, it includes: The feature processing module is further configured to grid the remote sensing image according to a preset size, and obtain the texture feature value in each grid and the number of boxes containing the target texture feature; The feature processing module is further configured to substitute the number of boxes containing target texture features into Formula III to obtain the fractal dimension of the remote sensing image, wherein Formula III is as follows: Where N(r) represents the number of boxes when the grid side length is r, and r represents the side length of the grid; The feature processing module is further configured to compare the fractal dimension with the actual distribution of each tree species area in the remote sensing image, wherein: If the fractal dimension is consistent with the actual distribution, the feature processing module determines that the fractal dimension is accurate; If the fractal dimension is inconsistent with the actual distribution, the feature processing module determines that the fractal dimension calculation is wrong and adjusts the preset size until the fractal dimension is consistent with the actual distribution.

5. The remote sensing identification system for forest stand dominant tree species based on texture feature fractal dimension according to claim 1, characterized in that: The identification module is configured to verify the classification model, and when the verified classification model generates a dominant tree species classification diagram, it includes: The recognition module is also configured to classify the test set based on the classification model, obtain the prediction result, and compare it with the actual category label to generate a confusion matrix; The recognition module is further configured to obtain the recall rate and precision rate of the classification model based on the confusion matrix, and determine the classification accuracy score of the classification model based on the recall rate and precision rate and the F1 score; and determine the accuracy of the classification model based on the relationship between the classification accuracy score and the preset classification accuracy score pre-configured by the recognition module, wherein: If the classification accuracy score is lower than the preset classification accuracy score, the recognition module determines that the accuracy of the classification model is low, and iteratively corrects the classification model until the classification accuracy score is higher than or equal to the preset classification accuracy score; If the classification accuracy score is higher than or equal to the preset classification accuracy score, the identification module determines that the classification model has high accuracy and generates the dominant tree species classification map based on the distribution of tree species characteristics in the remote sensing image identified by the classification model.

6. A method for remote sensing identification of dominant tree species in a forest stand based on texture feature fractal dimension, which is used in a remote sensing identification system for dominant tree species in a forest stand based on texture feature fractal dimension as claimed in claims 1 to 5, characterized in that: include: Use high-resolution remote sensing images to obtain multispectral data of the forest study area; Preprocessing the multispectral data and generating remote sensing images of the research area; Extracting image texture features of the remote sensing image of the study area based on the gray level co-occurrence matrix, determining the fractal dimension of the remote sensing image according to the image texture features; and selecting the preferred classification features according to the evaluation results of the feature combination; A classification model is established based on the preferred classification features, the classification model is verified, and the verified classification model generates a classification diagram of dominant tree species.

7. The method for remote sensing identification of dominant tree species in forest stands based on texture feature fractal dimension according to claim 6, characterized in that: When extracting the image texture features of the remote sensing image of the study area based on the gray level co-occurrence matrix, it includes: Acquire the spatial relationship between pixels in the remote sensing image after preprocessing, and generate a gray level co-occurrence matrix according to the gray level co-occurrence frequency of each pixel; Obtaining the probability distribution of each gray level in the gray level co-occurrence matrix based on normalization processing; According to the probability distribution of the occurrence of each grayscale, the brightness mean of the image texture, the entropy of the image texture, the contrast of the image texture, the grayscale value variance of the image texture, the angular second-order moment of the image texture, the dissimilarity of the image texture, and the linear correlation of the grayscale between pixels in the image texture are obtained; The image texture feature is established according to the brightness mean of the image texture, the entropy of the image texture, the contrast of the image texture, the grayscale value variance of the image texture, the angular second-order moment of the image texture, the dissimilarity of the image texture and the linear correlation of the grayscale between pixels in the image texture.

8. The method for remote sensing identification of dominant tree species in forest stands based on texture feature fractal dimension according to claim 6, characterized in that: The preprocessing of the multispectral data includes: Obtain the radiation brightness value in the multispectral data, and substitute the radiation brightness value into formula I to obtain the ground object reflectivity ρ, wherein formula I is as follows: Wherein, L is the radiation brightness value, L p is the path radiance, E s is the solar radiation energy, θ is the solar zenith angle, and T is the atmospheric transmittance; The multispectral data is corrected according to the relationship between the ground object reflectance ρ and the preset ground object reflectance ρ preconfigured by the data processing module, wherein: If the reflectivity ρ of the ground object is inconsistent with the preset reflectivity ρ of the ground object, the multispectral data is corrected to obtain the response parameters of the remote sensing image acquisition equipment, and the response parameters are substituted into Formula II to determine the corrected radiation brightness value L, wherein Formula II is as follows: Among them, DN min is the minimum response parameter of the remote sensing image acquisition equipment, DN max is the maximum response parameter of the remote sensing image acquisition equipment, L min is the minimum radiation brightness of the remote sensing image acquisition equipment, L max is the maximum value of the radiation brightness of the remote sensing image acquisition equipment; The geographic boundary of the study area is obtained, and a clipping range is defined based on the vector boundary. The data processing module is also configured to clip the corrected remote sensing image with a vector mask, extract the image of the study area, and define it as the remote sensing image of the study area.

9. The method for remote sensing identification of dominant tree species in forest stands based on texture feature fractal dimension according to claim 6, characterized in that: Determining the fractal dimension of the remote sensing image according to the image texture feature includes: Gridding the remote sensing image according to a preset size, and obtaining texture feature values ​​in each grid and the number of boxes containing target texture features; Substitute the number of boxes containing target texture features into Formula III to obtain the fractal dimension of the remote sensing image, wherein Formula III is as follows: Where N(r) represents the number of boxes when the grid side length is r, and r represents the side length of the grid; A comparison is performed between the fractal dimension and the actual distribution of each tree species area in the remote sensing image, wherein: If the fractal dimension is consistent with the actual distribution, it is determined that the fractal dimension is accurate; If the fractal dimension is inconsistent with the actual distribution, it is determined that the fractal dimension is calculated incorrectly, and the preset size is adjusted until the fractal dimension is consistent with the actual distribution.

10. The method for remote sensing identification of dominant tree species in forest stands based on texture feature fractal dimension according to claim 6, characterized in that: Verifying the classification model and generating a dominant tree species classification diagram using the verified classification model includes: Classify the test set based on the classification model, obtain the prediction results, and compare them with the actual category labels to generate a confusion matrix; Obtaining the recall rate and precision rate of the classification model based on the confusion matrix, and determining the classification accuracy score of the classification model based on the recall rate and precision rate and the F1 score; The accuracy of the classification model is determined according to the relationship between the classification accuracy score and a preset classification accuracy score pre-configured by the recognition module, wherein: If the classification accuracy score is lower than the preset classification accuracy score, it is determined that the accuracy of the classification model is low, and the classification model is iteratively corrected until the classification accuracy score is higher than or equal to the preset classification accuracy score; If the classification accuracy score is higher than or equal to the preset classification accuracy score, it is determined that the classification model has high accuracy, and the dominant tree species classification map is generated based on the distribution of tree species characteristics in the remote sensing image identified by the classification model.

Images