Monitoring and identification method of susceptible stages of pine wood nematode disease based on UAV visible light remote sensing

Abstract

This application provides a method for monitoring and identifying pine wood nematode diseased trees at their susceptible stages based on visible light remote sensing from unmanned aerial vehicles (UAVs). The method specifically relates to the technical field of reading and identifying image data for use in identifying the color of pine wood nematode disease. The method comprises the following steps: performing geometric registration, cropping, and other preprocessing on the acquired UAV visible light remote sensing data; calculating vegetation indices based on three visible light bands: the visible-band difference vegetation index (VDVI) and the excess green index (ExG); converting the visible light image and the corresponding vegetation index image into grayscale images and performing multi-threshold segmentation on the images; extracting discolored standing trees based on an optimal threshold value, and post-processing the extracted results to identify individual discolored standing trees. This method further improves the efficiency and accuracy of identifying pine wood nematode diseased trees based on visible light UAV remote sensing images.

Description

Technical Field

[0001] The present application relates to the technical field of pest and disease identification and monitoring, specifically to the technical field of reading and identifying image data, and more specifically to a method for identifying and monitoring the diseased stage of pine wood nematode-lesioned standing trees based on visible light remote sensing by unmanned aerial vehicle (UAV) technology. Background Art

[0002] Traditional monitoring and forecasting rely on manual surveys, which are not only time-consuming and labor-intensive but also subject to factors such as topography. This makes it difficult to accurately and in real time determine the occurrence, location, scope, and extent of pine wilt disease. This can lead to untimely prevention and control, widespread spread of the disease, and irreparable damage. While satellite remote sensing can monitor large areas, it is limited by factors such as revisit cycles, resolution, and weather, making it unable to meet the needs of short-term, high-precision monitoring of disasters. With the rapid development of remote sensing technology, drone remote sensing can obtain higher spatial resolution images than satellite remote sensing, offering advantages such as high flexibility and real-time performance. It can accurately identify discolored standing trees down to the individual tree, making it a key tool for pine wilt disease monitoring.

[0003] UAVs equipped with different cameras can acquire a variety of data types, such as visible light, multispectral, hyperspectral, thermal infrared, and other drone remote sensing data. Multi-type drone data is of great significance in the monitoring and identification of pine wood nematode diseased trees. Considering the cost of sensors, the difficulty of obtaining data, and the processing speed, drone visible light images are often used in actual business work to quickly identify discolored trees, while drone multispectral and hyperspectral data are mostly used for early monitoring research of pine wood nematode disease. Chen Weimei et al. used a two-level fusion deep learning model of VGG16 and improved YOLOv5 based on drone images to identify pine wood nematode diseased trees, with an identification accuracy of 85.58%. Zhang et al. used drone visible light images based on improved YOLOv5 to carry out pine wood nematode diseased trees monitoring, with an identification accuracy of 92.6%, an improvement of 3.3% over the original model. Su et al.

[20] proposed an aerial image detection model based on PWD-YOLOv8, which has an accuracy of 87.9% in identifying discolored trees. Liu Jincang et al. constructed feature vectors using the red, green, and blue bands and texture features of drones and applied a conditional random field model to classify discolored standing trees, achieving an average accuracy of 88.43%. Pine needles undergo significant color changes after infection with pine wilt disease. Using the spectral characteristics of infected pine trees in imagery, remote sensing monitoring of discolored standing trees can be achieved. Wang Xiaoqin et al. compared vegetation indices constructed based on visible light bands for extracting information about healthy green vegetation. The results showed that the visible-band difference vegetation index (VDVI) had higher extraction accuracy than the normalized red-green difference index (RGNDI) and the excess green index (ExG). Zou Yuzhen et al. compared the accuracy of four vegetation indices constructed from red and green bands (RGRI, GRRI, RGNDI, and ExR) for extracting discolored standing trees, finding that the ExR-based index achieved the highest extraction accuracy. Li Hao et al. used an image segmentation algorithm that combined the super green feature factor (ExG) and the maximum inter-class variance method to better realize the discriminant analysis of the disease degree of diseased pine trees.

