Fruit tree pest detection method based on machine vision

Abstract

The invention relates to the technical field of agricultural intelligent detection, in particular to a fruit tree disease and pest detection method based on machine vision, which comprises the following steps: acquiring visible light, infrared light and ultraviolet light images of a fruit tree at multiple angles through a multispectral camera to obtain complete fruit tree surface information, and preprocessing the acquired images to obtain a fruit tree disease and pest detection result; comprising the steps of denoising, illumination correction and image alignment, extracting feature vectors, including the color, shape and size of scabs, pest distribution and other features, of pest and disease damage from preprocessed images through a convolutional neural network, fusing the pest and disease damage features with environmental data by using a multi-modal data fusion algorithm, and enhancing the accuracy of pest and disease damage classification. Classifying and identifying diseases and pests based on a support vector machine classification algorithm; according to the dynamically adjusted threshold value and suggestions provided by the expert system, a disease and pest diagnosis report is generated, and prevention and control measure suggestions are provided; according to the invention, efficient and accurate detection and management of fruit tree diseases and insect pests can be realized, and the intelligent level of agricultural management is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural intelligent detection, and in particular to a method for detecting fruit tree diseases and insect pests based on machine vision. Background Art

[0002] The detection and management of fruit tree diseases and pests is a crucial part of the fruit tree planting process. Traditional methods of detecting fruit tree diseases and pests mainly rely on manual inspections, and visually observe the leaves, fruits and branches of fruit trees to judge the pest and disease situation. However, this manual detection method has great subjectivity and limitations, making it difficult to achieve comprehensive detection of large-scale orchards, and the detection efficiency is low, and it is easily restricted by weather, light and human resources. In addition, with the continuous expansion of the scale of fruit tree planting, the cost of manual detection methods has gradually increased, and it cannot meet the needs of modern agriculture for efficient and accurate pest and disease management. Therefore, fruit tree disease and pest detection methods based on automation and intelligent technology have gradually become the focus of research and application.

[0003] The existing fruit tree pest and disease detection systems based on image processing mainly rely on image analysis of a single spectrum (such as visible light), which cannot fully capture the multidimensional characteristics of pests and diseases, especially when the color changes of the lesions on the surface of the fruit trees are not obvious or the lighting conditions are complex, the detection accuracy is low. In addition, most existing systems fail to fully combine the impact of environmental factors (such as temperature, humidity, soil moisture, etc.) on the development of pests and diseases, resulting in inaccurate or unstable detection results in actual applications when faced with differences in the performance of pests and diseases under different environmental conditions. Although some systems use machine learning algorithms for pest and disease classification and identification, most models lack a dynamic adjustment mechanism and cannot optimize the detection threshold according to the type and growth stage of the fruit tree, resulting in prominent problems of false detection and missed detection. Current technology also lacks an effective mechanism for generating pest and disease control recommendations, making it difficult to provide personalized prevention and control plans, limiting the application effect of intelligent pest and disease management systems in practice. Summary of the invention

[0004] The invention provides a method for detecting fruit tree diseases and insect pests based on machine vision.

[0005] The method for detecting fruit tree diseases and insect pests based on machine vision includes the following steps:

[0006] S1, image acquisition: Use a multispectral camera to collect images of different parts of the fruit tree. The collected images include visible light images, infrared images and ultraviolet images. The complete surface information of the fruit tree is obtained through multi-angle shooting;

[0007] S2, image preprocessing: preprocess the collected images, including denoising, illumination correction and image alignment, and perform image enhancement through a deep learning model to ensure that the images have consistent clarity and contrast under different lighting conditions;

[0008] S3, Pest and disease feature extraction: Use convolutional neural network to extract pest and disease features from preprocessed images and construct feature vectors. The features include lesion color, shape, size and pest distribution;

[0009] S4, multimodal data fusion: using multimodal data fusion algorithm to fuse pest and disease characteristics with environmental data (such as temperature, humidity, soil moisture, etc.) to enhance the accuracy of pest and disease detection;

[0010] S5, pest and disease classification and identification: Based on the extracted feature vectors, the support vector machine model is used to classify pests and diseases, and the types of pests and diseases on fruit trees are automatically identified according to the classification results;

[0011] S6, dynamic threshold judgment and decision-making: Dynamically adjust the threshold of pest and disease detection for different fruit tree types and growth stages, combine the suggestions provided by the expert system, generate a pest and disease diagnosis report, and provide corresponding prevention and control measures.

