Vegetation coverage rate measuring method and system

Abstract

The invention discloses a vegetation coverage rate measurement method and system, and belongs to the technical field of vegetation coverage rate measurement, and the method comprises the steps: employing a feature matching-based image registration method for a heterogeneous image obtained through the shooting of a field handheld device, extracting the significant feature points in the image, calculating the similarity between the feature points, and obtaining the vegetation coverage rate of the heterogeneous image; determining the corresponding relation between the images according to the image matching result, further realizing geometric correction and spatial registration of the images, eliminating the influence of view angle difference and resolution inconsistency, and obtaining an overall image with consistent space; in order to guarantee the accuracy of the whole image, in the image fusion process, according to the feature point distribution condition of the image overlapping area, the fusion weight is adjusted in a self-adaptive mode, weighted averaging is carried out on pixel gray values in the overlapping area, it is guaranteed that the fused image is in smooth transition in the overlapping area, obvious splicing traces are avoided, and the image fusion accuracy is improved. And the visual quality and the measurement precision of the whole image are improved.

Description

Technical Field

[0001] The invention belongs to the technical field of vegetation coverage measurement, and in particular relates to a vegetation coverage measurement method and system. Background Art

[0002] There are many technical challenges in the process of measuring vegetation coverage using handheld devices in the field. First, the regional images taken by handheld devices in the field often have problems such as different perspectives and inconsistent resolutions. How to effectively fuse these heterogeneous images and construct complete, high-quality overall image data is a major challenge. Secondly, when extracting green vegetation, relying solely on color information is easily disturbed by factors such as lighting and shadows, resulting in inaccurate extraction results. Recognition methods based on features such as texture and shape may face the dilemma of high computational complexity and insufficient robustness.

[0003] Furthermore, even if the overall image is obtained, how to verify the accuracy is also a thorny issue. The traditional manual sampling verification method is time-consuming and labor-intensive, while the automated verification algorithm is difficult to balance accuracy and efficiency. Finally, in the process of comprehensive analysis and removal of overlapping parts, how to balance local optimization and global optimization to avoid over-segmentation or under-segmentation is also an issue worthy of in-depth discussion. Summary of the invention

[0004] In order to solve the above technical problems, the present invention proposes a method and system for measuring vegetation coverage to solve the problems existing in the above prior art.

[0005] To achieve the above object, the present invention provides a method for measuring vegetation coverage, comprising:

[0006] Install a depth camera on a handheld device in the field, and obtain a depth heterogeneous image of the area to be tested based on the depth camera;

[0007] Preprocessing the depth heterogeneous images to obtain a spatially consistent overall image, and fusing the spatially consistent overall image to obtain a fused data set;

[0008] Building a multi-feature fusion model for vegetation identification based on a random forest classifier, inputting the fusion data set into the multi-feature fusion model for vegetation identification for calculation, and obtaining a binary segmentation result of vegetation;

[0009] Processing the binary segmentation result of the vegetation based on the region merging method of graph theory and the minimum spanning tree algorithm to obtain a merged vegetation region;

[0010] Calculating a vegetation index value for each pixel in the merged vegetation area, generating a detailed distribution map of vegetation based on the vegetation index value, and calculating an estimated value of vegetation coverage of the merged vegetation area;

[0011] The detailed distribution map of vegetation and the estimated value of vegetation coverage are combined to generate a final vegetation coverage measurement result.

[0012] Preferably, the process of preprocessing the depth heterogeneous image to obtain a spatially consistent overall image includes:

[0013] The significant feature points in the depth heterogeneous image are extracted, the similarity between the significant feature points is calculated, and the depth heterogeneous image is geometrically corrected and spatially registered based on the similarity to obtain a spatially consistent overall image.

[0014] Preferably, the process of fusing the spatially consistent whole images to obtain a fused data set includes:

[0015] Extracting feature points of the spatially consistent overall image and calculating the distribution density of the feature points;

[0016] Determine the image overlapping area based on the distribution density of the feature points, and calculate the initial fusion weight of each pixel in the overlapping area;

[0017] For each pixel in the overlapping area, obtain the grayscale value of its corresponding position in the image to be fused, and perform weighted average on the grayscale values ​​according to the initial fusion weight to obtain the fused grayscale value of the pixel;

[0018] The grayscale values ​​of the fused pixels in the overlapping area are spliced ​​with the original grayscale values ​​of the pixels in the non-overlapping area to obtain a preliminary fused image;

[0019] Performing image enhancement processing on the preliminary fused image to generate the fused data set.