[0004] In summary, research on using drone visible light imagery to monitor standing trees with pine wilt disease has made considerable progress. However, current research on drone identification of standing trees with pine wilt disease has two challenges. First, current research focuses on improving model algorithms, but due to a lack of samples, it is difficult to generalize to different regions. Second, in the current research on monitoring and identifying discolored wood, there is still potential for extracting and identifying vegetation indices based on drone visible light imagery containing only red, green, and blue bands. To address these issues, the present invention calculates vegetation indices based on drone visible light imagery and utilizes a threshold segmentation method to achieve operational level identification of discolored wood. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for identifying standing trees infected with pine wood nematode disease color based on visible light remote sensing images from drones, so as to further improve the efficiency and accuracy of identifying standing trees infected with pine wood nematode disease color based on visible light remote sensing data from drones.

[0006] The technical solutions adopted by the present invention are as follows:

[0007] S1. Obtain high-resolution UAV visible light remote sensing images within the detection area;

[0008] S2. Calculate vegetation index (ExG, VDVI) based on three visible light bands;

[0009] Furthermore, the formula for calculating the vegetation index (ExG, VDVI) based on the three visible light bands in step S2 is:

[0010] ;

[0011] ;

[0012] Where, 、 、 Represents the pixel values of the red, green and blue bands respectively.

[0013] S3, multi-threshold segmentation of visible light image and vegetation index;

[0014] Furthermore, step S3 performs threshold segmentation based on the vegetation index and the visible light image, specifically including the following steps:

[0015] S31, converting to grayscale values: converting the visible light image and the corresponding vegetation index image into grayscale images, with pixel values ranging from 0 to 255.

[0016] S32. Threshold segmentation: The image is segmented using the genetic algorithm-modified multi-threshold maximum inter-class variance method (Otsu). To avoid misclassification due to a small number of thresholds, the number of thresholds is set to 10. Threshold segmentation is performed on the visible light and vegetation index images converted to grayscale values in step S31.

[0017] S4. Extraction and post-processing of discolored standing timber.

[0018] Furthermore, in step 4, discolored standing trees are identified and the results are post-processed. The specific steps include:

[0019] Based on the grayscale-converted visible light and vegetation index image threshold segmentation results from step 3, extract vegetation and non-vegetation (e.g., shadows) regions, as well as a preliminary extraction of discolored standing trees. Mask the preliminary extraction raster of discolored standing trees using the vegetation regions obtained through threshold segmentation of the preprocessed drone visible light imagery. Convert the masked results to vectors, calculate the area of each patch, and finally, screen for discolored standing trees based on patch area.

[0020] The vegetation index calculated based on the three visible light bands in step S2 includes the overgreen vegetation index, the normalized green-red difference index, the normalized green-blue difference index, the red-green ratio index and the visible light band difference vegetation index.

[0021] The calculation formula of the overgreen vegetation index is: ;

[0022] Among them, G represents the reflectivity of the green light band, R represents the reflectivity of the red light band, and B represents the reflectivity of the blue light band; this index highlights vegetation information by enhancing the relatively high reflectivity characteristics of green vegetation in the green light band.

[0023] The calculation formula of the normalized green-red difference index is: This index uses the difference between green and red light bands to reflect the growth status and coverage of vegetation. The green light band has a stronger reflection on vegetation, while the red light band is easily absorbed by vegetation. After normalization, the difference between the two can better reflect the characteristics of vegetation.

[0024] The calculation formula of the normalized green-blue difference index is: ; This index is based on the reflectance difference between the green and blue light bands and can be used to analyze the characteristics of vegetation; similar to NGRDI, the difference between the green and blue light bands is normalized to highlight the different reflectance characteristics of vegetation and other land objects in these two bands.