[0012] Optionally, the S1 includes:

[0013] S11, Multispectral camera installation and calibration: The multispectral camera is fixedly installed at different positions around the fruit trees to ensure that the camera covers all parts of the fruit trees, and the camera is calibrated;

[0014] S12, visible light image acquisition: using a visible light sensor to acquire images of leaves, fruits and trunks of fruit trees;

[0015] S13, infrared image acquisition: using infrared sensors to collect thermal images of fruit trees and detect the difference in thermal radiation between fruit tree leaves and fruits;

[0016] S14, ultraviolet image acquisition: using an ultraviolet camera to acquire images of the surface of the fruit tree in the ultraviolet band, and detecting the fluorescent reflection characteristics of the diseased area on the surface of the fruit tree;

[0017] S15, image data integration and transmission: the collected visible light, infrared and ultraviolet image data are integrated and transmitted in real time to the preset image preprocessing module through the preset wireless transmission unit to ensure data integrity and transmission speed.

[0018] Optionally, S2 includes:

[0019] S21, image denoising: performing denoising on the collected image to eliminate random noise in the image;

[0020] S22, illumination correction: performing illumination correction for images under different illumination conditions;

[0021] S23, image alignment: aligning images at different angles;

[0022] S24, Image Enhancement: Enhance the preprocessed images through a deep learning model to ensure that the images have consistent clarity and contrast under different lighting conditions.

[0023] Optionally, the S3 includes:

[0024] S31, Feature extraction model design: Design a feature extraction model based on convolutional neural network to extract pest and disease related features from preprocessed images;

[0025] S32, lesion feature extraction: extracting lesion features from the feature map output by the feature extraction model, where the lesion features include color, shape and size;

[0026] S33, pest feature extraction: identifying the distribution of pests present in the image;

[0027] S34, feature vector construction: Integrate the lesion characteristics and pest distribution characteristics into a feature vector for subsequent classification and identification.

[0028] Optionally, the S4 includes:

[0029] S41, environmental data collection: collect environmental data through sensors installed around the fruit trees;

[0030] S42, standardization of multimodal feature data: standardize pest and disease characteristics and environmental data to make them have the same dimension;

[0031] S43. Multimodal data fusion algorithm: A multimodal data fusion algorithm based on weighted fusion to fuse pest and disease characteristic data with environmental data;

[0032] S44, data fusion optimization: Optimize the weight parameters in the fusion algorithm based on historical data.

[0033] Optionally, the S5 includes:

[0034] S51, training of support vector machine model: using historical pest and disease data and environmental data to train the support vector machine model;

[0035] S52, classification process of support vector machine model: using the trained support vector machine model to classify the extracted pest and disease features;

[0036] S53, verification and adjustment of classification results: verify the classification results of the support vector machine model.

[0037] Optionally, S5 further includes:

[0038] S54, Pest and disease type identification: Based on the classification results of the support vector machine model, the specific types of pests and diseases of fruit trees are automatically identified;

[0039] S55, verification of pest and disease identification results: verifying the pest and disease identification results;

[0040] S56, generation of prevention and control measures recommendations: Generate targeted prevention and control recommendations based on the pest and disease identification results.

[0041] Optionally, the S6 includes:

[0042] S61, threshold setting rules: setting thresholds for different pest and disease detection based on the type, variety, and growth stage of the fruit tree;

[0043] S62, dynamic threshold adjustment mechanism: real-time monitoring of the growth environment and development of pests and diseases of fruit trees, and dynamic adjustment of detection thresholds based on data changes.

[0044] Optionally, S6 further includes:

[0045] S63, pest and disease diagnosis report generation: Generate a pest and disease diagnosis report based on the classification results of the support vector machine model and the dynamically adjusted threshold value, combined with the suggestions provided by the expert system;

[0046] S64, generation of prevention and control measures recommendations: Based on the pest and disease diagnosis report and combined with the expert system, generate targeted prevention and control measures recommendations.

[0047] Beneficial effects of the present invention:

[0048] The present invention uses a multi-angle, multi-band (visible light, infrared light, ultraviolet light) image acquisition method of a multi-spectral camera to comprehensively capture different parts and surface information of fruit trees, ensuring the richness and integrity of the detection data. At the same time, pre-processing steps such as image denoising, illumination correction, and image alignment are adopted, and image enhancement is performed through a deep learning model to eliminate the influence of ambient illumination and ensure that the image has consistent clarity and contrast under different conditions. This multi-dimensional, high-quality image input lays a solid foundation for subsequent feature extraction and classification of pests and diseases, and significantly improves the accuracy and robustness of the detection system.