[0020] Preferably, the process of obtaining the binary segmentation result of vegetation includes:

[0021] After color conversion processing is performed on the fused data set, information is extracted to obtain color features, texture features and shape features;

[0022] Normalizing the color features, texture features, and shape features to obtain a multi-feature fusion vector;

[0023] Constructing a random forest classifier, training the random forest classifier based on a historical data set, and obtaining a multi-feature fusion model for vegetation recognition;

[0024] The multi-feature fusion vector is input into the multi-feature fusion model of vegetation recognition for calculation to obtain the binary segmentation result of the vegetation.

[0025] Preferably, the process of extracting information after color conversion processing is performed on the fused data set to obtain color features, texture features and shape features includes:

[0026] After converting the fused data set from RGB space to HSV space, extracting green channel information, and obtaining color features of green vegetation;

[0027] Calculating the texture features of the fused data set based on the gray level co-occurrence matrix method to obtain the texture features of the vegetation area;

[0028] According to the shape characteristics of the vegetation area, the shape features of the vegetation are extracted.

[0029] Preferably, the process of obtaining the merged vegetation area includes:

[0030] The region merging method based on graph theory converts the binary segmentation result of the vegetation into an undirected weighted graph, and gradually merges regions with high similarity through a minimum spanning tree algorithm until all regions are merged into a whole and then output to generate the merged vegetation region.

[0031] Preferably, the process of generating a detailed distribution map of vegetation based on the vegetation index value and calculating the estimated value of vegetation coverage of the merged vegetation area comprises:

[0032] Obtaining the vegetation area of ​​the merged vegetation area based on pixel statistics, calculating the total proportion of the merged vegetation area, and obtaining a preliminary vegetation coverage estimation value based on the total proportion;

[0033] Acquire multispectral information of the overall image, and calculate a vegetation index value for each pixel based on the spectral information;

[0034] Calculating the NDVI value of the vegetation index value of each pixel, and generating a vegetation index distribution map based on the NDVI value;

[0035] The vegetation index distribution map is binarized to obtain a detailed vegetation distribution map.

[0036] Preferably, the process of generating the final vegetation coverage measurement result includes:

[0037] Counting the number of pixels with a value of 1 in the fine distribution map of the vegetation, and dividing the result by the total number of pixels in the image, to obtain a fine vegetation coverage rate based on the vegetation index;

[0038] The preliminary vegetation coverage estimation value and the refined vegetation coverage based on the vegetation index are weighted averaged to obtain a final vegetation coverage measurement result.

[0039] To achieve the above object, the present invention also provides a vegetation coverage measurement system, comprising:

[0040] A data acquisition module, used to install a depth camera on a field handheld device, and acquire a depth heterogeneous image of the area to be measured based on the depth camera;

[0041] A data fusion module is used to preprocess the depth heterogeneous images to obtain a spatially consistent overall image, and fuse the spatially consistent overall image to obtain a fused data set;

[0042] A model calculation module is used to construct a multi-feature fusion model for vegetation identification based on a random forest classifier, input the fusion data set into the multi-feature fusion model for vegetation identification for calculation, and obtain a binary segmentation result of vegetation;

[0043] A merging module, used for processing the binary segmentation result of the vegetation based on the region merging method of graph theory and the minimum spanning tree algorithm to obtain a merged vegetation region;

[0044] An estimation calculation module, used to calculate the vegetation index value of each pixel in the merged vegetation area, generate a fine distribution map of vegetation based on the vegetation index value, and simultaneously calculate an estimated value of vegetation coverage of the merged vegetation area;

[0045] The result generation module is used to combine the vegetation fine distribution map and the vegetation coverage estimation value to generate a final vegetation coverage measurement result.