[0025] The formula for calculating the red-green ratio index is: ; This index simply and intuitively reflects the ratio of the reflectivity of the red light band and the green light band; since vegetation has strong absorption in the red light band and strong reflection in the green light band, this ratio can indicate the growth status and coverage of vegetation to a certain extent.

[0026] The calculation formula of the visible light band difference vegetation index is: ; The vegetation index of the three visible light bands is comprehensively considered. By integrating and calculating the information of the three bands, it can more comprehensively reflect the characteristics of vegetation; its value range is the same as that of the normalized vegetation index.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] The present invention identifies standing trees discolored by pine wood nematode disease based on visible light remote sensing images from unmanned aerial vehicles (UAVs). Using high-resolution UAV visible light data, the present invention proposes a method for rapidly extracting discolored standing trees using threshold segmentation based on vegetation index without constructing a sample dataset, achieving good recognition results.

[0029] Compared with the traditional regular inspection and monitoring of pine wood nematode disease by monitors, the present invention can discover discolored standing trees hidden in the forest, saving a lot of manpower and material resources. At the same time, the survey area covers a larger area, greatly improving work efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 The present invention shows a flow chart of the method for identifying standing trees with pine wood nematode disease color based on UAV visible light remote sensing images.

[0031] Figure 2 A schematic diagram of the study area of Example 2 is shown.

[0032] Figure 3 The figure shows the statistical graph of pixel values of different ground objects in the three bands of visible light.

[0033] Figure 4 The statistical graphs of five vegetation indices of different land features converted into grayscale values are shown.

[0034] Figure 5 The results of partial image threshold segmentation are shown.

[0035] Figure 6 The figure shows the extraction results of discolored standing trees based on different vegetation indices according to the present invention.

[0036] Figure 7 The comparison of color-changing wood extraction results based on five vegetation indices in a local area is shown. DETAILED DESCRIPTION

[0037] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0038] Example 1

[0039] like Figure 1 As shown, this embodiment discloses a method for remote sensing monitoring and identification of standing trees with pine wood nematode disease color, comprising the following steps:

[0040] S1. Obtain high-resolution UAV visible light remote sensing images within the detection area;

[0041] S2. Calculate vegetation index (ExG, VDVI) based on three visible light bands;

[0042] Furthermore, the formula for calculating the vegetation index (ExG, VDVI) based on the three visible light bands in step S2 is:

[0043] ;

[0044] ;

[0045] Where, 、 、 Represents the pixel values of the red, green and blue bands respectively.

[0046] S3, multi-threshold segmentation of visible light image and vegetation index;

[0047] Furthermore, step S3 performs threshold segmentation based on the vegetation index and the visible light image, specifically including the following steps:

[0048] S31, converting to grayscale values: converting the visible light image and the corresponding vegetation index image into grayscale images, with pixel values ranging from 0 to 255.

[0049] S32, threshold segmentation: The image is threshold segmented using the multi-threshold Otsu algorithm improved by the genetic algorithm. To avoid misclassification due to a small number of thresholds, the number of thresholds for segmentation is set to 10. Threshold segmentation is performed on the visible light and vegetation index images converted to grayscale values in step S31.

[0050] S4. Extraction and post-processing of discolored standing timber.

[0051] Furthermore, in step 4, discolored standing trees are identified and the results are post-processed. The specific steps include:

[0052] Based on the grayscale-converted visible light and vegetation index image threshold segmentation results from step 3, extract vegetation and non-vegetation (e.g., shadows) regions, as well as a preliminary extraction of discolored standing trees. Mask the preliminary extraction raster of discolored standing trees using the vegetation regions obtained through threshold segmentation of the preprocessed drone visible light imagery. Convert the masked results to vectors, calculate the area of each patch, and finally, screen for discolored standing trees based on patch area.

[0053] The vegetation index calculated based on the three visible light bands in step S2 includes the overgreen vegetation index, the normalized green-red difference index, the normalized green-blue difference index, the red-green ratio index and the visible light band difference vegetation index.

