A method for identifying pine trees susceptible to pine wilt disease in the early stages

Abstract

The present invention provides a method for identifying pine trees susceptible to pine wilt disease in the early stages, comprising the following steps: S1, using a spectral camera to capture remote sensing images of pine tree samples at different time periods, processing the spectral images, and obtaining a spectral curve for each pine tree; S2, using the first-order derivative to measure the band characteristics of the spectral curve and screen out three characteristic bands λ1, λ2, and λ3 that are sensitive to pine wilt disease; S3, calculating the Pine Wield Disease Vegetation Degeneration Index (PWDERI) based on the spectral reflectance values ​​of the characteristic bands according to the following formula; S4, using ENVI combined with the Pine Wield Disease Vegetation Degeneration Index (PWDERI) to establish a decision tree analysis model for early pine wilt disease monitoring and discriminate pine tree samples. This method has a simple algorithm, requires a small number of bands, and has high throughput, and can effectively identify early asymptomatic pine trees infected with pine wilt disease. Furthermore, this method is a non-destructive testing method using optical instruments, does not require damage to the morphological structure of the pine trees, and has excellent application and promotion value.

Description

Technical Field

[0001] The invention relates to a method for identifying pine trees susceptible to pine wilt disease in the early stage, belonging to the technical field of early diagnosis of pine wilt disease. Background Art

[0002] Pine wilt disease, also known as pine wilt disease, is a devastating infectious disease of pine trees caused by pine nematodes as the main pathogen, combined with vector insects, host pine trees and environmental factors. Pine wilt disease has attracted worldwide attention due to its high infectiousness and mortality rate. Pine wilt disease originated in North America, but it has spread rapidly in Asia. Since the first discovery of pine wilt disease in Nanjing in 1982, it has spread and expanded in many provinces and cities in my country, destroying a large number of pine forest resources and natural ecological landscapes, causing huge ecological and economic losses, and affecting nearly 60 million hm2 of my country. 2 Pine forest vegetation poses a huge threat.

[0003] The early monitoring methods for pine wilt disease mainly include forest symptom diagnosis method, gum flow method, pathogen nematode identification method, pathogen nematode identification method, etc. However, most of these traditional pine wilt disease monitoring methods require plant protection workers to go to the field for investigation and sampling, which has problems such as heavy labor, time-consuming and low efficiency. In addition, they are highly subjective and one-sided, and are prone to errors, which leads to insufficient rigor in the early monitoring results of pine wilt disease.

[0004] Remote sensing technology offers the advantages of flexibility, high efficiency, and high data accuracy. It can conduct in-depth monitoring in areas with poor transportation conditions and difficult access for both people and vehicles. Remote sensing monitoring methods can quickly obtain images of the study area. Remote sensing monitoring and prediction systems developed by analyzing remote sensing images using the spatial analysis capabilities of geographic information systems (GIS) combined with pest and disease monitoring and prediction models have become a powerful tool for early detection of pine wood nematode disease. However, traditional remote sensing analysis methods utilize screened sensitive bands, directly selecting the spectral reflectance of the characteristic bands and then modeling and analyzing them. This approach has the disadvantage of using only the raw spectral information, without performing operations between spectra to highlight the characteristics of the detected object. Summary of the Invention

[0005] The present invention provides a method for identifying pine trees susceptible to pine wood nematode disease in the early stage, which can effectively solve the above problems.

[0006] The present invention is achieved in that:

[0007] A method for identifying pine trees susceptible to pine wilt disease in the early stage comprises the following steps:

[0008] S1, uses a spectral camera to capture remote sensing images of pine tree samples at different times, processes the spectral images, and obtains the spectral curve of each pine tree;

[0009] S2, using the first-order derivative to measure the band characteristics of the spectral curve, screened out three characteristic bands λ1, λ2 and λ3 that are sensitive to pine wood nematode disease;

[0010] S3. Calculate the pine wilt vegetation decline index (PWDERI) based on the spectral reflectance value of the characteristic band according to the following formula. The calculation formula of PWDERI is as follows:

[0011] Where Pλx represents the reflectivity value at wavelength λx;

[0012] S4. ENVI was used in combination with the Pine Wood Nematode Vegetation Decline Index (PWDERI) to establish a decision tree analysis model for early pine wood nematode disease monitoring and to identify pine tree samples.