[0046] Compared with the prior art, the present invention has the following advantages and technical effects:

[0047] The present invention discloses a method for measuring vegetation coverage based on images of a field handheld device. For heterogeneous images taken by a field handheld device, an image registration method of feature matching is used to achieve a spatially consistent overall image. In the process of image fusion, the fusion weight of the overlapping area is adaptively adjusted to ensure a smooth transition. When extracting vegetation, multiple features such as color, texture and shape are comprehensively utilized to construct a random forest classification model. The overlapping parts are removed by a regional merging method based on graph theory, and the final coverage is calculated by combining the area ratio and the vegetation index. To verify the accuracy, high-precision satellite images are used as a reference to perform pixel-level comparison and confusion matrix analysis. The present invention solves the problems of heterogeneity of field handheld device images and low vegetation extraction accuracy, and realizes high-precision measurement of vegetation coverage. Through multi-source data fusion, multi-feature classification and accuracy verification, the reliability and practicability of the measurement results are improved, providing an effective tool for vegetation monitoring and ecological assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0049] Figure 1 The figure is a flow chart of a method for measuring vegetation coverage according to an embodiment of the present invention. DETAILED DESCRIPTION

[0050] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0051] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0052] Embodiment 1

[0053] like Figure 1 As shown, this embodiment provides a method for measuring vegetation coverage, including:

[0054] Step S101, for heterogeneous images taken by handheld devices in the wild, an image registration method based on feature matching is used to extract significant feature points in the image and calculate the similarity between the feature points to determine the correspondence between the images, thereby achieving geometric correction and spatial registration of the images, eliminating the influence of perspective differences and inconsistent resolutions, and obtaining a spatially consistent overall image.

[0055] Step S102, to ensure the accuracy of the overall image, during the image fusion process, the fusion weight is adaptively adjusted according to the distribution of feature points in the overlapping area of ​​the images, and the grayscale values ​​of the pixels in the overlapping area are weighted averaged to ensure that the fused image has a smooth transition in the overlapping area, avoid obvious stitching marks, and improve the visual quality and measurement accuracy of the overall image.

[0056] Step S103, when extracting green vegetation information, a multi-feature fusion model for vegetation recognition is constructed by comprehensively utilizing multiple features such as color, texture, and shape. Green channel information is extracted through color space conversion, the texture features of the image are calculated, and the shape features of the vegetation are extracted through morphological operations. These features are input into a random forest classifier for training and prediction to obtain a binary segmentation result of the vegetation.

[0057] Step S104, in order to remove the overlapping parts in the vegetation extraction results, a region merging method based on graph theory is used. The vegetation segmentation results are converted into an undirected weighted graph, and the regions with high similarity are gradually merged through the minimum spanning tree algorithm until all regions are merged into a whole. In the merging process, by setting the similarity threshold, the merging granularity is controlled to avoid over-segmentation and under-segmentation.

[0058] Step S105, for the merged vegetation area, calculate the proportion of its area to the entire image to obtain a preliminary vegetation coverage estimation value. To further improve the reliability of the estimation, the multispectral information of the overall image is used to calculate the vegetation index value of each pixel, and the vegetation index map is binarized according to the set threshold to obtain a detailed distribution map of vegetation. The area estimation value is combined with the vegetation index distribution map, and the final vegetation coverage measurement result is obtained by weighted averaging.

[0059] Step S106, in order to verify the accuracy of the vegetation coverage measurement results, high-precision satellite remote sensing images are used as reference data. The satellite images are geometrically corrected and registered to make them consistent with the field handheld device images in space. Then, the satellite images are classified for vegetation to obtain the reference vegetation coverage true value. The field handheld device measurement results are compared with the reference true value at the pixel level, the confusion matrix is ​​calculated, and the accuracy of the measurement results is evaluated, including overall accuracy, user accuracy, mapping accuracy and other indicators, to ensure the reliability and practicality of the field handheld device vegetation coverage measurement method.

[0060] Specifically:

[0061] Install a depth camera on a handheld device in the field, and obtain a depth heterogeneous image of the area to be tested based on the depth camera;

[0062] Preprocessing the depth heterogeneous images to obtain a spatially consistent overall image, and fusing the spatially consistent overall image to obtain a fused data set;

[0063] Furthermore, the process of preprocessing the depth heterogeneous image to obtain a spatially consistent overall image includes:

[0064] The significant feature points in the depth heterogeneous image are extracted, the similarity between the significant feature points is calculated, and the depth heterogeneous image is geometrically corrected and spatially registered based on the similarity to obtain a spatially consistent overall image.