[0054] The calculation formula of the overgreen vegetation index is: ;

[0055] Among them, G represents the reflectivity of the green light band, R represents the reflectivity of the red light band, and B represents the reflectivity of the blue light band; this index highlights vegetation information by enhancing the relatively high reflectivity characteristics of green vegetation in the green light band.

[0056] The calculation formula of the normalized green-red difference index is: This index uses the difference between green and red light bands to reflect the growth status and coverage of vegetation. The green light band has a stronger reflection on vegetation, while the red light band is easily absorbed by vegetation. After normalization, the difference between the two can better reflect the characteristics of vegetation.

[0057] The calculation formula of the normalized green-blue difference index is: ; This index is based on the reflectance difference between the green and blue light bands and can be used to analyze the characteristics of vegetation; similar to NGRDI, the difference between the green and blue light bands is normalized to highlight the different reflectance characteristics of vegetation and other land objects in these two bands.

[0058] The formula for calculating the red-green ratio index is: ; This index simply and intuitively reflects the ratio of the reflectivity of the red light band and the green light band; since vegetation has strong absorption in the red light band and strong reflection in the green light band, this ratio can indicate the growth status and coverage of vegetation to a certain extent.

[0059] The calculation formula of the visible light band difference vegetation index is: ; The vegetation index of the three visible light bands is comprehensively considered. By integrating and calculating the information of the three bands, it can more comprehensively reflect the characteristics of vegetation; its value range is the same as that of the normalized vegetation index.

[0060] Example 2

[0061] This case study takes Fushun City, Liaoning Province as the research area.

[0062] Fushun City is located in the mid-latitudes (E124°13′06″, N41°56′15″). The study area has a temperate continental monsoon climate, characterized by high temperatures and abundant rainfall in summer and drought and little rainfall in winter. The average annual temperature is 6.8°C, and the average annual precipitation is 789.5 mm. The study area is a plain with a generally flat terrain and minimal undulations. The study area is rich in forest resources, primarily planted Korean pine (Pinus koraiensis Siebold & Zuccarini), Chinese pine (Pinustabuliformis Carriere), and Chinese larch (Larix gmelinii var. principis-rupprechtii (Mayr) Pilger). The main soil types include brown soil, dark brown soil, and meadow soil. Due to incomplete treatment of infested trees, the presence of host plants (pine trees) and host insects (pine alternating beetles), and a suitable living environment, pine wilt disease has not been completely eradicated in Fushun City.

[0063] In this example, a typical area of Hada Forest Farm was selected for drone data collection and manual verification.

[0064] The experimental case of this embodiment carried out a field observation experiment on pine wood nematode disease in the study area on July 18, 2023.

[0065] The drone visible light images used in this example were captured using a DJI M300 drone equipped with a Zenmuse H20T camera. The visible light remote sensing images were collected between 10:00 AM and 2:00 PM under clear, windless conditions. The experiment ultimately achieved a spatial resolution of 4 cm, which meets the monitoring requirements for pixel-level metric analysis.

[0066] Verification of the captured remote sensing images revealed that they adequately covered the actual survey area. This process ultimately yielded usable drone-derived visible light remote sensing images for identifying discolored standing trees.

[0067] The first step of the present invention is data collection. On July 18, 2023, a visible light image of the Hada Forest Farm area was obtained by drone. The original image data was pre-processed by splicing, geometric registration, cropping, etc., and finally a visible light remote sensing image of the drone with red, green and blue bands was obtained ( Figure 1 ), with a spatial resolution of 4 cm. Background remote sensing imagery from the two months prior to the control experiment confirmed that any discolored standing trees in the area at that time were newly infected with pine wilt disease that year. Visual interpretation and field surveys confirmed the true discolored standing trees in the study area.

