An intelligent extraction method for pine wood nematode diseased wood based on UAV hyperspectral imaging

Abstract

The present invention relates to the field of forestry health monitoring technology, specifically to an intelligent extraction method for pine wood nematode diseased wood based on unmanned aerial vehicle hyperspectral imaging, comprising the following steps: according to the target forest location data, considering the forest range and the energy consumption of the unmanned aerial vehicle flight, analyzing and calculating the flight path of the unmanned aerial vehicle, and generating a flight path planning result. In the present invention, by optimizing the flight path of the unmanned aerial vehicle, the efficiency of spectral data collection is improved, and the waste of resources caused by poor path planning is reduced. In combination with the real-time recording of the geographic coordinate information of the image data, the accurate correspondence of the data is guaranteed, and the spatial accuracy of data collection is improved. By analyzing the spectral characteristics of the diseased wood samples and matching and analyzing the hyperspectral data of the target area, the efficiency and accuracy of identifying the diseased wood are improved. The severity rating and the prediction of the disease spread trend strengthen the foresight of forestry management and improve the efficiency and accuracy of forestry health monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of forestry health monitoring, and in particular to an intelligent extraction method of pine wood nematode diseased wood based on unmanned aerial vehicle hyperspectral imaging. Background Art

[0002] The field of forestry health monitoring technology uses a variety of technical means to evaluate and monitor the health of forest ecosystems, including tree growth, the occurrence and spread of diseases, and the impact of environmental factors on forests. It combines remote sensing technology, geographic information systems, ecology, plant pathology, and computer science to provide forest health data from macro to micro levels. By using sensors for hyperspectral and multispectral imaging, it can remotely identify and evaluate the diseases and physiological states of trees and forests, enabling forestry managers to promptly detect tree health problems, formulate forestry management and disease control strategies, and maintain forest resources and biodiversity.

[0003] Among them, the pine wood nematode diseased wood extraction method aims to detect and identify trees infected by pine wood nematodes, capture the spectral data of trees through drones and hyperspectral images, and combine data analysis to distinguish between healthy and infected trees, identify trees affected by pine wood nematode disease, improve the efficiency and accuracy of monitoring, and achieve rapid and accurate detection of infected trees in forestry health monitoring, so that forestry managers can respond quickly and carry out effective disease management and control, reduce the impact of pine wood nematode disease on forest ecosystems, and improve the accuracy of diseased wood detection and the speed and efficiency of dealing with forestry problems.

[0004] Traditional pine wood nematode diseased wood extraction technology often has insufficient data integration in actual operation, resulting in the inability to accurately analyze the health status of forests in real time. Although traditional remote sensing technology can cover a wide area, it cannot accurately identify individual diseased trees or early diseases due to resolution limitations, limiting the timeliness and effectiveness of disease management. Diseased tree identification and disease assessment rely on the experience and judgment of experts, and lack of a rapid response mechanism, making it impossible to take prompt measures in the early stages of a disease outbreak, resulting in the widespread spread of the disease, making it impossible for forestry managers to respond in a timely manner when faced with sudden forest diseases, increasing the loss of forest resources and the risk of biodiversity. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an intelligent extraction method for pine wood nematode diseased wood based on unmanned aerial vehicle hyperspectral imaging.

[0006] In order to achieve the above object, the present invention adopts the following technical scheme, a method for intelligent extraction of pine wood nematode diseased wood based on UAV hyperspectral imaging, comprising the following steps: S1: Based on the target forest location data, take aerial photos of the entire target forest, perform radiation correction and orthorectification processing, and combine the image geographic coordinate information to generate UAV hyperspectral remote sensing orthophotos; S2: Based on the UAV hyperspectral remote sensing orthophoto and ground measured data, the crown spectral curves of pine wood nematode diseased trees with different degrees of damage are extracted, and the characteristics are analyzed to construct a spectral feature library of diseased trees; S3: using the diseased tree spectral feature library, analyzing the hyperspectral data of the target forest area, identifying the diseased trees through spectral feature matching, clustering the diseased tree pixels according to the pixel erosion and expansion algorithm, extracting the diseased tree coordinates, recording the location of the target diseased tree and adding marking information, and generating a diseased tree identification result; S4: Based on the diseased tree identification result, combined with the spectral feature database of diseased trees with different degrees of damage, by analyzing the hyperspectral data of the target diseased tree, the severity of the disease of the diseased tree is evaluated, the degree of impact of the disease is analyzed, and severity rating data is generated; S5: In combination with the severity rating data, by analyzing the geographical location information of multiple diseased trees, the location distribution characteristics of the diseased trees are evaluated, the spread path and speed of the disease are predicted, and the spread trend prediction data is generated.