[0065] Furthermore, the process of fusing the spatially consistent whole images to obtain a fused data set includes:

[0066] Extracting feature points of the spatially consistent overall image and calculating the distribution density of the feature points;

[0067] Determine the image overlapping area based on the distribution density of the feature points, and calculate the initial fusion weight of each pixel in the overlapping area;

[0068] For each pixel in the overlapping area, obtain the grayscale value of its corresponding position in the image to be fused, and perform weighted average on the grayscale values ​​according to the initial fusion weight to obtain the fused grayscale value of the pixel;

[0069] The grayscale values ​​of the fused pixels in the overlapping area are spliced ​​with the original grayscale values ​​of the pixels in the non-overlapping area to obtain a preliminary fused image;

[0070] Performing image enhancement processing on the preliminary fused image to generate the fused data set.

[0071] Specifically, aerial photography with handheld devices in the field can obtain a wide range of surface information, but the viewing angle of a single image is limited. In order to obtain a complete regional image, it is necessary to stitch together multiple images taken by handheld devices in the field. Due to differences in shooting angle, height, lighting, etc., these images have problems such as different viewing angles, inconsistent resolutions, and noise, and a series of preprocessing and registration steps are required. First, preprocess each image. For example, there is some random noise in an image, which affects the clarity of the image. Gaussian filtering can be used to remove noise. Suppose the grayscale value of a pixel in the image is 100, and the grayscale values ​​of its surrounding pixels are 90, 110, 95, and 105 respectively. After Gaussian filtering, the grayscale value of the pixel will become the weighted average of these values, and the weight is determined by the Gaussian function, thereby smoothing the image and reducing noise. In addition, the image may be underexposed, resulting in low contrast. Histogram equalization can be used to enhance the image contrast and make the image details clearer. This is like adjusting the brightness and contrast of a picture to make the image more "good-looking". The preprocessed image has higher quality, which is conducive to subsequent feature extraction. Next, extract salient feature points from the preprocessed image. For example, the SIFT algorithm is used to extract feature points. The SIFT algorithm identifies scale-invariant feature points in the image, such as corner points and edge points. These points can still be identified even after the image is scaled or rotated.

[0072] Get multiple images to be fused, such as multiple photos of the same area taken at different locations by a handheld device in the field, where there are overlapping parts between these photos. Extract the feature points of each image. The SIFT algorithm can be used to extract scale-invariant feature points in the image, such as corner points and edge points. Calculate the distribution density of feature points to count the number of feature points per unit area, such as how many feature points there are in each square pixel. Areas with high density of feature points are usually areas with rich textures in the image. Determine the image overlap area based on the distribution density of feature points. Areas with high density of feature points often correspond to overlapping areas of images. For example, areas in two images where feature points are dense can be considered as overlapping areas of the two images. Calculate the initial fusion weight of each pixel in the overlapping area. The initial weight can be determined based on the distance from the pixel to the boundary of the overlapping area. The closer the distance to the boundary, the smaller the weight, and the farther the distance to the boundary, the larger the weight. For example, the inverse distance function can be used to calculate the weight. The weight of pixels close to the boundary is close to 0, and the weight of pixels far from the boundary is close to 1. For each pixel in the overlapping area, obtain its grayscale value at the corresponding position in the image to be fused. For example, the grayscale value of a pixel in the first image is 100, and the grayscale value of the corresponding position in the second image is 150. The grayscale value is weighted averaged according to the initial fusion weight to obtain the fusion grayscale value of the pixel. Assuming that the initial fusion weight of the pixel is 0.3 in the first image and 0.7 in the second image, the fusion grayscale value of the pixel is 0.3×100+0.7×150=135. Determine whether the difference between the fusion grayscale values ​​of the current pixel and the adjacent pixel exceeds the preset smoothing threshold. For example, the preset smoothing threshold is 10. If the fusion grayscale value of the current pixel is 135 and the fusion grayscale value of the adjacent pixel is 150, the difference is 15, which exceeds the smoothing threshold. If it exceeds, adjust the fusion weights of the current pixel and the adjacent pixel until the smooth transition condition is met. For example, the weight of the current pixel can be adjusted to 0.4, the weight of the adjacent pixel can be adjusted to 0.6, and the fusion grayscale value can be recalculated until the difference is less than or equal to 10. This is done to eliminate splicing marks and make the fused image look more natural and smooth. Repeat the weighted average and smoothing threshold determination steps until the fused grayscale values ​​of all pixels in the overlapping area meet the smooth transition conditions and the stitching marks are eliminated. This is like finely polishing the image to ensure that the transition of the overlapping area is smooth and natural. The fused pixel grayscale values ​​in the overlapping area are spliced ​​with the original pixel grayscale values ​​in the non-overlapping area to obtain a preliminary fused image. The pixel grayscale values ​​in the non-overlapping area retain the original values ​​directly. For example, the left part of the first image is a non-overlapping area, so its original pixel grayscale values ​​are directly retained. Perform image enhancement processing on the preliminary fused image to improve the overall visual quality of the image and output the final high-precision fused image.

