Identification Method Of Rice Disease Based On Bayesian Network

As the main food crop in the world,rice is influencing people’s life.The yield and quality of rice is the concern of the national people.The disease of rice is the key to yield loss.How to quickly find and implement effective control of rice disease is an urgent problem.At present,our method of identifying rice diseases is still based on people’s subjective judgment.This method has high requirements for experience,low efficiency and no real-time function.With the progress and development of modern economy and society,the wide popularization of computer and the update of technology,artificial intelligence has been applied and developed rapidly in various fields.How to analyze and process the data collected by these intelligent devices,especially the images and data,to obtain useful information,and effectively and accurately deal with rice diseases,is an urgent problem for rice growers.Therefore,in this study,the most common and the highest incidence of rice diseases(rice blast,sheath blight,bacterial blight)were selected as the research objects,and the identification methods in the natural environment were proposed:First of all,this paper briefly introduces the basic theoretical knowledge of machine learning algorithm and image processing,which provides a set of corresponding technical route and ideas to solve the problem for the later research of rice disease recognition algorithm.Then,in the research of rice biological disease recognition algorithm,a machine learning method is used to recognize rice diseases.In this paper,starting from the analysis of rice sheath blight,rice blast and bacterial blight and other disease spots,we collected the disease spot images by taking photos with smart phones,and then used graying,image denoising,disease spot segmentation and other operations to predict the disease images of various types of rice,segmented the disease spot images of various types of rice,and established a corresponding rice leaf disease spot database.Then,according to the external characteristics of different types and lesions,the three principal components and disease characteristics need to be selected.Starting from the color,shape,texture and other factors,the corresponding disease characteristic parameters are extracted,a total of 63.In order to effectively reduce the noise and lengthy data processing time of the image,the principal component analysis method is used,The feature parameters of disease were optimized from 63 to 40.Finally,BP(Back Propagation)neural network classifier and Bayesian network classifier are used to establish rice disease recognition model.The optimized color feature parameter set,shape feature parameter set,texture feature parameter set and the combination of these three feature parameter sets are used to establish the model and analyze the classification and recognition effect of the two classifiers.The results show that Bayesian classifier has better classification effect.Furthermore,convolution neural network was used to optimize Bayesian network algorithm,and the recognition accuracy of color feature,shape feature and texture feature of three groups of rice diseases was improved to98%.The recognition effect is good,and Bayesian network can be effectively applied to rice disease recognition.