[0049] The present invention forms a more representative comprehensive feature vector by multimodal data fusion of pest and disease characteristics (color, shape, size) and environmental data (temperature, humidity, soil moisture). The weight of the features is dynamically assigned using a weighted fusion algorithm to optimize the correlation between pest and disease characteristics and environmental factors. The support vector machine classification algorithm is used to classify pests and diseases based on weighted fusion features, which can effectively distinguish different types of pests and diseases, and the classification results are more accurate. At the same time, combined with the diagnosis and suggestions provided by the expert system, the intelligent level of pest and disease identification is further improved, ensuring the scientificity and practicality of the diagnostic report and prevention and control measures.

[0050] The present invention monitors the growth environment and development of pests and diseases of fruit trees in real time by dynamically adjusting the threshold of pest and disease detection, ensuring that pests and diseases can be detected and diagnosed in a timely manner. Combined with the pest and disease diagnosis report and prevention and control measures recommended by the expert system, it can not only predict the potential development trend of pests and diseases, but also provide personalized prevention and control plans based on actual environmental conditions and historical data. This solution greatly improves the intelligent level of pest and disease management, ensures that fruit trees can be accurately controlled for pests and diseases at different growth stages, and effectively reduces economic losses. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0052] Figure 1 A schematic diagram of a method flow of an embodiment of the present invention;

[0053] Figure 2 Schematic diagram of S3 process of an embodiment of the present invention. DETAILED DESCRIPTION

[0054] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0055] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0056] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0057] like Figure 1-Figure 2 As shown, the method for detecting fruit tree pests and diseases based on machine vision includes the following steps:

[0058] S1, image acquisition: Use a multispectral camera to collect images of different parts of the fruit tree. The collected images include visible light images, infrared images and ultraviolet images. The complete surface information of the fruit tree is obtained through multi-angle shooting;

[0059] S2, image preprocessing: preprocess the collected images, including denoising, illumination correction and image alignment, and perform image enhancement through a deep learning model to ensure that the images have consistent clarity and contrast under different lighting conditions;

[0060] S3, Pest and disease feature extraction: Use convolutional neural network to extract pest and disease features from preprocessed images and construct feature vectors. The features include lesion color, shape, size and pest distribution;

[0061] S4, multimodal data fusion: using multimodal data fusion algorithm to fuse pest and disease characteristics with environmental data (such as temperature, humidity, soil moisture, etc.) to enhance the accuracy of pest and disease detection;

[0062] S5, pest and disease classification and identification: Based on the extracted feature vectors, the support vector machine model is used to classify pests and diseases, and the types of pests and diseases on fruit trees are automatically identified according to the classification results;

[0063] S6, dynamic threshold judgment and decision-making: Dynamically adjust the threshold of pest and disease detection for different fruit tree types and growth stages, combine the suggestions provided by the expert system, generate a pest and disease diagnosis report, and provide corresponding prevention and control measures.

[0064] S1 includes:

[0065] S11, Multispectral camera installation and calibration: The multispectral camera is fixedly installed at different positions around the fruit tree to ensure that the camera covers all parts of the fruit tree, and the camera is calibrated to ensure the image acquisition accuracy at different angles and spectra. The calibration process includes:

[0066] (1) Correct the optical distortion of the camera using a calibration plate;

[0067] (2) A light intensity adjustment mechanism is used to ensure that the multispectral camera can automatically adjust the exposure parameters under different lighting conditions to ensure the consistency of the collected images;

[0068] S12, visible light image acquisition: Use visible light sensors to acquire images of leaves, fruits and trunks of fruit trees. The acquired image resolution is 5000×4000 pixels to ensure that detailed information on the surface of the fruit trees can be obtained. The specific acquisition steps include:

[0069] (1) Set the camera focal length and shooting distance to ensure that all parts of the fruit tree are within the clear focus range;

[0070] (2) Through the multi-angle shooting mode, the fruit tree is rotated and photographed every 30 degrees, ultimately forming a 360-degree full-scale image acquisition;

[0071] S13, infrared image acquisition: Use infrared sensors to collect thermal images of fruit trees to detect the difference in thermal radiation between fruit tree leaves and fruits. The specific acquisition steps include:

[0072] (1) Set the temperature sensitivity range of the infrared camera to -10°C to 50°C to ensure that it can identify small temperature changes on the fruit trees;

[0073] (2) Collect infrared images at different time periods, record the temperature difference between day and night of fruit trees, and analyze the possible pest and disease conditions in areas with abnormal temperature differences;

