Pest and disease monitoring method for artificial promotion and renewal of natural quercus liaotungensis forest

Abstract

The invention relates to a pest and disease monitoring method for artificial promotion and renewal of a natural quercus liaotungensis forest. The method comprises the following steps: acquiring multispectral remote sensing image data and ground survey data of a natural quercus liaotungensis forest area; eliminating spectral interference of the multispectral remote sensing image data to obtain standardized remote sensing image data; spectral features are extracted from the standardized remote sensing image data and input into a disease and insect pest classification and recognition model, a disease and insect pest classification result is obtained, the disease and insect pest classification and recognition model is constructed by a deep learning model and obtained through training according to a training set, and the training set comprises different disease and insect pest types obtained according to ground survey data and corresponding spectral features; the disease and pest classification result is matched with a disease and pest spectral feature library, a disease and pest risk grade map is obtained, and the disease and pest spectral feature library is used for classified storage of spectral features under the influence of different degrees of diseases and pests. According to the invention, the accuracy and efficiency of monitoring diseases and insect pests in the natural forest area of Quercus liaotungensis are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of forestry monitoring, and particularly to a pest and disease monitoring method for artificial promotion of natural regeneration of Quercus wutaishanica forests. Background Art

[0002] During the process of artificial promotion of natural regeneration of Quercus wutaishanica forests, it is crucial to effectively monitor and control pests and diseases. Although the application of remote monitoring technology can achieve dynamic monitoring of pests and diseases in forest areas, there are still some technical problems to be solved. Firstly, due to the complex and diverse terrain of Quercus wutaishanica natural forests and the high canopy coverage, the spectral information of remote sensing images is easily interfered, resulting in low accuracy of pest and disease information extraction. Secondly, the spectral characteristics of pests and diseases at different development stages have small differences in remote sensing images, making it difficult to accurately identify and classify pests and diseases. Moreover, some pests and diseases with strong concealment are difficult to be effectively detected by remote sensing equipment, easily causing missed detections.

[0003] In view of the above problems, the present invention provides a pest and disease monitoring method for artificial promotion of natural regeneration of Quercus wutaishanica forests. Summary of the Invention

[0004] The object of the present invention is to provide a pest and disease monitoring method for artificial promotion of natural regeneration of Quercus wutaishanica forests, realizing automated and intelligent pest and disease monitoring and analysis, and improving the accuracy and efficiency of pest and disease monitoring in Quercus wutaishanica natural forest areas.

[0005] To achieve the above object, the present invention provides the following solutions:

[0006] A pest and disease monitoring method for artificial promotion of natural regeneration of Quercus wutaishanica forests, comprising:

[0007] Obtaining multi-spectral remote sensing image data and ground survey data of Quercus wutaishanica natural forest areas;

[0008] Eliminating spectral interference of the multi-spectral remote sensing image data to obtain standardized remote sensing image data;

[0009] Inputting the spectral features extracted from the standardized remote sensing image data into a pest and disease classification and recognition model to obtain a pest and disease classification result, wherein the pest and disease classification and recognition model is constructed by a deep learning model and trained according to a training set, and the training set includes different pest and disease types and corresponding spectral features obtained according to the ground survey data;

[0010] Matching the pest and disease classification result with a pest and disease spectral feature library to obtain a pest and disease risk level map, wherein the pest and disease spectral feature library is used to classify and store spectral features under the influence of pests and diseases at different degrees.

[0011] Optionally, obtaining the standardized remote sensing image data includes:

[0012] Obtain the vegetation coverage of the natural forest area of Quercus wutaishanica, divide the natural forest area of Quercus wutaishanica according to the vegetation coverage, and obtain different coverage areas;

[0013] According to the multispectral remote sensing image data, obtain the spectral values of different coverage areas;

[0014] Perform different corrections according to the spectral values of different coverage areas to obtain the standardized remote sensing image data.

[0015] Optionally, after obtaining the standardized remote sensing image data, it includes:

[0016] If there are abnormal spectral values in the standardized remote sensing image data, use the spectral feature matching algorithm to correct the multispectral remote sensing image data;

[0017] According to the corrected multispectral remote sensing image data, re-obtain the standardized remote sensing image data.

[0018] Optionally, establishing the pest and disease spectral feature library includes:

[0019] According to the standardized remote sensing image data, extract the spectral features of the natural forest area of Quercus wutaishanica;

[0020] According to the spectral features of the natural forest area of Quercus wutaishanica, obtain the spectral features of pests and diseases, and construct the pest and disease spectral feature library.