[0007] As a further solution of the present invention, the flight path planning data includes flight time data, geographic coverage information, and energy consumption calculation data; the geo-tagged hyperspectral data set includes image shooting timestamp, geographic coordinate information, and spectral band information; the diseased tree spectral feature library includes diseased tree spectral standard curve, disease type label, and spectral width range; the diseased tree identification results include diseased tree geographic location information, disease type information, and diseased tree marking information; the severity rating data includes diseased tree health index, disease severity score, and diseased tree growth stage data; the diffusion trend prediction data includes diffusion path prediction results, impact area analysis results, and diffusion speed prediction results.

[0008] As a further solution of the present invention, according to the target forest location data, the entire target forest is aerially photographed, and radiation correction and orthorectification are performed, and combined with the image geographic coordinate information, the steps of generating a drone hyperspectral remote sensing orthophoto are as follows: S101: according to the target forest location data, set the flight path of the UAV and perform full-area aerial photography, obtain aerial photography data, and generate an aerial image set; S102: Based on the aerial image set, performing radiation correction on the image data, adjusting the spectral value of the image according to the actual lighting conditions, eliminating the influence of environmental factors, and obtaining a radiation corrected image; S103: Based on the radiation correction image, ortho correction is performed to eliminate deformation caused by terrain influence, and the image geographic coordinate information is combined to correspond to the actual ground position to generate a UAV hyperspectral remote sensing orthoimage.

[0009] As a further solution of the present invention, based on the UAV hyperspectral remote sensing orthophoto and ground measured data, the crown spectral curves of pine wood nematode diseased trees with different degrees of damage are extracted, and characteristic analysis is performed on them. The steps of constructing a diseased tree spectral feature library are specifically as follows: S201: extracting spectral data of pine wood nematode diseased wood samples with different degrees of damage based on the UAV hyperspectral remote sensing orthophoto combined with ground measured data positioning, and generating a spectral feature set of the diseased wood samples; S202: Based on the spectral feature set of the diseased wood sample, feature analysis is performed on the spectral data to identify features associated with pine wood nematode disease in the spectral bands, and heterogeneous feature band information is screened to generate key spectral feature data; S203: Based on the key spectral feature data, a spectral feature database is constructed to record disease stage information corresponding to a plurality of spectral features, and a diseased tree spectral feature library is generated.

[0010] As a further solution of the present invention, the hyperspectral data of the target forest area is analyzed by using the diseased tree spectral feature library, the diseased trees are identified by spectral feature matching, and the diseased tree pixels are clustered according to the pixel corrosion and expansion algorithm, the diseased tree coordinates are extracted, the location of the target diseased tree is recorded and the marking information is added, and the steps of generating the diseased tree identification result are specifically as follows: S301: Based on the diseased tree spectral feature library, preprocess the hyperspectral data of the target forest area, remove noise and outliers, and generate data processing records; S302: Based on the data processing record, by comparing and matching with the spectral data of the diseased wood sample, calculating the similarity of the spectral data, identifying the pine wood nematode diseased wood, and generating a feature data comparison record; S303: Based on the feature data comparison record, the spatial location information of multiple diseased trees is recorded and marked, and a diseased tree identification result is generated.

[0011] As a further solution of the present invention, the specific formula for calculating the similarity of spectral data is: ; in, represents the cosine similarity value, The first spectral values, represents the corresponding spectral value in the spectral library of known pine nematode diseased wood. represents the total number of spectral data points, is the current index position in the index column.

[0012] As a further solution of the present invention, according to the diseased tree identification result, combined with the spectral feature database of diseased trees with different degrees of damage, by analyzing the hyperspectral data of the target diseased tree, the severity of the diseased tree is evaluated, the degree of impact of the disease is analyzed, and the steps of generating severity rating data are specifically as follows: S401: Based on the diseased tree identification result, the hyperspectral data of the identified diseased tree is analyzed, key spectral parameters including reflectance are recorded, and a diseased tree spectral data set is generated; S402: Analyze the spectral parameters of the diseased wood based on the diseased wood spectral data set, identify the development stage of the target diseased wood by comparing with the spectral data of the sample diseased wood, and generate development stage analysis data; S403: Based on the development stage analysis data, according to the disease development stages of multiple diseased trees, the disease severity score of the target diseased tree is calculated, the disease is classified and rated, and severity rating data is generated.