[0073] Building a multi-feature fusion model for vegetation identification based on a random forest classifier, inputting the fusion data set into the multi-feature fusion model for vegetation identification for calculation, and obtaining a binary segmentation result of vegetation;

[0074] Furthermore, the process of obtaining the binary segmentation result of vegetation includes:

[0075] After color conversion processing is performed on the fused data set, information is extracted to obtain color features, texture features and shape features;

[0076] Normalizing the color features, texture features, and shape features to obtain a multi-feature fusion vector;

[0077] Constructing a random forest classifier, training the random forest classifier based on a historical data set, and obtaining a multi-feature fusion model for vegetation recognition;

[0078] The multi-feature fusion vector is input into the multi-feature fusion model of vegetation recognition for calculation to obtain the binary segmentation result of the vegetation.

[0079] Furthermore, the process of extracting information after color conversion processing is performed on the fused data set to obtain color features, texture features and shape features includes:

[0080] After converting the fused data set from RGB space to HSV space, extracting green channel information, and obtaining color features of green vegetation;

[0081] Calculating the texture features of the fused data set based on the gray level co-occurrence matrix method to obtain the texture features of the vegetation area;

[0082] According to the shape characteristics of the vegetation area, the shape features of the vegetation are extracted.

[0083] By converting the input color image from RGB space to HSV space, the green channel information is extracted to obtain the color features of green vegetation. The gray-level co-occurrence matrix method is used to calculate the texture features of the image, including texture parameters such as contrast, correlation, energy and entropy, to obtain the texture features of the vegetation area. According to the shape characteristics of the vegetation area, morphological operations such as opening and closing operations are designed to extract the shape features of the vegetation and eliminate noise and small areas. The color features, texture features and shape features are normalized to construct a multi-feature fusion vector as the input feature of the random forest classifier. The random forest classifier is used to train and predict the multi-feature fusion vector, and the classification accuracy and robustness are improved through the ensemble learning method to obtain the binary segmentation result of the vegetation. The binary segmentation results are post-processed, such as morphological filtering, connected domain analysis, etc., to eliminate isolated small areas and noise and obtain the final vegetation segmentation mask. The vegetation segmentation mask is fused with the original image to highlight the vegetation area, realizing the extraction and visualization of green vegetation information.

[0084] Acquire multi-source remote sensing image data containing vegetation areas, and extract a variety of effective vegetation feature vectors according to the characteristics of different data sources. Through methods such as feature selection and feature transformation, the extracted multi-source features are optimized and combined to obtain a fused high-dimensional feature vector representation. The fused feature vector is trained using a random forest classifier, and the generalization ability and robustness of the classifier are improved by integrating multiple decision trees. During the training process, according to the distribution of sample data, the parameters of the random forest are reasonably set, such as the number of decision trees, the minimum number of samples of nodes, the size of feature subsets, etc., to balance classification accuracy and computational efficiency. The trained random forest classifier is used to predict the new remote sensing image, and the feature vector of each pixel is input into the classifier to obtain the probability value of the pixel belonging to vegetation or non-vegetation. According to the predicted vegetation probability map, a threshold is set to binarize it to obtain the binary segmentation result of vegetation.

[0085] Processing the binary segmentation result of the vegetation based on the region merging method of graph theory and the minimum spanning tree algorithm to obtain a merged vegetation region;

[0086] Furthermore, the process of obtaining the merged vegetation area includes:

[0087] The region merging method based on graph theory converts the binary segmentation result of the vegetation into an undirected weighted graph, and gradually merges regions with high similarity through a minimum spanning tree algorithm until all regions are merged into a whole and then output to generate the merged vegetation region.