[0021] Optionally, the deep learning model for constructing the pest and disease classification and recognition model is a convolutional neural network;

[0022] The convolutional neural network learns the features of different scales corresponding to the spectral features of different pest and disease types through multi-layer convolution and pooling operations, extracts the features of pests and diseases, and obtains the pest and disease classification results.

[0023] Optionally, matching the pest and disease classification results with the pest and disease spectral feature library to obtain the pest and disease risk level map includes:

[0024] If the pest and disease classification results match the pest and disease spectral feature library, generate pest and disease risk distribution data;

[0025] Input the pest and disease risk distribution data into a pre-trained support vector machine for classification to obtain the pest and disease risk level map.

[0026] Optionally, the method further includes:

[0027] Obtain the high-incidence areas of pests and diseases according to the pest and disease risk level map;

[0028] Obtain high-resolution remote sensing images within the high-incidence areas of the pests and diseases, and perform image enhancement processing to obtain the processed image data;

[0029] Obtain spectral features from the processed image data, and compare them with the pest and disease spectral feature library to identify suspicious areas;

[0030] Input the spectral features of the suspicious areas into the pest and disease classification and recognition model to obtain the hidden pest and disease classification results;

[0031] Generate hidden pest and disease risk distribution data according to the hidden pest and disease classification results, and update the pest and disease risk distribution data.

[0032] Optionally, identifying suspicious areas includes: if the matching degree between the spectral features obtained from the processed image data and the pest and disease spectral feature library exceeds a preset threshold, it is determined as the suspicious area.

[0033] The beneficial effects of the present invention are as follows: The present invention discloses an intelligent monitoring method for pests and diseases in the natural forest area of Quercus wutaishanica. This method first obtains multi-spectral remote sensing images and ground survey data, and eliminates spectral interference through calibration. Then, it extracts the spectral features of Quercus wutaishanica trees and pests and diseases, establishes a feature library, and combines a deep learning model to achieve accurate classification and recognition. Finally, the present invention integrates various technologies to realize automated and intelligent pest and disease monitoring, analysis, and decision support. This method not only improves the accuracy and efficiency of pest and disease monitoring in the natural forest area of Quercus wutaishanica, but also provides a scientific basis for forestry management, which is of great significance for protecting forest resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0035] Figure 1 It is a flowchart of an intelligent monitoring method for pests and diseases in the natural forest area of Quercus wutaishanica according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0036] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0037] To make the above objects, features, and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0038] Traditional forest pest and disease monitoring in forest areas mostly adopts the method of manual investigation. This monitoring method is time-consuming and laborious, and is not conducive to large-scale and accurate forest pest and disease monitoring. By using remote sensing technology for forest pest and disease monitoring, it is possible to quickly, frequently, and over a large range identify and evaluate forest pest and disease areas, and be able to monitor the dynamics of pests and diseases in real time. There have been many related studies and applications. However, pests and diseases are greatly affected by environmental factors such as temperature and humidity, and have biological complexity. At the same time, the spatio-temporal resolution of remote sensing satellite technology is limited, and there is still a lot of work to be done to more fully explore the pest and disease information contained in remote sensing images. In addition, how to comprehensively and systematically monitor forest pests and diseases based on remote sensing technology and continuously and automatically monitor forest pests and diseases accurately still faces challenges.

[0039] In the actual monitoring process of the present invention, necessary ground investigation work is carried out. By scientifically setting up sample plots, collecting pest and disease samples, and obtaining first-hand data such as their types, quantities, distributions, and damage degrees. At the same time, during the investigation process, attention should also be paid to observing the environmental conditions for the occurrence and development of pests and diseases, providing data support for subsequent prediction and early warning analysis. Finally, the correlation analysis between different investigation data should be strengthened. On the basis of fully considering various influencing factors, an investigation index system and technical regulations suitable for the monitoring of pests and diseases in the natural forest of Quercus wutaishanica should be constructed, continuously improving the pertinence and effectiveness of monitoring, and providing scientific guarantees for the sustainable management of the natural forest of Quercus wutaishanica.

[0040] As Figure 1 shown, this embodiment provides a method for monitoring pests and diseases for artificial promotion of regeneration in the natural forest of Quercus wutaishanica, including:

[0041] Obtaining multi-spectral remote sensing image data and ground investigation data of the natural forest area of Quercus wutaishanica;

[0042] Eliminating the spectral interference of the multi-spectral remote sensing image data to obtain standardized remote sensing image data;

[0043] Inputting the spectral features extracted from the standardized remote sensing image data into a pest and disease classification and recognition model to obtain a pest and disease classification result. Among them, the pest and disease classification and recognition model is constructed by a deep learning model and trained according to a training set, and the training set includes different pest and disease types and corresponding spectral features obtained according to the ground investigation data;

[0044] Matching the pest and disease classification result with a pest and disease spectral feature library to obtain a pest and disease risk level map, where the pest and disease spectral feature library is used to classify and store spectral features under the influence of pests and diseases at different levels.