[0013] As a further solution of the present invention, in combination with the severity rating data, by analyzing the geographical location information of multiple diseased trees, evaluating the location distribution characteristics of the diseased trees, predicting the spread path and speed of the disease, and generating the spread trend prediction data, the specific steps are: S501: Based on the severity rating data, and according to the marks and geographical location information of multiple diseased trees, a diseased tree distribution map is drawn to generate location distribution data; S502: Analyze the spatial relationship and distance between multiple diseased trees based on the location distribution data, identify disease diffusion clusters, and generate disease cluster analysis results; S503: Based on the disease cluster analysis results, identify the impact range of the disease, calculate the diffusion speed, predict the diffusion path, and generate diffusion trend prediction data.

[0014] As a further solution of the present invention, the specific formula for calculating the diffusion rate is: ; in, Represents the speed of disease spread, that is, the distance the disease spreads per unit time. Representative The observed disease distance indicates the distance from the initial point to the current point. Represents the corresponding time point, that is, the specific time when the disease was observed. is the average value of the disease distance of all observation points, is the average time of all observation points, is the observation sequence number, used to mark the observation point. is the total number of observations, indicating the number of data points involved in the calculation.

[0015] Compared with the prior art, the advantages and positive effects of the present invention are: In the present invention, the flight path of the UAV is optimized to improve the efficiency of spectral data collection and reduce resource waste caused by poor path planning. The geographic coordinate information of the real-time recording of image data is combined to ensure the accurate correspondence of the data and improve the spatial accuracy of data collection. The spectral characteristics of the diseased tree samples are analyzed and matched with the hyperspectral data of the target area for analysis, thereby improving the efficiency and accuracy of identifying diseased trees. The severity rating and the prediction of the disease spread trend strengthen the foresight of forestry management and improve the efficiency and accuracy of forestry health monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of the main steps of the present invention; Figure 2 It is a schematic diagram of the refinement of S1 of the present invention; Figure 3 It is a schematic diagram of the refinement of S2 of the present invention; Figure 4 It is a schematic diagram of the refinement of S3 of the present invention; Figure 5 It is a schematic diagram of the refinement of S4 of the present invention; Figure 6 It is a schematic diagram of the refinement of S5 of the present invention. DETAILED DESCRIPTION

[0017] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0018] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, "multiple" means two or more, unless otherwise clearly and specifically defined.

[0019] See also Figure 1 The present invention provides a technical solution, a method for intelligently extracting pine wood nematode diseased wood based on UAV hyperspectral imaging, comprising the following steps: S1: Based on the target forest location data, take aerial photos of the entire target forest, perform radiation correction and orthorectification processing, and combine the image geographic coordinate information to generate UAV hyperspectral remote sensing orthophotos; S2: Based on the UAV hyperspectral remote sensing orthophotos and ground-based measured data, the crown spectral curves of pine wood nematode diseased trees with different degrees of damage were extracted, and their characteristics were analyzed to construct a spectral feature library of diseased trees; S3: Use the diseased tree spectral feature library to analyze the hyperspectral data of the target forest area, identify the diseased trees through spectral feature matching, cluster the diseased tree pixels according to the pixel erosion and expansion algorithm, extract the diseased tree coordinates, record the location of the target diseased tree and add marking information to generate the diseased tree identification result; S4: Based on the results of diseased tree identification, combined with the database of spectral characteristics of diseased trees with different degrees of damage, the severity of the diseased tree is evaluated by analyzing the hyperspectral data of the target diseased tree, the degree of impact of the disease is analyzed, and severity rating data is generated; S5: Combined with the severity rating data, by analyzing the geographic location information of multiple diseased trees, the location distribution characteristics of the diseased trees are evaluated, the spread path and speed of the disease are predicted, and the spread trend prediction data is generated.

[0020] UAV hyperspectral remote sensing orthophotos include hyperspectral image data, image shooting timestamp and geographic coordinate information, and spectral band information. The diseased tree spectral feature library includes diseased tree spectral standard curve, disease type label, and spectral width range. The diseased tree identification results include diseased tree geographic location information, disease severity information, and diseased tree marking information. The severity rating data includes diseased tree health index, disease severity score, and diseased tree growth stage data. The diffusion trend prediction data includes diffusion path prediction results, impact area analysis results, and diffusion speed prediction results.