[0074] S14, ultraviolet image acquisition: using an ultraviolet camera to acquire images of the surface of the fruit tree in the ultraviolet band, and detecting the fluorescent reflection characteristics of the diseased area on the surface of the fruit tree. The specific acquisition steps include:

[0075] (1) Setting the UV camera’s wavelength range between 320nm and 400nm to identify possible disease spots or insect pest activity on the surface of fruit trees;

[0076] (2) Use ultraviolet light to illuminate the surface of the fruit tree, record the fluorescence response of the diseased area, and generate a high-contrast ultraviolet image;

[0077] S15, image data integration and transmission: Integrate the collected visible light, infrared and ultraviolet image data, and transmit them to the preset image preprocessing module in real time through the preset wireless transmission unit to ensure the integrity and transmission speed of the data, specifically including:

[0078] (1) Annotate multispectral images according to timestamps and spatial location information to ensure that images in different bands can correspond correctly;

[0079] (2) Use 5G wireless transmission technology to achieve real-time transmission of large amounts of image data and avoid detection information lags caused by transmission delays;

[0080] Through precise multispectral image acquisition solutions, the complete acquisition of visible light, infrared and ultraviolet images of fruit trees at different angles and bands is ensured, providing a high-quality data foundation for subsequent pest and disease detection.

[0081] S2 includes:

[0082] S21, image denoising: performing denoising on the collected image to eliminate random noise in the image and improve image quality, including the following steps:

[0083] (1) Use Gaussian filter to denoise the image. The filtering formula is as follows:

[0084]

[0085] Among them, I(x, y) is the original image, I′(x, y) is the denoised image, σ is the standard deviation, which controls the denoising strength, and x 0 ,y 0 is the center point of the image;

[0086] (2) Use median filtering to eliminate salt and pepper noise and prevent loss of details. This method is suitable for ultraviolet images and other noisy images.

[0087] S22, Light Correction: Perform light correction on images under different light conditions to ensure that the images have consistent contrast under different light conditions. The specific steps include:

[0088] (1) Adaptive histogram equalization technology is used to adjust the brightness and contrast of the image. The corrected image has uniform brightness and clearer details.

[0089] (2) Use the illumination estimation model to calculate the global illumination deviation of each image. The correction formula is:

[0090]

[0091] Among them, I(x, y) is the original image pixel value, L(x, y) is the calculated local illumination deviation, and I 校正 (x, y) is the corrected image pixel value;

[0092] S23, image alignment: Align images at different angles to ensure the accuracy of subsequent image enhancement and feature extraction, including:

[0093] (1) Use the SIFT (Scale Invariant Feature Transform) algorithm to extract key feature points in the image and perform feature point matching on images at different angles;

[0094] (2) Based on the RANSAC algorithm, the error of the matched feature points is estimated and eliminated to eliminate abnormal points in the image registration process and improve the alignment accuracy;

[0095] (3) Align all feature points in the image through affine transformation to ensure pixel-level matching of images at different angles;

[0096] S24, Image Enhancement: Enhance the preprocessed image through a deep learning model to ensure that the image has consistent clarity and contrast under different lighting conditions. The specific steps include:

[0097] (1) An image enhancement model based on a convolutional neural network is used. The model includes three convolutional layers, each of which contains a ReLU activation function. The enhancement formula is expressed as:

[0098] I 增强 =f(I 输入 )=σ(W 3 *(σ(W 2 *(σ(W 1 *I 输入 +b 1 ))+b 2 ))+b 3 );

[0099] Among them, I 输入 is the input image, I 增强 is the enhanced image, W 1 , W 2 , W 3 are the convolution weight matrices, b 1 , b 2 , b 3 is the bias term, σ is the ReLU activation function;

[0100] (2) Optimize the image enhancement model based on convolutional neural network through training data sets to ensure that the model can adapt to different types of lighting and environmental changes and improve the effect of image enhancement;

[0101] Through image preprocessing, the noise and lighting effects in the image are eliminated, and the clarity and contrast of the image are enhanced through the deep learning model, ensuring the accuracy of subsequent pest and disease feature extraction and identification.