[0045] Specifically, the multispectral remote sensing image data of the natural forest area of Quercus wutaishanica includes visible light, near-infrared, and short-wave infrared bands. When obtaining the data, the time phase and spatial resolution need to be considered, such as selecting Landsat 8 OLI images during the growing season.

[0046] Furthermore, obtaining standardized remote sensing image data includes:

[0047] Obtain the vegetation coverage of the natural forest area of Quercus wutaishanica, divide the natural forest area of Quercus wutaishanica according to the vegetation coverage, and obtain different coverage areas;

[0048] Obtain the spectral values of different coverage areas based on the multispectral remote sensing image data;

[0049] Perform different corrections based on the spectral values of different coverage areas to obtain standardized remote sensing image data.

[0050] Specifically, the division of vegetation coverage can be based on the normalized difference vegetation index (NDVI) value. The vegetation index is extracted based on band operations. For example, the normalized difference vegetation index NDVI = (NIR - Red) / (NIR + Red). If NDVI < 0.3, it is a low-coverage area; if 0.3 ≤ NDVI < 0.6, it is a medium-coverage area; if NDVI ≥ 0.6, it is a high-coverage area. Correcting for the spectral value differences in different coverage areas is a key step in improving the quality of remote sensing data. Radiometric correction can eliminate the sensor system error and convert the original digital number (DN) to apparent radiance. Atmospheric correction can remove the effects of atmospheric scattering and absorption to obtain the surface reflectance. For example, for high-coverage areas, the dark target method can be used for atmospheric correction, using water bodies or shadow areas as dark pixels; for low-coverage areas, the 6S model can be selected for more accurate atmospheric correction. The corrected spectral values are used to generate standardized remote sensing image data, and this process helps to eliminate the differences between different sensors and data at different times.

[0051] Even further, after obtaining the standardized remote sensing image data, it includes:

[0052] If there are abnormal spectral values in the standardized remote sensing image data, then use the spectral feature matching algorithm to correct the multispectral remote sensing image data;

[0053] Based on the corrected multispectral remote sensing image data, re-obtain the standardized remote sensing image data.

[0054] Specifically, if abnormal spectral values are found in the standardized data, such as the reflectance of the near-infrared band of some pixels being abnormally high, the spectral feature matching algorithm can be used for correction. This algorithm compares the spectral features of the target pixel with the surrounding normal pixels and finds the most similar spectral curve for replacement.

[0055] Furthermore, establishing a spectral feature library for pests and diseases includes:

[0056] Extract the spectral features of the natural forest area of Quercus wutaishanica according to the standardized remote sensing image data;

[0057] Based on the spectral features of the natural forest area of Quercus wutaishanica, obtain the spectral features of pests and diseases, and construct a spectral feature library for pests and diseases.

[0058] Specifically, the extraction of spectral features of Quercus wutaishanica trees is the basis for pest and disease monitoring. By analyzing the standardized remote sensing image data, the spectral differences between healthy trees and trees affected by pests and diseases can be obtained. For example, the reflectance of healthy Quercus wutaishanica leaves is relatively high in the near-infrared band, while the reflectance of leaves affected by pests and diseases will decrease significantly. When constructing the spectral feature library for pests and diseases, the spectral features under different degrees of pest and disease infestation can be classified and stored, such as the spectral curves of slightly, moderately, and severely damaged ones. By comparing the spectral features of the target area with the samples in the feature library, the risk of pests and diseases can be preliminarily judged. If the reflectance of the near-infrared band of Quercus wutaishanica in a certain area is significantly lower than that of the healthy samples and is similar to the damaged samples in the feature library, there may be a risk of pests and diseases.

[0059] Furthermore, the deep learning model for constructing the pest and disease classification and recognition model is a convolutional neural network;

[0060] The convolutional neural network learns the features of different scales corresponding to the spectral features of different pest and disease types through multi-layer convolution and pooling operations, extracts the features of pests and diseases, and obtains the pest and disease classification results.

[0061] Even further, match the pest and disease classification results with the spectral feature library for pests and diseases to obtain a pest and disease risk level map, including:

[0062] If the pest and disease classification results match the spectral feature library for pests and diseases, generate pest and disease risk distribution data;

[0063] Input the pest and disease risk distribution data into a pre-trained support vector machine for classification to obtain a pest and disease risk level map.