[0021] See also Figure 2 According to the target forest location data, the whole target forest is photographed, and radiation correction and orthorectification are performed. Combined with the image geographic coordinate information, the steps to generate the UAV hyperspectral remote sensing orthophoto are as follows: S101: according to the target forest location data, set the flight path of the UAV and perform full-area aerial photography, obtain aerial photography data, and generate an aerial image set; In sub-step S101, positioning is performed based on the latitude and longitude data of the target forest. During the positioning process, the GPS positioning system is used to obtain the precise location of the forest. Then, the flight path of the drone is set based on these data. The path planning adopts the waypoint flight mode, which allows preset flight points. The drone will automatically fly along these preset points to ensure coverage of the entire target forest area. During the flight, the high-resolution camera carried by the drone continuously shoots the ground. The shooting interval is optimized according to the flight speed and the shutter speed of the camera to ensure the continuity and coverage of the image. The original aerial data obtained in this process is then stored in the internal storage card of the drone. After the operation is completed, the data is transmitted to the ground station to generate an aerial image set containing geographic coordinates.

[0022] S102: Based on the aerial image set, the image data is radiometrically corrected, the spectral value of the image is adjusted according to the actual lighting conditions, the influence of environmental factors is eliminated, and a radiometrically corrected image is obtained; In sub-step S102, radiation correction software is used to perform radiation correction based on the aerial image set. This correction process includes converting the digital spectral values ​​in the image into the true values ​​of the ground reflectance, using the radiation transfer model and the atmospheric correction algorithm, taking into account the actual lighting conditions, such as the sun's altitude and meteorological data, and adjusting the spectral value of each image pixel to compensate for the image deviation caused by atmospheric scattering and absorption. The correction process takes into account the multi-layer structure of the atmosphere, uses the atmospheric characteristics of the model from the surface to the satellite transmission path, and analyzes the image pixel by pixel. The adjustment result is output as a radiation corrected image, which reflects the true surface color and details.

[0023] S103: Based on the radiometric correction image, orthorectification is performed to eliminate deformation caused by terrain influence, and the image geographic coordinate information is combined to correspond to the actual ground position to generate a UAV hyperspectral remote sensing orthoimage; In sub-step S103, orthorectification is performed based on the radiation correction image, using a digital elevation model and orthorectification software. This process includes using the image's geographic coordinate information to match it with the digital elevation model data, calculating the pixel offset caused by the terrain, and making adjustments. The key to orthorectification is to eliminate the impact of terrain undulations on the image so that it can accurately correspond to the actual position on the ground. The stitching and transition of each image are considered during the correction process to ensure that there are no height difference errors and color distortions at the stitching. The output result is a drone hyperspectral remote sensing orthoimage, which is used in geographic information systems to support accurate map making and ground feature analysis.

[0024] See also Figure 3 Based on the UAV hyperspectral remote sensing orthophotos and ground-based measured data, the crown spectral curves of pine wood nematode diseased trees with different degrees of damage were extracted and their characteristics were analyzed. The specific steps for constructing the diseased wood spectral feature library are as follows: S201: Based on the UAV hyperspectral remote sensing orthophoto combined with ground-based data positioning, the spectral data of pine wood nematode diseased wood samples with different degrees of damage were extracted to generate the spectral feature set of diseased wood samples; In sub-step S201, spectral data based on geocoding is used to collect spectra of wood samples affected by pine wood nematode disease. Hyperspectral imaging technology is used to collect spectral data to provide continuous band coverage and capture spectral changes when wood is affected by disease. The principal component analysis algorithm is used to extract characteristic bands and identify spectral intervals with significant changes. Accurate spectrum capture is performed on each sample to record changes in spectral reflectance caused by physiological changes caused by pine wood nematode disease. The generated spectral feature set of diseased wood samples includes band data and spectral features. The data provides a basis for further feature analysis and disease diagnosis.

[0025] S202: Based on the spectral feature set of the diseased wood samples, the spectral data is analyzed to identify the features associated with pine wood nematode disease in the spectral bands, and the heterogeneous feature band information is screened to generate key spectral feature data; In sub-step S202, feature analysis of spectral data is performed based on the spectral feature set of diseased wood samples. Spectral analysis software is used to automatically distinguish and record spectral bands related to pine wood nematode disease through feature recognition. Machine learning algorithms, including principal component analysis and linear discriminant analysis, are used in the feature analysis process to extract key information from complex data and reduce data redundancy, identify spectral bands that play a key role in characterizing the disease process, and record key band information including wavelength, reflectivity, and correlation between bands and disease development. The generated key spectral feature data lists the target bands and lays the foundation for establishing a comprehensive spectral feature database.