[0068] The second step of the present invention is to calculate the vegetation index based on the three bands of visible light: (1) spectral feature analysis; (2) vegetation index calculation. First, based on the visible light image of the drone, samples are selected with discolored standing trees, healthy pine trees, grassland, and bare soil as research objects, and the pixel values of the three bands of visible light corresponding to each pixel are counted, such as Figure 3 Based on the spectral characteristics of the ground objects and the formulas of relevant vegetation indices, the normalized red-green difference index (RGNDI), excess red index (ExR), vegetation atmospherically resistant index (VARI), visible-band difference vegetation index (VDVI), and excess green index (ExG) were calculated.

[0069]

[0070] The third step of the present invention is to screen the monitoring indicators and determine the classification threshold: threshold segmentation is performed based on the vegetation index and the visible light image: (1) conversion to grayscale values; (2) vegetation index analysis of the grayscale values; (3) threshold segmentation. The conversion to grayscale values refers to converting the visible light image and the corresponding vegetation index image into grayscale images with pixel values ranging from 0 to 255. Figure 4 The grayscale values of the vegetation indices corresponding to sample pixels for discolored standing trees, healthy pine trees, grassland, and bare soil were calculated after conversion to grayscale images. The five vegetation indices effectively distinguished discolored standing trees from the other three types of ground features. The RGNDI, ExR, and VARI values of discolored standing trees were significantly higher than those of the other three types of ground features, while the ExG and VDVI values were significantly lower. The image was threshold segmented using the genetic algorithm-modified multi-threshold Otsu method. To avoid misclassification due to a small number of thresholds, the number of thresholds was set to 10. To extract discolored standing trees and vegetation and non-vegetated (e.g., shadow) areas using vegetation index threshold segmentation, the genetic algorithm-modified Otsu method was used to perform threshold segmentation on the grayscale images of the five vegetation indices and the grayscale images converted from the preprocessed drone visible light imagery.

[0071] The fourth step of the present invention is to extract and post-process discolored standing trees, and perform accuracy evaluation. According to the multi-threshold segmentation results, combined with the visible light image of the drone, the optimal thresholds for extracting discolored standing trees after the five vegetation indices RGNDI, ExR, VARI, VDVI, and ExG are converted into grayscale values within the study area are: 182, 144, 242, 39, and 82, respectively. Combined with the formula and meaning of each vegetation index, pixels above the optimal threshold for VARI, RGNDI, and ExR are selected to extract discolored standing trees, and pixels below the optimal threshold for VDVI and ExG are selected to extract discolored standing trees. Pixels above the optimal threshold of 92 of the pre-processed drone visible light image converted to grayscale are vegetation areas, and the discolored standing trees initially extracted using the vegetation index are subsequently masked based on this. The multi-threshold segmentation results and extraction results of ExR and ExG converted to grayscale values and the pre-processed drone visible light image are shown as follows. Figure 5 shown.

[0072] The vegetation area obtained by threshold segmentation of pre-processed drone visible light images was used to mask the initial extraction raster map of discolored standing trees. The masked result was converted into a vector, and the area of each patch was calculated. Finally, the discolored standing trees were screened based on the patch area. The final discolored standing tree patches in the study area were obtained through threshold segmentation and post-processing ( Figure 6 ).

[0073] This paper uses three statistical metrics: precision, recall, and F1 score. Precision indicates the accuracy of positive examples in the prediction results; recall indicates the proportion of positive examples that are correctly predicted; and F1 score is the harmonic mean of precision and recall. The calculation formulas for each metric are as follows.

[0074] (1);

[0075] (2);

[0076] (3);

[0077] Where P represents positive samples, N represents negative samples, T represents correct predictions, and F represents incorrect predictions. TP is the true positive class, which refers to the number of trees identified as color-changing trees that are actually color-changing trees; FN is the false negative class, which refers to the number of trees identified as non-color-changing trees that are actually color-changing trees. Both of them represent actual positive samples; FP is the false positive class, which refers to the number of trees identified as color-changing trees that are actually not color-changing trees; TN is the true negative class, which refers to the number of trees identified as non-color-changing trees that are actually non-color-changing trees. Both of them represent actual negative samples.