[0102] S3 includes:

[0103] S31, feature extraction model design: Design a feature extraction model based on convolutional neural network to extract pest and disease related features from preprocessed images, including the following structure:

[0104] (1) The input layer of the feature extraction model receives the preprocessed image with a size of 5000 × 4000 pixels and is divided into multiple channels (visible light, infrared light, ultraviolet light, etc.);

[0105] (2) The convolution layer uses a 3×3 convolution kernel to extract features. The convolution calculation formula is expressed as:

[0106]

[0107] Among them, f 卷积 (x, y) is the pixel value of the output feature map, W ij is the convolution kernel weight, I(x+i, y+j) is the pixel value of the input image, b is the bias, M and N represent the row and column size of the convolution kernel respectively;

[0108] S32, lesion feature extraction: extract lesion features from the feature map output by the feature extraction model. The lesion features include color, shape and size. The specific steps are as follows:

[0109] (1) Analyze the color distribution of the lesion using a preset color feature extraction unit, extract the color histogram of the RGB space, and generate a color vector of the lesion;

[0110] (2) Using a preset shape feature extraction unit and an edge detection algorithm (such as Canny edge detection) to identify the contour of the lesion and extract its geometric shape features, such as area, perimeter, roundness, etc.;

[0111] (3) Calculate the area ratio of the lesion in the image using the preset size feature extraction unit, which is calculated as:

[0112]

[0113] Among them, A r Indicates the area ratio of the lesion, N b Represents the total number of pixels in the lesion area, N t represents the total number of pixels in the entire leaf area;

[0114] S33, Pest feature extraction: Identify the distribution of pests in the image. The specific steps are as follows:

[0115] (1) Use a deep learning model (based on the Faster R-CNN architecture) to detect pest objects and identify the location and size of pests in images;

[0116] (2) Extract pest distribution feature vectors, including the size, location, and number of pests, and generate pest distribution maps based on the actual location distribution of pests;

[0117] S34, feature vector construction: Integrate the lesion characteristics and pest distribution characteristics into a feature vector for subsequent classification and identification. The feature vector structure is as follows:

[0118] (1) Color feature vector: three-dimensional color feature based on RGB histogram;

[0119] (2) Shape feature vector: contains the geometric shape information of the lesion (such as area, perimeter, roundness, etc.);

[0120] (3) Size feature vector: the relative size of the lesions and pests in the image;

[0121] (4) Pest distribution feature vector: spatial distribution of pests and their quantity information;

[0122] The convolutional neural network is used to extract pest and disease features from the preprocessed images, effectively obtaining the multi-dimensional features of fruit tree pests and diseases (including the color, shape, size and distribution of lesions), providing rich feature data for subsequent classification and identification.

[0123] S4 includes:

[0124] S41, Environmental data collection: Collect environmental data through sensors installed around fruit trees, including:

[0125] (1) Temperature sensor collection: collect the air temperature around the fruit trees, with a sampling frequency of once every 10 seconds, and a temperature range of -10°C to 50°C;

[0126] (2) Humidity sensor collection: collects the air humidity of the fruit tree growth environment, with the collected humidity range of 0% to 100%, which is used to analyze the impact of environmental humidity on pests and diseases;

[0127] (3) Soil moisture sensor collection: Soil moisture data at the roots of fruit trees is collected through humidity sensors buried in the soil. The sampling frequency is once every minute and the collection range is 0 to 100% moisture content.

[0128] S42, standardization of multimodal feature data: standardize the pest and disease characteristics and environmental data to make them have the same dimension, including:

[0129] (1) Normalize the characteristics of lesions and pests (such as color, shape, size, etc.) using the following normalization formula:

[0130]

[0131] Among them, x is the original eigenvalue, x min and x max are the minimum and maximum values ​​of the feature respectively, and x′ is the normalized feature value;

[0132] (2) Standardize environmental data (such as temperature, humidity, and soil moisture) using the Zscore standardization formula:

[0133]

[0134] Among them, x is the original environmental data, μ is the mean, σ is the standard deviation, and z is the standardized value;

[0135] S43. Multimodal data fusion algorithm: A multimodal data fusion algorithm based on weighted fusion is used to fuse pest and disease characteristic data with environmental data. Specifically, it includes the following steps:

[0136] (1) Feature fusion: The normalized pest and disease feature data and the standardized environmental data are combined into a joint feature vector, denoted as:

[0137] F=[f 病虫害 , f 温度 , f 湿度 , f 土壤湿度 ];

[0138] Among them, f 病虫害 The feature vector representing the pest, f 温度 , f 湿度 , f 土壤湿度 are the characteristic vectors of temperature, humidity, and soil moisture, respectively;

[0139] (2) Weighted fusion: Assign weights to each feature according to its importance. The fusion formula is:

[0140] F 融合 =α 1 f 病虫害 +α 2 f 温度 +α 3 f 湿度 +α 4 f 土壤湿度 ;

