Research On Image Recognition Of Tea Plant Diseases Based On Convolutional Neural Network

Tea is one of the most consumed drinks in China,which is popular among the people.With climate change and the increase of resistance to tea diseases,tea diseases have become one of the biggest threats to tea production.Therefore,it is very necessary to identify tea plant diseases accurately and quickly and put forward treatment measures.In the past,in the process of tea production practice,it is generally identified by the naked eye and experience of tea farmers,but this method has its limitations.Once the identification is not accurate,it will delay the best time to deal with the disease,resulting in the reduction of yield and economic losses.In this study,four kinds of common tea plant disease images in Hunan Province were collected,and the data set of tea plant disease was established after preprocessing.Based on deep learning and convolution neural network algorithm,the tea plant disease image recognition was completed,and the tea plant disease recognition system was built,which can enable users to quickly and accurately identify the tea plant disease and understand the disease information,which has strong practical significance.The main contents and the work of this paper are as follows:(1)aiming at the current open tea disease data set,taking the tea plant diseases with high incidence rate in Hunan as the object,we use Python Scrapy crawler crawling the agricultural technology website and the camera to capture the tea disease picture in two ways to collect pictures.After manual cleaning and image preprocessing,the image data set tpdid was constructed(2)Image recognition of tea plant diseases based on convolution neural network algorithm.RESNET network and efficient network,which are good in the field of image recognition,are selected to build the tea plant disease recognition model.The recognition accuracy of the two models for the data set of tea plant disease is83.48% and 88.72% respectively.In order to further improve the accuracy of tea plant disease recognition,an integrated network model of tea plant disease recognition,embedding,was designed based on DML deep mutual learning strategy.Resnet-50 and efficientnet-b4 were trained at the same time.Efficientnet-b4 learned each other’s parameters and optimized its own network to obtain the embedding network model,The image recognition accuracy of tea plant diseases was improved to 93.31%(3)With the development of information technology and the popularity of intelligent mobile terminals,the tea plant disease identification system based on wechat applet is further developed.The front-end display uses wechat applet frontend technology and web bootstrp framework,the data layer uses My SQL as the database,and the back-end uses tornado framework for development.The system can quickly and accurately identify tea plant diseases and return the identification results to the front-end display of wechat small program,which realizes the online identification of tea plant diseases and promotes the development of agricultural informatization of tea industry.