[0088] Specifically, in the process of extracting green vegetation information, constructing an undirected weighted graph is an important step to achieve vegetation area overlap. First, the vegetation areas in the vegetation extraction result image are converted into nodes in graph theory. Assume that there are multiple independent vegetation areas in the image, and each area is regarded as a node. For example, there are five significant vegetation areas in the image, marked as nodes A, B, C, D and E respectively. The similarity between nodes is used as the weight of the edge to construct an undirected weighted graph. The similarity is calculated based on features such as color, texture and shape. For example, nodes A and B have a high similarity in color features, but a low similarity in texture and shape features. The similarity between nodes is measured by methods such as Euclidean distance or Mahalanobis distance. Assume that the Euclidean distance calculation result shows that the color feature distance between nodes A and B is 2, the texture feature distance is 5, and the shape feature distance is 3. Combining these distances, the similarity between nodes A and B is 4. Calculate the similarity between all nodes to form a node similarity matrix. Each element in the matrix represents the similarity between a pair of nodes.

[0089] According to the similarity matrix, the nodes with the highest similarity are gradually merged through the minimum spanning tree algorithm. Assume that nodes A and B are initially merged, and a new node AB is formed after merging. The node features and similarity matrix are updated. The feature of the new node AB is the weighted average of the features of nodes A and B, and the similarity matrix is ​​also updated accordingly to reflect the similarity between the new node and other nodes. Set the similarity threshold to control the merging granularity. Assume that the threshold is 5. When the similarity between nodes is less than 5, the merging process stops.

[0090] Calculating a vegetation index value for each pixel in the merged vegetation area, generating a detailed distribution map of vegetation based on the vegetation index value, and calculating an estimated value of vegetation coverage of the merged vegetation area;

[0091] Furthermore, the process of generating a detailed distribution map of vegetation based on the vegetation index value and calculating the estimated value of vegetation coverage of the merged vegetation area includes:

[0092] Obtaining the vegetation area of ​​the merged vegetation area based on pixel statistics, calculating the total proportion of the merged vegetation area, and obtaining a preliminary vegetation coverage estimation value based on the total proportion;

[0093] Acquire multispectral information of the overall image, and calculate a vegetation index value for each pixel based on the spectral information;

[0094] Calculating the NDVI value of the vegetation index value of each pixel, and generating a vegetation index distribution map based on the NDVI value;

[0095] The vegetation index distribution map is binarized to obtain a detailed vegetation distribution map.

[0096] For example, the processed park image can be input into the random forest classifier to effectively distinguish between vegetation and non-vegetation areas. Finally, the binary segmentation results output by the classifier are post-processed, such as applying morphological filtering and connected domain analysis to eliminate isolated small areas and noise, and obtain the final vegetation segmentation mask. This mask can be fused with the original image to highlight the vegetation area, thereby achieving effective extraction and visualization of green vegetation information. This method has important application value in urban greening monitoring, agricultural research and other fields, and can provide important information about vegetation coverage and health status.

[0097] The detailed distribution map of vegetation and the estimated value of vegetation coverage are combined to generate a final vegetation coverage measurement result.

[0098] Furthermore, the process of generating the final vegetation coverage measurement result includes:

[0099] Counting the number of pixels with a value of 1 in the fine distribution map of the vegetation, and dividing the result by the total number of pixels in the image, to obtain a fine vegetation coverage rate based on the vegetation index;

[0100] The preliminary vegetation coverage estimation value and the refined vegetation coverage based on the vegetation index are weighted averaged to obtain a final vegetation coverage measurement result.

[0101] The merged vegetation area image is obtained, and the area of ​​the vegetation area is obtained by pixel statistics. The area ratio of the vegetation area to the entire image is calculated to obtain a preliminary vegetation coverage estimate. The multispectral information of the overall image is obtained, and the normalized vegetation index NDVI value is calculated for each pixel position to obtain a vegetation index distribution map. According to the preset NDVI threshold, the vegetation index distribution map is binarized, and pixels less than the threshold are set to 0, and pixels greater than or equal to the threshold are set to 1, to obtain a fine binary distribution map of the overall vegetation. The number of pixels with a value of 1 in the binary distribution map is counted and divided by the total number of pixels in the image to obtain a distribution map based on Refined vegetation coverage of vegetation index; weighted average of preliminary area estimated coverage and refined vegetation index coverage, with the weight coefficient determined according to the reliability of the two methods, to obtain the final vegetation coverage measurement result; using decision tree algorithm, with vegetation index and spectral information as features, and coverage as target, to train regression prediction model for rapid estimation of vegetation coverage in subsequent unknown areas; for continuous multi-phase remote sensing images, extract vegetation coverage of each phase, and obtain dynamic change trend of vegetation coverage through time series analysis, and further analyze its correlation with environmental factors such as climate and topography.