[0141] Among them, α 1 , α 2 , α 3 , α 4 They are the weights of pest and disease characteristics, temperature, humidity, and soil moisture, respectively. The weight values ​​are set according to the statistical results of historical data to ensure that the contribution of different characteristics to the fusion results is reasonable;

[0142] S44, data fusion optimization: Based on historical data, the weight parameters in the fusion algorithm are optimized to ensure that the fused feature vector can more accurately reflect the comprehensive characteristics of pests and diseases. The optimization process includes:

[0143] (1) Use the cross-validation method to adjust the weights and evaluate the impact of different weight combinations on detection accuracy through the validation set;

[0144] (2) According to the optimization results, determine the optimal weight parameter combination to improve the accuracy of the fused data;

[0145] Through multimodal data fusion, the influencing factors of pest and disease characteristics and environmental data are combined, and a weighted fusion algorithm is used to achieve effective data fusion, thereby enhancing the accuracy and robustness of pest and disease detection and ensuring accurate pest and disease detection under different environmental conditions.

[0146] S5 includes:

[0147] S51, training of support vector machine model: using historical pest and disease data and environmental data, training the support vector machine model, the steps include:

[0148] (1) Prepare a training data set, which includes pest and disease feature vectors (such as color, shape, size, etc.) and corresponding environmental data (such as temperature, humidity, etc.), and label the pest and disease category of each sample;

[0149] (2) The radial basis kernel function (RBF kernel) is used as the kernel function of the support vector machine model. The kernel function formula is expressed as:

[0150] K(x i , x j ) = exp(-γ||x i -x j || 2 );

[0151] Among them, x i , x j is the input feature vector, γ is the kernel parameter, which controls the nonlinear mapping ability of the model;

[0152] (3) Adjust the regularization parameter C and kernel parameter γ of the support vector machine model and select the optimal parameter combination through cross-validation to maximize the classification accuracy of the classifier;

[0153] S52, classification process of support vector machine model: using the trained support vector machine model to classify the extracted pest and disease features, the steps include:

[0154] (1) Input the weighted fusion feature vector F 融合 =α 1 f 病虫害 +α 2 f 温度 +α 3 f 湿度 +α 4 f 土壤湿度 To the trained support vector machine model;

[0155] (2) The support vector machine model calculates the classification score of each pest category based on the input fusion feature vector, and the classification decision is expressed as:

[0156]

[0157] Among them, α i is the weight of the support vector, K(x i , x j ) is the kernel function, y is the classification result, indicating the category label of pests and diseases;

[0158] S53, verification and adjustment of classification results: verify the classification results of the support vector machine model, and further optimize the model through historical data to ensure the accuracy of the classification results;

[0159] The weighted fused feature vector is input into the support vector machine model, and the classification capability of the support vector machine model is used to accurately classify pests and diseases. The combination of multimodal features improves the accuracy and efficiency of classification, providing reliable classification results for subsequent pest and disease identification.

[0160] The S5 also includes:

[0161] S54, Identification of pest and disease types: Based on the classification results of the support vector machine model, the specific types of pests and diseases on the fruit trees are automatically identified. The steps are as follows:

[0162] (1) According to the output of the support vector machine model, a pre-built pest and disease database is matched. The database stores detailed information on common pests and diseases, including disease symptom descriptions, pest types, possible disease stages, etc.;

[0163] (2) Generate a pest and disease identification report based on the matching results of the database. The report content includes: the name of the pest and disease, the location of occurrence, the development stage and the possible degree of damage;

[0164] S55, Verification of pest and disease identification results: Verify the pest and disease identification results to ensure the accuracy and rationality of the identification. The specific steps include:

[0165] (1) Compare the identified pest and disease types with the current environmental data (such as temperature, humidity, soil moisture, etc.) to analyze whether the identified pest and disease meets the current environmental conditions;

[0166] (2) If the identified pests and diseases are inconsistent or unreasonable with the environmental conditions, readjust the parameters or weighted features of the support vector machine model and update the identification results to ensure the accuracy of the pest and disease type;

[0167] S56, generation of prevention and control measures: Generate targeted prevention and control suggestions based on the pest identification results. The steps are as follows:

[0168] (1) Extracting control measures information matching the identification results from the pest and disease database, including detailed information such as recommended pesticide type, application dosage, and application timing;

[0169] (2) The generated prevention and control suggestions are output in the form of a report, which lists the feasibility and expected effects of the prevention and control measures and provides implementation suggestions;

[0170] By matching the classification results of the support vector machine model with the pest and disease database, the accurate identification of fruit tree pests and diseases is ensured. The identification results are verified in combination with the current environmental data, and targeted prevention and control measures are generated based on the identification results, thereby effectively improving the intelligence and precision of pest and disease management.