[0013] As a further preference, the spectral camera is carried by a drone or a remote sensing satellite.

[0014] As a further preference, the characteristic wavelength band λ1 is 445nm-510nm, λ2 is 644nm-679nm, and λ3 is 700nm-997nm.

[0015] As a further preference, the characteristic wavelength band λ1 is 506 nm, λ2 is 672 nm, and λ3 is 731 nm.

[0016] As a further priority, step S4 specifically includes the following steps:

[0017] S41, open the decision tree analysis in ENVI and import the PWDERI model;

[0018] S42, the optimal band ranges λ1, λ2 and λ3 obtained by analyzing the spectral data are substituted into the decision tree model for classification;

[0019] S43, the first post-classification process was performed using Majority Analysis in ENVI, and the second post-classification process was performed using Clump in ENVI;

[0020] S44, judging the pine tree sample according to the degree of aggregation of the retained pixel points. If the degree of aggregation of the retained pixel points is greater than 50%, the pine tree is judged to be an early-stage diseased tree of pine wilt disease.

[0021] As a further preference, the majority analysis (MajorityAnalysis) is to replace the category of the central pixel with the category of the pixel that occupies a dominant position in the transformation kernel.

[0022] As a further priority, the clustering process (Clump) is to cluster and merge adjacent similar classification areas using mathematical morphological operators, merge the selected classifications into one piece using a dilation operation, and then perform an erosion operation on the classification image using a transformation kernel.

[0023] The beneficial effects of the present invention are:

[0024] The present invention takes hyperspectral images of diseased pine trees at different stages, then solves the spectral curve, uses the first-order derivative algorithm to measure the spectral bands to screen out sensitive characteristic bands, and calculates the vegetation decay index of pine wilt disease based on the characteristic bands. Finally, ENVI is combined with the Pine Wield Vegetation Decline Index (PWDERI) to establish a decision tree analysis model for early pine wilt disease monitoring. This method can quickly and accurately identify early diseased pine trees in hyperspectral images, providing a reliable basis for subsequent pine wilt disease grade detection.

[0025] The method of the present invention uses fewer bands, is easy to calculate, and has a high speed. It can effectively identify early asymptomatic carriers of pine wood nematode disease. At the same time, the method belongs to non-destructive testing using optical instruments and does not require destruction of the morphological structure of pine trees, and has very good application and promotion value. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments. It should be understood that the following drawings only illustrate certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other relevant drawings can be obtained based on these drawings without paying any creative work.

[0027] Figure 1 This is a sample map of Qingkou Town, Minhou County.

[0028] Figure 2 This is the healthy early stage disease reflectance spectrum curve provided by Example 1 of the present invention.

[0029] Figure 3 This is the healthy and early stage disease first-order derivative spectral reflectance curve provided by Example 1 of the present invention.

[0030] Figure 4 This is the accuracy discrimination diagram provided by Example 1 of the present invention.

[0031] Figure 5 This is the classification discrimination diagram provided by Example 1 of the present invention.

[0032] Figure 6 This is a classification diagram of different band combinations provided by Example 1 of the present invention.

[0033] Figure 7 Sample map of Hongwei Town, Minhou County.

[0034] Figure 8 This is a diagram showing the effects of different index processing provided by Example 2 of the present invention.

[0035] Figure 9 This is a diagram of different index processing accuracies provided by Example 2 of the present invention.

[0036] Figure 10 This is the optimal vegetation index selection diagram provided by Example 3 of the present invention.