[0026] S203: constructing a spectral feature database based on the key spectral feature data, recording disease stage information corresponding to a plurality of spectral features, and generating a diseased tree spectral feature library; In sub-step S203, a spectral feature database is constructed based on key spectral feature data. The database is constructed and maintained using a database management system, and supports data access and query operations. The database records a variety of spectral features and corresponding pine wood nematode disease stage information. Each feature is associated with a specific disease description, including the early, middle and late stages of the disease. The data is processed by an integrated data analysis module that supports multivariate analysis and pattern recognition, automatically matching new spectral data with existing features to confirm the disease stage. The generated diseased tree spectral feature library provides a tool for on-site disease monitoring and early diagnosis, making disease management more efficient.

[0027] See also Figure 4, using the diseased tree spectral feature library, the hyperspectral data of the target forest area is analyzed, the diseased trees are identified by spectral feature matching, and the diseased tree pixels are clustered according to the pixel erosion and expansion algorithm, the diseased tree coordinates are extracted, the location of the target diseased tree is recorded and the marking information is added. The specific steps for generating the diseased tree identification result are as follows: S301: Based on the diseased tree spectral feature library, preprocess the hyperspectral data of the target forest area, remove noise and outliers, and generate data processing records; In sub-step S301, data preprocessing is performed on the hyperspectral data of the target forest area. Signal processing techniques, including noise filtering and outlier detection, are used. Wavelet transform and outlier removal algorithms are used to remove environmental noise and equipment errors from the data. Wavelet transform identifies and filters noise by decomposing the different frequency components of the signal. The outlier removal algorithm detects and removes significantly different values. The preprocessed data improves the accuracy and reliability of subsequent analysis. The generated data processing records record the operation process and results of each step, including the specific parameter settings for denoising and outlier processing and the comparison of data before and after processing, which provides an important basis for subsequent data analysis and application.

[0028] S302: Based on the data processing record, by comparing and matching the spectral data of the diseased wood sample, calculating the similarity of the spectral data, identifying the pine wood nematode diseased wood, and generating a feature data comparison record; The specific formula for calculating the similarity of spectral data is: ; in, represents the cosine similarity value, The first spectral values, represents the corresponding reflectance value in the spectral library of known pine nematode-infected wood. represents the total number of spectral data points, is the current index position in the index column.

[0029] formula: ; Detailed explanation of the formula and the process of formula calculation and derivation: The formula is used to calculate the cosine similarity between the spectral data of the target tree and the spectral data of the known pine wood nematode diseased tree. The obtained similarity value is used to evaluate the matching degree and identify the diseased tree; Parameter meaning and setting value: is the total number of spectral data points, assuming ; The target trees are The average pixel value of the band, assuming there are four data points, 0.1, 0.15, 0.25, and 0.35, reflects the spectral reflectance of the tree at a specific wavelength; Representative of known pine wilt diseased trees The spectral intensity of the points, assuming the corresponding data are 0.13, 0.17, 0.3, and 0.4, represents the spectral reflectance of the standard infected wood; Substitute the parameters into the formula for calculation: ; Compute the sum of dot products: ; Calculate the denominator : ; Calculate the denominator : ; Calculate cosine similarity: ; result The spectral data of the target trees are very close to the known spectral data of pine wood nematode diseased trees, indicating that the target trees may be infected by pine wood nematodes. The similarity value provides a clear indicator for forestry managers to help take management and control measures, limit the spread of the disease and adopt effective control strategies.

[0030] S303: Based on the feature data comparison record, the spatial location information of multiple diseased trees is recorded and marked, and a diseased tree identification result is generated; In sub-step S303, based on the feature data comparison records, the spatial location information of multiple diseased trees is recorded and marked, the location of the diseased trees is marked and stored using geographic information system technology, and the spectral identification results are accurately associated with the geographic location. The location information of each diseased tree is recorded using GPS coordinates and marked on the map. The generated diseased tree identification results are displayed in the form of a map, showing the location and disease information of each diseased tree. The marking helps forestry managers quickly locate and treat affected trees, ensuring the timeliness and effectiveness of disease control measures.