[0171] S6 includes:

[0172] S61, threshold setting rules: based on the type, variety, and growth stage of the fruit tree, set the thresholds for different pest and disease detection. The specific steps are as follows:

[0173] (1) Establish a standard threshold library for pests and diseases based on the types of fruit trees (such as apple trees, pear trees, peach trees, etc.), and each type of fruit tree has different threshold ranges for different pest and disease detection indicators (such as diseased area, number of pests, etc.);

[0174] (2) Dynamically adjust the threshold of pest and disease detection according to the growth stage of the fruit tree (such as seedling stage, flowering stage, fruiting stage, etc.), so that the threshold changes of pest and disease detection in different growth stages are consistent with the actual growth conditions of the fruit tree;

[0175] (3) Using machine learning algorithms, we can optimize the accuracy of threshold adjustment by analyzing historical pest and disease detection data, so that the threshold is more consistent with the actual performance of fruit tree pests and diseases;

[0176] S62, dynamic threshold adjustment mechanism: real-time monitoring of the growth environment and pest and disease development of fruit trees, and dynamic adjustment of detection thresholds based on data changes, including:

[0177] (1) Automatically adjust the threshold based on real-time collected environmental data (such as temperature, humidity, soil moisture, etc.) and pest and disease detection results;

[0178] (2) Using historical detection data and current environmental conditions to predict the possible development trend of pests and diseases, and dynamically adjust detection thresholds to improve early warning capabilities for potential pest and disease outbreaks;

[0179] (3) Automatically lower the detection threshold as pests and diseases develop to ensure that diseases are detected and intervened at an early stage;

[0180] By combining the characteristics of fruit tree types, varieties and growth stages, the thresholds for pest and disease detection are dynamically adjusted. By combining real-time data with historical data, the accuracy of detection and early warning capabilities are optimized, ensuring that fruit tree pests and diseases can be effectively identified at the appropriate time.

[0181] The S6 also includes:

[0182] S63, generation of pest and disease diagnosis report: based on the classification results of the support vector machine model and the dynamically adjusted threshold, combined with the suggestions provided by the expert system, a pest and disease diagnosis report is generated. The specific steps are as follows:

[0183] (1) Compare the classification results of the support vector machine model with the dynamically adjusted pest and disease detection threshold, identify pest and disease information that exceeds the threshold, and generate a preliminary pest and disease diagnosis report, including the type, severity, and affected fruit tree parts of the pest and disease;

[0184] (2) Combined with the rules and knowledge base in the expert system, further analyze the expansion risk and possible development trend of pests and diseases, and finally form a complete diagnostic report, which includes the potential impact of pests and diseases, recommended countermeasures and follow-up observation points;

[0185] S64, generation of prevention and control measures: Based on the pest diagnosis report and in combination with the expert system, generate targeted prevention and control measures. The specific steps are as follows:

[0186] (1) Based on the type, severity and development trend of pests and diseases reported in the diagnosis report, the expert system calls the corresponding control solution library to generate specific control measures, including the type of drug to be used, dosage, timing of application and operation guidelines;

[0187] (2) In the event of a serious or emergency situation involving pests and diseases, the system will automatically generate warning information and, combined with the prevention and control operations recommended by the expert system, send warning notifications to relevant personnel to prompt emergency treatment;

[0188] (3) Personalize pest and disease control recommendations, optimize control measures based on the growth stage of the fruit trees, local environmental conditions, and historical pest and disease data, and ensure that the plans are highly consistent with the actual conditions of the fruit trees;

[0189] By combining the classification results of the support vector machine, dynamically adjusted thresholds and expert systems, pest and disease diagnosis reports and personalized prevention and control measures are generated to ensure accurate diagnosis and effective prevention and control of fruit tree pests and diseases.

[0190] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0191] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for detecting fruit tree pests and diseases based on machine vision, characterized in that: The following steps are involved: S1, image acquisition: Use a multispectral camera to collect images of different parts of the fruit tree. The collected images include visible light images, infrared images and ultraviolet images. The complete surface information of the fruit tree is obtained through multi-angle shooting; S2, image preprocessing: preprocess the collected images, including denoising, illumination correction and image alignment, and perform image enhancement through a deep learning model; S3, Pest and disease feature extraction: Use convolutional neural network to extract pest and disease features from preprocessed images and construct feature vectors. The features include lesion color, shape, size and pest distribution; S4, multimodal data fusion: using multimodal data fusion algorithm to fuse pest and disease characteristics with environmental data; S5, pest and disease classification and identification: Based on the extracted feature vectors, the support vector machine model is used to classify pests and diseases, and the types of pests and diseases on fruit trees are automatically identified according to the classification results; S6, dynamic threshold judgment and decision-making: Dynamically adjust the threshold of pest and disease detection for different fruit tree types and growth stages, combine the suggestions provided by the expert system, generate a pest and disease diagnosis report, and provide corresponding prevention and control measures.