[0037] Figure 11 This is a multi-spectral band matching selection diagram provided by Example 3 of the present invention. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the invention for which protection is sought, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0039] Example 1

[0040] In 2019, the experiment selected a test plot in the pine wood nematode epidemic area of ​​Masson pine in Qingkou, Minhou. The map is as follows: Figure 1 From August to November of that year, a push-broom hyperspectral imager with a wavelength range of 389.8nm-997.4nm, a spectral resolution of 3.5nm, and 176 spectral bands was used on a drone. By capturing hyperspectral images of wild pine trees infected at different stages and then processing the hyperspectral images, the spectral curves of each healthy pine tree and early diseased pine tree were obtained, as shown in the figure below. Figure 2 As shown. The first-order derivative is used to measure the spectral band characteristics, as shown Figure 3As shown. Three characteristic bands λ1, λ2 and λ3 that are sensitive to pine wilt disease were screened out. Among them, the characteristic band λ1 ranges from 445nm to 510nm, and the wavelength λ1 with the largest difference is the 506nm band; the characteristic band λ2 ranges from 644nm to 679nm, and the wavelength λ2 with the largest difference is the 672nm band; the characteristic band λ3 ranges from 700nm to 997nm, and the wavelength λ3 with the largest difference is 731nm. The Pine Wilt Disease Early Reflectance Index (PWDERI) was calculated based on the characteristic bands. The calculation formula for the early attenuation index is: In, P λx represents the reflectance value at the wavelength of λx. Finally, the Pine Wood Nematode Early Decay Index (PWDERI) was combined with the decision tree analysis model and substituted into the August hyperspectral image for inversion. The discrimination results were then verified using the November image. The verification steps are as follows:

[0041] Step 1: According to the comparison of the hyperspectral image in August 2019 with the hyperspectral image in November 2019, 64 early diseased trees marked by the ROI tool in the test site of the pine wilt disease epidemic area of ​​​​Pinus massoniana in August were used as later verification samples; Step 2: Open the decision tree analysis in ENVI5.3 and bring in the PWDERI model; Step 3: Select the three bands λ1, λ2, and λ3 in the optimal range determined in this embodiment and substitute them into the decision tree model for classification; Step 4: Post-processing of the classification map, the classification image is subjected to 3 Clum clustering and 3 MajorityParameters scatter removal; Step 5: If the early sample point remains above 50% in the post-classification processing Npts (number of points), it is judged as an early diseased sample and a total of 60 early diseased trees are judged; Step 6: The classification image is subjected to accuracy discrimination with the early sample selected in step 1 to obtain the discrimination accuracy. Among them, the highest discrimination rate of λ1=506nm, λ2=672nm and λ3=731nm is 82.71%. Clustering (Clump) uses mathematical morphological operators (erosion and dilation) to cluster and merge adjacent similar classification areas. The selected classifications are then merged together using a dilation operation, and then the classification image is eroded using the transformation kernel. Majority Analysis replaces the category of the central pixel with the pixel category that occupies the majority (the largest number of pixels) in the transformation kernel to remove redundant classification points.

[0042] The specific discrimination accuracy is as follows Figure 4 As shown. The classification discrimination diagram is as follows Figure 5 As shown. The classification diagram of different bands is as follows Figure 6 shown.

[0043] Figure 4 : Band reorganization and accuracy discrimination were performed. The blue light range is 389.8nm-519.9nm, and the optimal discrimination area of ​​the first filter is 444.5nm-503.3nm; the red light range is 631.1nm-693.2nm, and the optimal discrimination area of ​​the second filter is 644.8nm-679.3nm; the red edge area to the near infrared has a greater discrimination value. Through the combination of the optimal discrimination areas, the optimal discrimination area of ​​the third filter is selected.

[0044] Figure 5 : The hyperspectral pine wilt vegetation decline index (PWDERI) decision tree model established based on the optimal band combination can accurately identify early diseased pine trees in August.