[0031] See also Figure 5 According to the results of diseased tree identification, combined with the database of spectral characteristics of diseased trees with different degrees of damage, the severity of the diseased tree is evaluated by analyzing the hyperspectral data of the target diseased tree, the degree of impact of the disease is analyzed, and the steps for generating severity rating data are as follows: S401: Based on the diseased tree identification results, the hyperspectral data of the identified diseased trees are analyzed, key spectral parameters including reflectance and absorption peaks are recorded, and a diseased tree spectral data set is generated; In sub-step S401, based on the results of diseased tree identification, the hyperspectral data of each diseased tree is analyzed, and the hyperspectral imaging technology is used to capture the reflectance information from visible light to near-infrared spectrum. After collecting data through the spectral imager, the spectral analysis software is used for preliminary processing to automatically identify and record key spectral parameters, including the reflectance of each band and the absorption peak of a specific wavelength. During the data analysis process, attention is paid to the detection of absorption peaks. The absorption peaks indicate changes in the concentrations of various chemical substances. The changes are direct indicators of the pathological state. By comparing the changes in each band and recording the bands with significant changes, the generated diseased tree spectral data set includes the spectral characteristics of each diseased tree, providing the necessary data basis for disease analysis and stage identification.

[0032] S402: Analyze the spectral parameters of the diseased wood based on the diseased wood spectral data set, identify the development stage of the target diseased wood by comparing it with the spectral data of the sample diseased wood, and generate development stage analysis data; In sub-step S402, spectral analysis methods are used to analyze the diseased tree spectral data set. The methods include using principal component analysis and spectral matching technology to distinguish the different development stages of diseased trees. Principal component analysis is used to reduce the dimension of spectral data and highlight the main features. Spectral matching technology compares the spectral data of known sample diseased trees and identifies the disease development stage of the target diseased tree by calculating the similarity scores of different spectra. The process involves a large amount of data comparison and pattern recognition to ensure accuracy and reduce calculation time. The generated development stage analysis data shows which development stage each diseased tree is in, which is crucial for the formulation of subsequent control measures.

[0033] S403: Based on the development stage analysis data, according to the disease development stages of multiple diseased trees, the disease severity score of the target diseased tree is calculated, the disease is classified and rated, and severity rating data is generated; In the above content, based on the development stage analysis data, by analyzing the disease development stage of a single diseased tree, according to the formula , calculate the severity score of the target diseased tree ; In the formula, represents the severity score of the disease, Represents the disease stage score, represents the duration factor, Represents the disease influencing factor; Detailed explanation of the formula and the process of formula calculation and derivation: Score of disease stage of hypothetical diseased trees , duration factor , disease influencing factors ,calculate : ; The result 7.2 indicates that the severity score of the target diseased tree is 7.2. The calculation process is used to quantify the disease progression and severity of a single diseased tree, providing a standardized basis for the classification and rating of the disease.

[0034] See also Figure 6 , combined with severity rating data, by analyzing the geographical location information of multiple diseased trees, evaluating the location distribution characteristics of diseased trees, predicting the spread path and speed of the disease, and generating the spread trend prediction data, the specific steps are: S501: Based on the severity rating data, according to the marks and geographical location information of multiple diseased trees, a diseased tree distribution map is drawn to generate location distribution data; In sub-step S501, the severity rating data and the marks and geographic location information of each diseased tree are used to draw a diseased tree distribution map using geographic information system technology, including inputting the geographic location data and diseased tree rating data into the GIS software. The software uses the data to generate a spatial distribution map of diseased trees, showing the locations of diseased trees with different ratings. Through map visualization, forestry managers can intuitively see the distribution of diseased trees and effectively allocate resources and control epidemics. Combined with the data stratification function, diseased trees can be screened and viewed according to their severity. The generated location distribution data contains coordinate information and the severity rating of diseased trees, providing basic data for further analysis.

[0035] S502: Analyze the spatial relationship and distance between multiple diseased trees based on the location distribution data, identify disease diffusion clusters, and generate disease cluster analysis results; In sub-step S502, based on the location distribution data, the spatial analysis tools nearest neighbor analysis and kriging interpolation are used to help analyze the spatial relationship and mutual distance between diseased trees. The density and distribution pattern of diseased tree clusters are calculated through statistical methods. Kriging interpolation is used to predict the spatial continuity of diseased tree distribution and identify disease spread clusters. The analysis results show the risk areas where diseases gather, providing a scientific basis for the formulation of prevention and control measures. The generated disease cluster analysis results describe the clustering pattern and potential spread area of ​​the disease. This information is crucial for the formulation of epidemic monitoring and control strategies.