[0078] VARI and RGNDI both achieved precision exceeding 90%, but their recall and F1 values were relatively low. VDVI and ExG performed better overall, extracting more complete patches of discolored standing trees. Their precisions exceeded 85%, at 88.14% and 87.70%, respectively, and their recalls exceeded 80%, at 80.00% and 82.31%, respectively. ExG achieved the highest F1 value, at 84.92%. In this study area, however, ExR performed poorly, with neither a recall nor an F1 value exceeding 70%. Overall, within the study area, where trees are sparsely forested and grasses are abundant, VDVI and ExG performed better for identifying discolored standing trees.

[0079] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above and that the invention can be embodied in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered in all respects as illustrative and non-restrictive, and the scope of the invention is defined by the appended claims, not the foregoing description, and all variations within the meaning and range of equivalents of the claims are intended to be included therein. Any reference sign in a claim should not be construed as limiting the claim to which it relates.

Claims

1. A method for monitoring and identifying the susceptible stages of pine wood nematode disease based on UAV visible light remote sensing, characterized by: The following steps are involved: S1. Obtain high-resolution UAV visible light remote sensing images within the detection area; S2. Calculate vegetation indices based on three visible light bands: visible light band difference vegetation index and super green index; S3, multi-threshold segmentation of visible light image and vegetation index; S4, extraction and post-processing of discolored standing wood; Step S2 is to calculate the vegetation index based on the three bands of visible light: (1) spectral feature analysis; (2) vegetation index calculation. First, based on the visible light image of the drone, samples are selected with discolored standing trees, healthy pine trees, grasslands, and bare soil as research objects, and the pixel values of the three bands of visible light corresponding to each pixel are counted. According to the spectral characteristics of the ground objects and the formula of the relevant vegetation index, the normalized red-green difference index, the over-red index, the visible light atmospheric impedance index, the visible light band difference vegetation index, and the super-green index are calculated; the formula for calculating the vegetation index based on the three bands of visible light in step S2 is: ExG=2ρ green -r red -r blue Where, ρ red , ρ green , ρ blue Represents the pixel values of red, green and blue bands respectively, ExG represents the super green index, and VDVI represents the visible light band difference vegetation index; Or, the calculation formula of the visible light band difference vegetation index is: The vegetation index of the three visible light bands is comprehensively considered. By integrating and calculating the information of the three bands, it can more comprehensively reflect the characteristics of vegetation. Its value range is the same as that of the normalized vegetation index. In step S3, threshold segmentation is performed based on the vegetation index and the visible light image, which specifically includes the following steps: S31, converting to grayscale values: converting the visible light image and the corresponding vegetation index image into grayscale images with pixel values ranging from 0 to 255; S32, threshold segmentation: The image is threshold segmented using a multi-threshold maximum inter-class variance method improved by a genetic algorithm. To avoid misclassification due to a small number of thresholds, the number of segmentation thresholds is set to 10. The visible light and vegetation index images converted to grayscale values in step S31 are threshold segmented. Step S3 is the screening of monitoring indicators and the determination of classification thresholds: threshold segmentation is performed based on vegetation index and visible light image: (1) conversion to grayscale value; (2) vegetation index analysis of grayscale value; (3) threshold segmentation, wherein conversion to grayscale value refers to converting visible light image and corresponding vegetation index image into grayscale image, with pixel value range of 0 to 255, and counting the grayscale values of vegetation index corresponding to sample pixels of discolored standing trees, healthy pine trees, grassland, and bare soil after conversion to grayscale image. The five vegetation indices can better distinguish discolored standing trees from the other three types of ground features, among which R of discolored standing trees is 0. The GNDI, ExR, and VARI values are significantly higher than those of the other three types of ground objects, while the ExG and VDVI values are significantly lower than those of the other three types of ground objects. The image is threshold segmented using the multi-threshold Otsu improved by the genetic algorithm. To avoid misclassification due to a small number of thresholds, the number of segmentation thresholds is set to 10. In order to extract discolored standing trees and vegetation and non-vegetation (shadow, etc.) areas based on vegetation index threshold segmentation, the Otsu improved by the genetic algorithm is used to perform threshold segmentation on the images after the five vegetation indices are converted to grayscale values and the pre-processed UAV visible light images are converted to grayscale values. In step 4, discolored standing trees are identified and the results are post-processed. The specific steps include: Based on the threshold segmentation results of the visible light and vegetation index images converted to grayscale values obtained in step 3, the vegetation and non-vegetation area extraction results and the preliminary extraction results of discolored standing trees are obtained respectively; the vegetation area obtained by threshold segmentation based on the preprocessed UAV visible light image is used to mask the preliminary extraction raster map of discolored standing trees, and the masked results are converted into vectors. The area of each image patch is calculated, and finally, the discolored standing trees are screened according to the image patch area. Three statistical indicators are selected: precision, recall, and F1 score. Precision indicates the accuracy of positive samples in the prediction results; recall indicates the proportion of positive samples that are correctly predicted; and F1 score is the harmonic mean of precision and recall. The calculation formulas for each indicator are as follows: Where P represents positive samples, N represents negative samples, T represents correct predictions, and F represents incorrect predictions. TP is the true positive class, which refers to the number of trees identified as color-changing trees that are actually color-changing trees; FN is the false negative class, which refers to the number of trees identified as non-color-changing trees that are actually color-changing trees. Both of them represent actual positive samples; FP is the false positive class, which refers to the number of trees identified as color-changing trees that are actually not color-changing trees; TN is the true negative class, which refers to the number of trees identified as non-color-changing trees that are actually non-color-changing trees. Both of them represent actual negative samples.