[0045] Figure 6 :Different bands are matched with classification diagrams, indicating that in the blue light band range (see Figure 6 Figure a), the best blue light band range is 444.5nm-503.3nm; followed by the red light band range (see Figure b), the best red light band range is 644.8nm-679.3nm; finally, in the near-infrared band region (see Figure c and d), the best effect is selected from the red edge part of 700.2nm-777.8nm, and the best effect is selected from the near-infrared part of 792.1nm-997.4nm; among them, the optimal band combination is 679.3nm-500nm-767.1nm.

[0046] Blue light range 445nm-510nm; red light range 644nm-679nm; near infrared range 700nm-997nm;

[0047] Example 2

[0048] The experiment selected a test site in the pine wood nematode epidemic area of ​​Masson pine in Hongwei Town, Minhou County, the map of which is as follows: Figure 7 As shown. Hyperspectral images of the sample plots were taken from June to November 2021. By taking hyperspectral images of wild pine trees infected at different times and then processing the hyperspectral images, the spectral curves of each healthy pine tree and early diseased pine tree were obtained. The first-order derivative was used to measure the spectral band characteristics, and three characteristic bands λ1, λ2 and λ3 that are sensitive to pine wood nematode disease were screened out. Early diseased trees were inferred based on the dead field images taken later, and the hyperspectral images in June, where the early symptoms first appeared, were processed with PWDERI index, DR index, SIPI index, NDRE index and MSR750 index, and the processing results are shown as follows. Figure 8As shown. Among them, the decision tree model analysis effect of the pine wilt disease vegetation decline index (PWDERI) is the best. Finally, the pine wilt disease early decay index (PWDERI), DR index, SIPI index, NDRE index, MSR750 index are combined with the decision tree analysis to establish a model and substituted into the June hyperspectral image for inversion, and the November image is used to verify the discrimination results. The calculation of the early decay index (PWDERI) is the same as in Example 1. The verification steps are as follows:

[0049] Step 1: Compare the diseased conditions of trees in June and November in the experimental plot by ENVI5.3 software, and use ROI tool to mark out early diseased pine wood in the June plot for later precision discrimination; Step 2: Open the decision tree analysis tool in ENVI5.3 and substitute the decision tree model used in this study; Step 3: Select model bands λ1, λ2, and λ3 according to the optimal range determined by the present embodiment, wherein λ1=506nm, λ2=672nm, and λ3=731nm; Step 4: The classified image is subjected to 3 Clum clustering and 3 MajorityParameters scatter removal; Step 5: After the classification, if Npts (the number of points) remains more than 50%, it is judged to be an early diseased sample, and a total of 21 early diseased trees are distinguished; Step 6: The classified image is subjected to precision discrimination with the early samples selected in step 1, and compared with the late diseased samples, 16 diseased trees are correctly distinguished. Clustering (Clump) and main analysis (MajorityAnalysis) are the same as in Example 1.

[0050] The specific discrimination accuracy is as follows Figure 9 Among them, the PWDERI (Premature Weakness Disease Index) had the highest accuracy.

[0051] from Figure 8 It can be seen from the different index processing effect diagrams that PWDERI processing is more accurate than other processing. Figure 8 It can be seen that the recognition accuracy of PWDERI processing is higher.

[0052] Example 3

[0053] The experiment selected a test site in the epidemic area of ​​pine wood nematode disease in Masson pine in Hongwei Town, Minhou County. From May to November 2021, the sample site was visited to obtain multispectral images. The spectral bands identified by the early hyperspectral were used as the basis for multispectral band selection. The bands of 450nm (B1), 560nm (B2), 650nm (B3), 730nm (B4), 840nm (B5), 505nm (C1), 680nm (C2), and 750nm (C4) were selected as the camera lens configuration. At the same time, dead field images taken in different time periods were used to infer early diseased trees as verification samples. Finally, the decision tree model established by the multispectral band vegetation decline disease index (PWDERI) was carried out to establish a discriminant model based on multispectral images. The calculation of the early decay index (PWDERI) is the same as in Example 1.