[0036] S503: Based on the results of the disease cluster analysis, identify the impact range of the disease, calculate the diffusion speed, and predict the diffusion path to generate diffusion trend prediction data; The specific formula for calculating the diffusion rate is: ; in, Represents the speed of disease spread, that is, the distance the disease spreads per unit time. Representative The observed disease distance indicates the distance from the initial point to the current point. Represents the corresponding time point, that is, the specific time when the disease was observed. is the average value of the disease distance of all observation points, is the average time of all observation points, is the observation sequence number, used to mark the observation point. is the total number of observations, indicating the number of data points involved in the calculation.

[0037] formula: ; Detailed explanation of the formula and the process of formula calculation and derivation: The formula is used to calculate the spread rate of pine wood nematode disease, and the results are used to predict future spread trends and plan prevention and control strategies; Parameter meaning and setting value: is the number of observation points, assuming it is 4; is the disease diffusion distance at the i-th observation point, which is assumed to be 2 km, 2.5 km, 3 km, and 3.5 km respectively, reflecting the cumulative disease diffusion distance from the epidemic point to each time point. is the corresponding time point, assuming it is 10 days, 20 days, 30 days, and 40 days, reflecting the observation time of the corresponding distance, and They are and The average value of kilometer, sky; Substitute the parameters into the formula for calculation: ; ; ; ; result It shows that the disease spreads an average of 0.05 kilometers per day, providing quantitative data support for forestry managers to evaluate and plan the implementation time window and resource allocation of management measures.

[0038] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. An intelligent extraction method for pine wood nematode diseased wood based on UAV hyperspectral imaging, characterized in that: The following steps are involved: According to the location data of the target forest, the entire target forest is photographed, and radiation correction and orthorectification are performed. Combined with the geographic coordinate information of the image, a UAV hyperspectral remote sensing orthophoto is generated; Based on the UAV hyperspectral remote sensing orthophoto and ground measured data, the crown spectral curves of pine wood nematode diseased trees with different degrees of damage are extracted, and the characteristics are analyzed to construct a spectral feature library of diseased trees; The diseased tree spectral feature library is used to analyze the hyperspectral data of the target forest area, identify the diseased trees through spectral feature matching, cluster the diseased tree pixels according to the pixel erosion and expansion algorithm, extract the diseased tree coordinates, record the location of the target diseased tree and add marking information to generate a diseased tree identification result; According to the diseased tree identification results, combined with the database of spectral characteristics of diseased trees with different degrees of damage, by analyzing the hyperspectral data of the target diseased tree, the severity of the disease of the diseased tree is evaluated, the degree of impact of the disease is analyzed, and severity rating data is generated; In combination with the severity rating data, by analyzing the geographical location information of multiple diseased trees, the location distribution characteristics of the diseased trees are evaluated, the spread path and speed of the disease are predicted, and the spread trend prediction data is generated.

2. The intelligent extraction method of pine wood nematode diseased wood based on unmanned aerial vehicle hyperspectral imaging according to claim 1 is characterized in that: The drone hyperspectral remote sensing orthophoto includes hyperspectral image data, image shooting timestamp and geographic coordinate information, and spectral band information. The diseased tree spectral feature library includes diseased tree spectral standard curves, disease type labels, and spectral width ranges. The diseased tree identification results include diseased tree geographic location information, disease severity information, and diseased tree marking information. The severity rating data includes diseased tree health index, disease severity score, and diseased tree growth stage data. The diffusion trend prediction data includes diffusion path prediction results, impact area analysis results, and diffusion speed prediction results.

3. The intelligent extraction method of pine wood nematode diseased wood based on UAV hyperspectral imaging according to claim 1 is characterized in that: According to the target forest location data, the entire target forest is photographed, and radiation correction and orthorectification are performed. Combined with the image geographic coordinate information, the steps to generate the UAV hyperspectral remote sensing orthophoto are as follows: According to the target forest location data, the flight path of the drone is set and full-area aerial photography is carried out to obtain aerial data and generate an aerial image set; Based on the aerial image set, the image data is radiometrically corrected, the spectral value of the image is adjusted according to the actual lighting conditions, the influence of environmental factors is eliminated, and a radiometrically corrected image is obtained; Based on the radiation correction image, ortho correction is performed to eliminate the deformation caused by the terrain, and the image geographic coordinate information is combined to correspond to the actual ground position to generate a UAV hyperspectral remote sensing orthoimage.

