| Peanut is an important cash crop.In the traditional planting process,peanuts are very vulnerable to the threat of disease.In order to reduce the threat of disease and improve crop production,growers often spray and irrigate a large number of pesticides.This will result in a large number of pesticide residues and environmental pollution.With the continuous development of the times,reducing the impact of diseases,reducing the use of pesticides and improving the quality of crops are the pursuit of modern people for food safety.How to detect lesions early and classify them scientifically is an important basis to reduce and improve the efficiency of drug use.The promotion of digital agriculture is the key to solve such problems.In this paper,computer vision technology and image processing technology were used to select two common diseases in peanut cultivation: black spot and brown spot.A relatively complete technology of automatic collection,preprocessing and segmentation of lesion area is formed..which can collect,pretreat and segment the lesion area.By extracting the color,shape and texture features of the lesion area,the extraction of feature parameters and the construction of discriminant model were realized;The leaf spot disease of peanut was classified by the method of support vector machine SVM;the optimal penalty factor coefficient was found by particle swarm optimization algorithm and the variance in the kernel function was determined by cross-validation method to improve the accuracy of disease classification.In this study,computer vision technology was used as a basic condition to realize real-time on-line detection,identification and classification of peanut leaf diseases in the field.It is of guiding significance to the discovery,classification and control of diseases in peanut cultivation.The main research work of this paper includes the following parts:1.Acquisition and preprocessing of image.Data acquisition is carried out by the data acquisition device,and then the data signal is transmitted to the computer.The computer first saves the original data,and then calls the image processing device.Since the original data is processed under natural light,firstly,the original data need to be denoised by adaptive median filtering.Then the color,texture,shape feature K value extraction and threshold segmentation are carried out.2.Improvement of Super-Green algorithm.Ultra-green algorithm is widely used in green crop imagesegmentation.In this paper,the traditional super green algorithm is improved according to the characteristics of peanut diseases.The improved super green algorithm can effectively obtain clear peanut disease images.The main steps are as follows: The characteristics of noise in peanut leaves are analyzed.Noise is generally divided into fine granular noise,salt and pepper noise and so on.Firstly,the adaptive median filtering algorithm is used to remove the fine granular noise and salt and pepper noise.Secondly,the other types of noise will be remove by filling selective holes.Finally,a clear peanut disease map is obtained by using MATLAB tools.It can effectively restore the peanut disease spots and provide guarantee for the accurate identification of the number of Peanut Diseases in the later stage.It can quickly and accurately identify and diagnose plant diseases,and is conducive to overcoming the subjectivity,experience and inefficiency of human visual recognition system.3.Classification of lesions.Particle swarm optimization(PSO)algorithm based on support vector machine(SVM)is used to classify the extracted feature data.Using this method,two kinds of lesions were successfully screened out and the correct recognition rate was very high.The higher the penalty factor is,the lower the tolerance is.The research content of this paper has practical significance and obvious application prospects for improving the real-time accurate identification of Peanut Diseases and promoting the development of digital agriculture in China. |
PDF Full Text Request