[0054] Step 1: Compare the disease status of trees in the experimental plot of Hongwei Town, Minhou County from May to November, and use the ROI annotation tool in ENVI5.3 to mark the early diseased pine trees in the early month plot; Step 2: Open the decision tree analysis tool in ENVI5.3, select a vegetation index suitable for multispectral discrimination, and substitute it into the decision tree model used in this embodiment; Step 3: Perform three Clum clustering and three Majority Parameters scatter point removal on the classified image; Step 4: After the classification, if the Npts (number of points) remains above 50%, it is judged as an early diseased sample; Step 5: Perform accuracy discrimination on the sample in the classified image in step 4 and the early sample selected in step 1 to determine the discrimination accuracy. Clustering (Clump) and main analysis (Majority Analysis) are the same as in Example 1.

[0055] The specific discrimination accuracy is as follows Figure 10 The optimal vegetation index is selected as shown in Figure 11 As shown. Figure 11 It can be seen that the discrimination effect of different band combinations after lens matching is found to be able to play a role in early detection

[0056] In summary, the embodiment of the present invention uses a customized filter based on the characteristic regional spectral band to perform narrow-band detection imaging on the pine trees in the detection area, and uses a decision tree-based classification model for early pine wilt disease to convert the spectral information of the three characteristic bands of the pine trees susceptible to early pine wilt disease into corresponding indices, and screen out the pixels of the diseased pine trees based on the threshold value of the index, and finally identify the early diseased plants based on the enrichment of the pixels. This method has a simple algorithm, requires few bands, and has a high throughput, and can effectively identify early asymptomatic infected pine trees infected with pine wilt disease; at the same time, this method belongs to non-destructive testing with optical instruments, does not require destroying the morphological structure of the pine trees, and has very good application and promotion value.

[0057] The foregoing description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Those skilled in the art will readily appreciate that various modifications and variations are possible. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of the present invention are intended to be within the scope of protection of the present invention.

Claims

1. A method for identifying pine trees susceptible to pine wilt disease in the early stage, characterized in that: The following steps are involved: S1, uses a spectral camera to capture remote sensing images of pine tree samples at different times, processes the spectral images, and obtains the spectral curve of each pine tree; S2, using the first-order derivative to measure the band characteristics of the spectral curve, screened out three characteristic bands λ1, λ2 and λ3 that are sensitive to pine wood nematode disease; the characteristic band λ1 is 506nm, λ2 is 672nm, and λ3 is 731nm; S3. Calculate the pine wilt vegetation decline index (PWDERI) based on the spectral reflectance values ​​of the characteristic bands using the following formula: , where Pλx represents the reflectivity value at wavelength λx; S4, using ENVI combined with the pine wilt disease vegetation decline index PWDERI to establish a decision tree analysis model for early pine wilt disease monitoring and discriminate pine samples; Step S4 specifically includes the following steps: S41, open the decision tree analysis in ENVI and import the PWDERI model; S42, the optimal band ranges λ1, λ2 and λ3 obtained from the spectral data analysis are substituted into the decision tree model for classification; S43, the first post-classification process was performed using Majority Analysis in ENVI, and the second post-classification process was performed using Clump in ENVI; S44, judging the pine tree sample according to the degree of aggregation of the retained pixel points. If the degree of aggregation of the retained pixel points is greater than 50%, it is judged to be an early-stage diseased tree of pine wilt disease.

2. The method for identifying pine trees susceptible to pine wilt disease in the early stage according to claim 1, characterized in that: The spectral camera is carried by a UAV or a remote sensing satellite.

3. The method for identifying pine trees susceptible to pine wilt disease in the early stage according to claim 1, characterized in that: The majority analysis is to replace the category of the central pixel with the category of the pixel that occupies the majority position in the transformation kernel.

4. The method for identifying pine trees susceptible to pine wilt disease in the early stage according to claim 1, characterized in that: The clustering process (Clump) is to cluster and merge adjacent similar classification areas using mathematical morphological operators, merge the selected classifications into one piece using a dilation operation, and then perform an erosion operation on the classification image using a transformation kernel.

Images