4. The intelligent extraction method of pine wood nematode diseased wood based on unmanned aerial vehicle hyperspectral imaging according to claim 1 is characterized in that: Based on the UAV hyperspectral remote sensing orthophoto and ground measured data, the crown spectral curves of pine wood nematode diseased trees with different degrees of damage are extracted, and the characteristics are analyzed to construct the diseased wood spectral feature library in the following steps: Based on the UAV hyperspectral remote sensing orthophoto combined with ground measured data positioning, the spectral data of pine wood nematode diseased wood samples with different degrees of damage are extracted to generate a spectral feature set of the diseased wood samples; Based on the spectral feature set of the diseased wood samples, feature analysis is performed on the spectral data to identify features associated with pine wood nematode disease in the spectral bands, screen heterogeneous feature band information, and generate key spectral feature data; Based on the key spectral feature data, a spectral feature database is constructed to record disease stage information corresponding to a variety of spectral features, and a diseased tree spectral feature library is generated.

5. The intelligent extraction method of pine wood nematode diseased wood based on UAV hyperspectral imaging according to claim 1 is characterized in that: The diseased tree spectral feature library is used to analyze the hyperspectral data of the target forest area, identify the diseased trees through spectral feature matching, cluster the diseased tree pixels according to the pixel erosion and expansion algorithm, extract the diseased tree coordinates, record the location of the target diseased tree and add marking information, and generate the diseased tree identification results in the following steps: Based on the diseased tree spectral feature library, data preprocessing is performed on the hyperspectral data of the target forest area to remove noise and outliers and generate data processing records; Based on the data processing record, by comparing and matching the spectral data of the diseased wood sample, calculating the similarity of the spectral data, identifying the pine wood nematode diseased wood, and generating a feature data comparison record; Based on the feature data comparison record, the spatial location information of multiple diseased trees is recorded and marked to generate a diseased tree identification result.

6. The method for intelligent extraction of pine wood nematode diseased wood based on unmanned aerial vehicle hyperspectral imaging according to claim 5 is characterized in that: The specific formula for calculating the similarity of spectral data is: ； in, represents the cosine similarity value, The first spectral values, represents the corresponding spectral value in the spectral library of known pine nematode diseased wood. represents the total number of spectral data points, is the current index position in the index column.

7. The method for intelligent extraction of pine wood nematode diseased wood based on unmanned aerial vehicle hyperspectral imaging according to claim 1, characterized in that: According to the diseased tree identification results, combined with the spectral feature database of diseased trees with different degrees of damage, by analyzing the hyperspectral data of the target diseased tree, the severity of the diseased tree is evaluated, the degree of impact of the disease is analyzed, and the steps of generating severity rating data are specifically as follows: Based on the diseased tree identification results, the hyperspectral data of the identified diseased trees are analyzed, key spectral parameters including reflectance are recorded, and a diseased tree spectral data set is generated; Based on the diseased tree spectral data set, the spectral parameters of the diseased tree are analyzed, and by comparing with the spectral data of the sample diseased tree, the development stage of the disease of the target diseased tree is identified, and development stage analysis data is generated; Based on the development stage analysis data, according to the disease development stages of multiple diseased trees, the disease severity score of the target diseased tree is calculated, the disease is classified and rated, and severity rating data is generated.

8. The intelligent extraction method of pine wood nematode diseased wood based on UAV hyperspectral imaging according to claim 1 is characterized in that: In combination with the severity rating data, by analyzing the geographical location information of multiple diseased trees, evaluating the location distribution characteristics of the diseased trees, predicting the spread path and speed of the disease, and generating the spread trend prediction data, the specific steps are: Based on the severity rating data, and according to the marks and geographical location information of a plurality of diseased trees, a diseased tree distribution map is drawn to generate location distribution data; Based on the location distribution data, analyzing the spatial relationship and distance between multiple diseased trees, identifying disease diffusion clusters, and generating disease cluster analysis results; Based on the disease cluster analysis results, the impact range of the disease is identified, the diffusion speed is calculated, and the diffusion path is predicted to generate diffusion trend prediction data.

9. The intelligent extraction method of pine wood nematode diseased wood based on unmanned aerial vehicle hyperspectral imaging according to claim 8 is characterized in that: The specific formula for calculating the diffusion rate is: ； in, Represents the speed of disease spread, that is, the distance the disease spreads per unit time. Representative The observed disease distance indicates the distance from the initial point to the current point. Represents the corresponding time point, that is, the specific time when the disease was observed. is the average value of the disease distance of all observation points, is the average time of all observation points, is the observation sequence number, used to mark the observation point. is the total number of observations, indicating the number of data points involved in the calculation.

Images