[0102] Assuming that the total number of pixels in an image is 1000×1000, after regional merging, the number of pixels in the vegetation area is 300,000, then the area of ​​the vegetation area accounts for 30% of the entire image, which is the preliminary estimated value of vegetation coverage. This method is simple and intuitive, but it may be affected by image resolution and segmentation accuracy. In order to improve the estimation accuracy of vegetation coverage, the multispectral information of the overall image is further obtained. Multispectral images contain reflectance data of different bands, which can reflect the spectral characteristics of vegetation. For each pixel position, its normalized vegetation index (NDVI) value is calculated. The calculation formula of NDVI is: (NIR-Red) / (NIR+Red), where NIR represents the reflectance of the near-infrared band and Red represents the reflectance of the red light band. Through calculation, an NDVI distribution map can be obtained, and the value of each pixel in the map represents the health of the vegetation at that location. According to the preset NDVI threshold, the NDVI distribution map is binarized. Assuming the threshold is set to 4, the pixel values ​​less than 4 are set to 0, indicating non-vegetation areas; pixel values ​​greater than or equal to 4 are set to 1, indicating vegetation areas. After this processing, a fine binary distribution map of vegetation is obtained. The number of pixels with a value of 1 in the statistical map is assumed to be 350,000, divided by the total number of pixels in the image, 1,000,000, and the fine vegetation coverage based on the vegetation index is 35%. This method utilizes the spectral characteristics of vegetation and can more accurately reflect the actual distribution of vegetation.

[0103] Embodiment 2

[0104] This embodiment provides a vegetation coverage measurement system, including:

[0105] A data acquisition module, used to install a depth camera on a field handheld device, and acquire a depth heterogeneous image of the area to be measured based on the depth camera;

[0106] A data fusion module is used to preprocess the depth heterogeneous images to obtain a spatially consistent overall image, and fuse the spatially consistent overall image to obtain a fused data set;

[0107] A model calculation module is used to construct a multi-feature fusion model for vegetation identification based on a random forest classifier, input the fusion data set into the multi-feature fusion model for vegetation identification for calculation, and obtain a binary segmentation result of vegetation;

[0108] A merging module, used for processing the binary segmentation result of the vegetation based on the region merging method of graph theory and the minimum spanning tree algorithm to obtain a merged vegetation region;

[0109] An estimation calculation module, used to calculate the vegetation index value of each pixel in the merged vegetation area, generate a fine distribution map of vegetation based on the vegetation index value, and simultaneously calculate an estimated value of vegetation coverage of the merged vegetation area;

[0110] The result generation module is used to combine the vegetation fine distribution map and the vegetation coverage estimation value to generate a final vegetation coverage measurement result.

[0111] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for measuring vegetation coverage, characterized in that: The following steps are involved: Install a depth camera on a handheld device in the field, and obtain a depth heterogeneous image of the area to be tested based on the depth camera; Preprocessing the depth heterogeneous images to obtain a spatially consistent overall image, and fusing the spatially consistent overall image to obtain a fused data set; Building a multi-feature fusion model for vegetation identification based on a random forest classifier, inputting the fusion data set into the multi-feature fusion model for vegetation identification for calculation, and obtaining a binary segmentation result of vegetation; Processing the binary segmentation result of the vegetation based on the region merging method of graph theory and the minimum spanning tree algorithm to obtain a merged vegetation region; Calculating a vegetation index value for each pixel in the merged vegetation area, generating a detailed distribution map of vegetation based on the vegetation index value, and calculating an estimated value of vegetation coverage of the merged vegetation area; The detailed distribution map of vegetation and the estimated value of vegetation coverage are combined to generate a final vegetation coverage measurement result.

2. The method for measuring vegetation coverage according to claim 1, characterized in that: The process of preprocessing the depth heterogeneous image to obtain a spatially consistent overall image includes: The significant feature points in the depth heterogeneous image are extracted, the similarity between the significant feature points is calculated, and the depth heterogeneous image is geometrically corrected and spatially registered based on the similarity to obtain a spatially consistent overall image.