[0064] Specifically, obtain the pest and disease sample data collected from ground surveys and construct a sample database. Use the deep learning model to train the sample data to obtain a pest and disease classification and recognition model. Extract the spectral features of Quercus wutaishanica trees from the standardized remote sensing image data to generate spectral feature data. Input the extracted spectral feature data into the pest and disease classification and recognition model to obtain the pest and disease classification results. If the classification results match the pre-established spectral feature library for pests and diseases, generate pest and disease risk distribution data. Use the support vector machine algorithm to classify the pest and disease risk distribution data to obtain a pest and disease risk level map. Based on the pest and disease risk level map, correct the abnormal spectral values in the standardized remote sensing image data and update the spectral feature library for pests and diseases.

[0065] Obtaining the sample data of pests and diseases collected from ground surveys is the basis for constructing a monitoring system for pests and diseases of Quercus wutaishanica. For example, multiple sampling points are set in the main distribution areas of Quercus wutaishanica to collect samples such as leaves and bark of different types of pests and diseases, and record their spectral characteristics. These sample data will be used to construct a sample database to support subsequent model training. The application of deep learning models can significantly improve the accuracy of pest and disease identification. Taking the convolutional neural network (CNN) as an example, the collected sample images can be used for training so that the model can automatically extract the characteristics of pests and diseases. Through multi-layer convolution and pooling operations, the model can learn features at different scales, thereby better classifying pests and diseases. The support vector machine (SVM) algorithm performs well in classifying the pest and disease risk distribution data. SVM can find the optimal classification hyperplane in the high-dimensional feature space to effectively distinguish different risk levels. For example, the risk level can be divided into three levels: low, medium, and high. SVM outputs the risk level of each region by learning the input variables of the pest and disease risk distribution data.

[0066] Furthermore, the method further includes:

[0067] Obtaining the high-incidence areas of pests and diseases according to the pest and disease risk level map;

[0068] Obtaining high-resolution remote sensing images within the high-incidence areas of pests and diseases, and performing image enhancement processing to obtain the processed image data;

[0069] Obtaining spectral characteristics from the processed image data, and comparing them with the pest and disease spectral characteristic library to identify suspicious areas;

[0070] Inputting the spectral characteristics of the suspicious areas into the pest and disease classification and identification model to obtain the hidden pest and disease classification results;

[0071] Generating hidden pest and disease risk distribution data according to the hidden pest and disease classification results to update the pest and disease risk distribution data.

[0072] Furthermore, identifying the suspicious areas includes: if the matching degree between the spectral characteristics obtained from the processed image data and the pest and disease spectral characteristic library exceeds a preset threshold, it is determined as a suspicious area.

[0073] Specifically, multispectral data of high-incidence areas are obtained from high-resolution remote sensing images, and image enhancement algorithms are used to preprocess the multispectral data. According to the preprocessed image data, spectral features are extracted and compared with a pre-established spectral feature library. If the matching degree between the spectral features of the target area and the spectral features of pests and diseases exceeds a preset threshold, it is judged as a suspicious area, and the hidden pest and disease identification mechanism is triggered. A pest and disease classification and identification model is used to classify the spectral features of the suspicious area to determine the classification results of hidden pests and diseases. According to the classification results, a distribution map of hidden pests and diseases is generated, and relevant data in the pest and disease spatial distribution database are updated. A clustering algorithm is used to divide the distribution area of hidden pests and diseases to generate a refined pest and disease risk zoning map. Combining with the pest and disease risk zoning map, the remote sensing image data of the high-incidence area are corrected, and the abnormal records in the high-resolution image database are updated.

[0074] In this embodiment, according to the distribution map of hidden pests and diseases, combined with temperature values, humidity values and soil values, the pest and disease risk zoning range is corrected to generate a refined pest and disease risk zoning map. A clustering algorithm is used to analyze the pest and disease risk zoning map to divide the distribution areas of pests and diseases with different risk levels, and the risk level data in the pest and disease spatial distribution database are updated. Combining with the risk level data in the pest and disease spatial distribution database, the abnormal records in the remote sensing image data of the high-incidence area are corrected, and the high-resolution image database is updated. According to the updated high-resolution image database and pest and disease spatial distribution database, the environmental factor model for the occurrence and development of pests and diseases is optimized, and the model parameters are adjusted.

[0075] Through the above steps, a comprehensive monitoring method for Quercus wutaishanica pests and diseases can be obtained. This method combines remote sensing technology and machine learning algorithms, can timely detect pest and disease risks, and provides a scientific basis for forestry management departments. Continuously updating and optimizing the spectral feature library can further improve the adaptability and accuracy of the system, and make an important contribution to the health management of Quercus wutaishanica.