2. The method for detecting fruit tree pests and diseases based on machine vision according to claim 1, characterized in that: The S1 includes: S11, multispectral camera installation and calibration: fixedly install the multispectral camera at different positions around the fruit trees and calibrate the camera; S12, visible light image acquisition: using a visible light sensor to acquire images of leaves, fruits and trunks of fruit trees; S13, infrared image acquisition: using infrared sensors to collect thermal images of fruit trees and detect the difference in thermal radiation between fruit tree leaves and fruits; S14, ultraviolet image acquisition: using an ultraviolet camera to acquire images of the surface of the fruit tree in the ultraviolet band, and detecting the fluorescent reflection characteristics of the diseased area on the surface of the fruit tree; S15, image data integration and transmission: the collected visible light, infrared and ultraviolet image data are integrated and transmitted in real time to a preset image preprocessing module through a preset wireless transmission unit.

3. The method for detecting fruit tree pests and diseases based on machine vision according to claim 2, characterized in that: The S2 includes: S21, image denoising: performing denoising on the collected image to eliminate random noise in the image; S22, illumination correction: performing illumination correction for images under different illumination conditions; S23, image alignment: aligning images at different angles; S24, Image enhancement: Enhance the preprocessed image through a deep learning model.

4. The method for detecting fruit tree pests and diseases based on machine vision according to claim 3, characterized in that: The S3 includes: S31, Feature extraction model design: Design a feature extraction model based on convolutional neural network to extract pest and disease related features from preprocessed images; S32, lesion feature extraction: extracting lesion features from the feature map output by the feature extraction model, where the lesion features include color, shape and size; S33, pest feature extraction: identifying the distribution of pests present in the image; S34, feature vector construction: Integrate the lesion characteristics and pest distribution characteristics into a feature vector for classification and identification.

5. The method for detecting fruit tree pests and diseases based on machine vision according to claim 4, characterized in that: The S4 includes: S41, environmental data collection: collect environmental data through sensors installed around the fruit trees; S42, multimodal feature data standardization: standardize pest and disease characteristics and environmental data; S43. Multimodal data fusion algorithm: A multimodal data fusion algorithm based on weighted fusion to fuse pest and disease characteristic data with environmental data; S44, data fusion optimization: Optimize the weight parameters in the fusion algorithm based on historical data.

6. The method for detecting fruit tree pests and diseases based on machine vision according to claim 1, characterized in that: The S5 includes: S51, training of support vector machine model: using historical pest and disease data and environmental data to train the support vector machine model; S52, classification process of support vector machine model: using the trained support vector machine model to classify the extracted pest and disease features; S53, verification and adjustment of classification results: verify the classification results of the support vector machine model.

7. The method for detecting fruit tree pests and diseases based on machine vision according to claim 6, characterized in that: The S5 further includes: S54, Pest and disease type identification: Based on the classification results of the support vector machine model, the specific types of pests and diseases of fruit trees are automatically identified; S55, verification of pest and disease identification results: verifying the pest and disease identification results; S56, generation of prevention and control measures recommendations: Generate targeted prevention and control recommendations based on the pest and disease identification results.

8. The method for detecting fruit tree pests and diseases based on machine vision according to claim 1, characterized in that: The S6 includes: S61, threshold setting rules: setting thresholds for different pest and disease detection based on the type, variety, and growth stage of the fruit tree; S62, dynamic threshold adjustment mechanism: real-time monitoring of the growth environment and development of pests and diseases of fruit trees, and dynamic adjustment of detection thresholds based on data changes.

9. The method for detecting fruit tree diseases and insect pests based on machine vision according to claim 8, characterized in that: The S6 further includes: S63, pest and disease diagnosis report generation: Generate a pest and disease diagnosis report based on the classification results of the support vector machine model and the dynamically adjusted threshold value, combined with the suggestions provided by the expert system; S64, generation of prevention and control measures recommendations: Based on the pest and disease diagnosis report and combined with the expert system, generate targeted prevention and control measures recommendations.

Images