3. The method for measuring vegetation coverage according to claim 1, characterized in that: The process of fusing the spatially consistent overall images to obtain a fused data set includes: Extracting feature points of the spatially consistent overall image and calculating the distribution density of the feature points; Determine the image overlapping area based on the distribution density of the feature points, and calculate the initial fusion weight of each pixel in the overlapping area; For each pixel in the overlapping area, obtain the grayscale value of its corresponding position in the image to be fused, and perform weighted average on the grayscale values ​​according to the initial fusion weight to obtain the fused grayscale value of the pixel; The grayscale values ​​of the fused pixels in the overlapping area are spliced ​​with the original grayscale values ​​of the pixels in the non-overlapping area to obtain a preliminary fused image; Performing image enhancement processing on the preliminary fused image to generate the fused data set.

4. The method for measuring vegetation coverage according to claim 1, characterized in that: The process of obtaining the binary segmentation result of vegetation includes: After color conversion processing is performed on the fused data set, information is extracted to obtain color features, texture features and shape features; Normalizing the color features, texture features, and shape features to obtain a multi-feature fusion vector; Constructing a random forest classifier, training the random forest classifier based on a historical data set, and obtaining a multi-feature fusion model for vegetation recognition; The multi-feature fusion vector is input into the multi-feature fusion model of vegetation recognition for calculation to obtain the binary segmentation result of the vegetation.

5. The method for measuring vegetation coverage according to claim 4, characterized in that: The process of extracting information after color conversion processing is performed on the fused data set to obtain color features, texture features and shape features includes: After converting the fused data set from RGB space to HSV space, extracting green channel information, and obtaining color features of green vegetation; Calculating the texture features of the fused data set based on the gray level co-occurrence matrix method to obtain the texture features of the vegetation area; According to the shape characteristics of the vegetation area, the shape features of the vegetation are extracted.

6. The method for measuring vegetation coverage according to claim 1, characterized in that: The process of obtaining the merged vegetation area includes: The region merging method based on graph theory converts the binary segmentation result of the vegetation into an undirected weighted graph, and gradually merges regions with high similarity through a minimum spanning tree algorithm until all regions are merged into a whole and then output to generate the merged vegetation region.

7. The method for measuring vegetation coverage according to claim 1, characterized in that: The process of generating a detailed distribution map of vegetation based on the vegetation index value and calculating the estimated value of vegetation coverage of the merged vegetation area comprises: Obtaining the vegetation area of ​​the merged vegetation area based on pixel statistics, calculating the total proportion of the merged vegetation area, and obtaining a preliminary vegetation coverage estimation value based on the total proportion; Acquire multispectral information of the overall image, and calculate a vegetation index value for each pixel based on the spectral information; Calculating the NDVI value of the vegetation index value of each pixel, and generating a vegetation index distribution map based on the NDVI value; The vegetation index distribution map is binarized to obtain a detailed vegetation distribution map.

8. The method for measuring vegetation coverage according to claim 7, characterized in that: The process of generating the final vegetation coverage measurement result includes: Counting the number of pixels with a value of 1 in the fine distribution map of the vegetation, and dividing the result by the total number of pixels in the image, to obtain a fine vegetation coverage rate based on the vegetation index; The preliminary vegetation coverage estimation value and the refined vegetation coverage based on the vegetation index are weighted averaged to obtain a final vegetation coverage measurement result.

9. A system for measuring vegetation coverage according to any one of claims 1 to 8, characterized in that: include: A data acquisition module, used to install a depth camera on a field handheld device, and acquire a depth heterogeneous image of the area to be measured based on the depth camera; A data fusion module is used to preprocess the depth heterogeneous images to obtain a spatially consistent overall image, and fuse the spatially consistent overall image to obtain a fused data set; A model calculation module is used to construct a multi-feature fusion model for vegetation identification based on a random forest classifier, input the fusion data set into the multi-feature fusion model for vegetation identification for calculation, and obtain a binary segmentation result of vegetation; A merging module, used for processing the binary segmentation result of the vegetation based on the region merging method of graph theory and the minimum spanning tree algorithm to obtain a merged vegetation region; An estimation calculation module, used to calculate the vegetation index value of each pixel in the merged vegetation area, generate a fine distribution map of vegetation based on the vegetation index value, and simultaneously calculate an estimated value of vegetation coverage of the merged vegetation area; The result generation module is used to combine the vegetation fine distribution map and the vegetation coverage estimation value to generate a final vegetation coverage measurement result.

Images