[0076] The embodiments described above are only descriptions of the preferred embodiments of the present invention, and do not limit the scope of the present invention. Without departing from the design spirit of the present invention, various deformations and improvements made by those of ordinary skill in the art to the technical solutions of the present invention shall fall within the protection scope determined by the claims of the present invention.

Claims

1. A method for monitoring pests and diseases for artificially promoting regeneration of natural Quercus liaotungensis forests, characterized in that: include: Acquire multispectral remote sensing image data and ground survey data of natural Quercus liaotungensis forest areas; Eliminating spectral interference of the multispectral remote sensing image data to obtain standardized remote sensing image data; Extracting spectral features from the standardized remote sensing image data and inputting them into a pest and disease classification and recognition model to obtain a pest and disease classification result, wherein the pest and disease classification and recognition model is constructed by a deep learning model and obtained by training based on a training set, and the training set includes different pest and disease types and corresponding spectral features obtained based on the ground survey data; The pest classification result is matched with a pest spectral feature library to obtain a pest risk level map, wherein the pest spectral feature library is used to classify and store spectral features under different degrees of pest influence.

2. The method for monitoring pests and diseases for artificially promoting regeneration of natural Quercus liaotungensis forest according to claim 1, characterized in that: Acquisition of standardized remote sensing image data includes: Obtaining the vegetation coverage of the natural Quercus liaotungensis forest area, and dividing the natural Quercus liaotungensis forest area according to the vegetation coverage to obtain different coverage areas; According to the multispectral remote sensing image data, spectral values ​​of different coverage areas are obtained; Different corrections are performed according to the spectral values ​​of different coverage areas to obtain the standardized remote sensing image data.

3. The method for monitoring pests and diseases for artificially promoting regeneration of natural Quercus liaotungensis forest according to claim 2, characterized in that: After obtaining the standardized remote sensing image data, the following steps are included: If there are abnormal spectral values ​​in the standardized remote sensing image data, a spectral feature matching algorithm is used to correct the multispectral remote sensing image data; The standardized remote sensing image data is reacquired according to the corrected multispectral remote sensing image data.

4. The method for monitoring pests and diseases for artificially promoting regeneration of natural Quercus liaotungensis forests according to claim 1, characterized in that: Establishing the spectral feature library of pests and diseases includes: Extracting the spectral characteristics of the natural forest area of ​​Quercus liaotungensis according to the standardized remote sensing image data; According to the spectral characteristics of the natural forest area of ​​Quercus liaotungensis, the spectral characteristics of pests and diseases are obtained, and the spectral characteristic library of pests and diseases is constructed.

5. The method for monitoring pests and diseases for artificially promoting regeneration of natural Quercus liaotungensis forests according to claim 1, characterized in that: The deep learning model for constructing the pest and disease classification and recognition model is a convolutional neural network; The convolutional neural network learns the features of different scales of spectral features corresponding to different types of pests and diseases through multi-layer convolution and pooling operations, extracts the features of the pests and diseases, and obtains the pest and disease classification results.

6. The method for monitoring pests and diseases for artificially promoting regeneration of natural Quercus liaotungensis forests according to claim 1, characterized in that: Matching the pest classification result with the pest spectral feature library to obtain a pest risk level map includes: If the pest classification result matches the pest spectral feature library, then pest risk distribution data is generated; The pest and disease risk distribution data is input into a pre-trained support vector machine for classification to obtain a pest and disease risk level map.

7. The method for monitoring pests and diseases for artificially promoting regeneration of natural Quercus liaotungensis forests according to claim 6, characterized in that: The method further comprises: Obtaining areas with high incidence of pests and diseases according to the pest and disease risk level map; Obtain high-resolution remote sensing images of the pest and disease high-incidence area, and perform image enhancement processing to obtain processed image data; Acquire spectral features from the impact data after the treatment, and compare them with the spectral feature library of the pests and diseases to identify suspicious areas; Inputting the spectral features of the suspicious area into the pest classification and recognition model to obtain the hidden pest classification results; According to the hidden pest classification result, hidden pest risk distribution data is generated to update the pest risk distribution data.

8. The method for monitoring pests and diseases for artificially promoting regeneration of natural Quercus liaotungensis forests according to claim 7, characterized in that: Identifying a suspicious area includes: if the matching degree between the spectral feature obtained from the processed impact data and the spectral feature library of the pests and diseases exceeds a preset threshold, then determining it as the suspicious area.

Images