2. The method for monitoring and identifying the susceptible stage of pine wood nematode disease based on UAV visible light remote sensing according to claim 1, characterized in that: The vegetation index calculated based on the three visible light bands in step S2 includes the overgreen vegetation index, the normalized green-red difference index, the normalized green-blue difference index, the red-green ratio index and the visible light band difference vegetation index.

3. The method for monitoring and identifying the susceptible stage of pine wood nematode disease based on UAV visible light remote sensing according to claim 2, characterized in that: The calculation formula of the overgreen vegetation index is: ExG=2G-RB; Among them, G represents the reflectivity of the green light band, R represents the reflectivity of the red light band, and B represents the reflectivity of the blue light band; this index highlights vegetation information by enhancing the relatively high reflectivity characteristics of green vegetation in the green light band.

4. The method for monitoring and identifying the susceptible stage of pine wood nematode disease based on UAV visible light remote sensing according to claim 2, characterized in that: The calculation formula of the normalized green-red difference index is: This index uses the difference between green and red light bands to reflect the growth status and coverage of vegetation. The green light band has a stronger reflection on vegetation, while the red light band is easily absorbed by vegetation. After normalization, the difference between the two can better reflect the characteristics of vegetation.

5. The method for monitoring and identifying the susceptible stage of pine wood nematode disease based on UAV visible light remote sensing according to claim 2, characterized in that: The calculation formula of the normalized green-blue difference index is: This index is based on the reflectance difference between the green and blue light bands and can be used to analyze the characteristics of vegetation. Similar to the NGRDI, it normalizes the difference between the green and blue light bands to highlight the different reflectance characteristics of vegetation and other land objects in these two bands.

6. The method for monitoring and identifying the susceptible stage of pine wood nematode disease based on UAV visible light remote sensing according to claim 2, characterized in that: The red-green ratio index calculation formula is: This index simply and intuitively reflects the ratio of the reflectivity of the red light band and the green light band. Since vegetation has stronger absorption in the red light band and stronger reflection in the green light band, this ratio can indicate the growth status and coverage of vegetation to a certain extent.

Images