Research And Implementation Of Intelligent Diagnosis System For Rice Field Diseases And Pests Based On Android And Deep Learning

There are various species of diseases and pests in rice fields,which are easily confused because of the similarity and difference among species.In addition,different diseases and pest types correspond to different control methods.Therefore,it is an important premise to ensure the real-time and accuracy of identifying rice pests and diseases.At present,the diagnostic methods for diseases and pests in rice fields mainly rely on manual identification,which has the disadvantages of strong subjectivity and poor real-time.At present,there are few taxonomists of diseases and pests.In addition,pest forecasting technicians and farmers lack of professional knowledge.Therefore,it is urgent to develop a quick and convenient intelligent diagnosis tool for rice diseases and pests.For the above problems,in this paper,49 kinds of rice pests and 47 kinds of diseases are studied,and an intelligent diagnosis system of rice pests and diseases based on Android and deep learning is developed.The system provides a convenient and accurate automatic diagnosis tool for pest forecasting technicians and farmers.The main research contents and results are as follows(1)The Algorithm of image recognition for rice field diseases and insect pests is studied.In this paper,49 kinds of pests and 47 kinds of diseases in rice fields are studied.Under the framework of Caffe,the migration learning method is used to fine tune the convolutional neural network parameters.In addition,five popular deep learning models are trained,including CaffeNet,GoogleNet,VGGNet,ResNet,and DenseNet.The best model for identifying diseases and pests in rice fields obtained on the test set is the DenseNet model.The accuracy of identifying pests and diseases is 94.8%and 91.4%,respectively(2)An APP for intelligent diagnosis of diseases and pests in rice fields is developed.The client APP is mainly composed of five functional modules.It includes user registration,diseases and pests information inquiry,diseases and pests location information,intelligent diagnosis of diseases and pests and remote expert identification.After signing up for an account,the user can log into the system swiping the screen or searching keywords.In this way,the user can browse and inquire the biology of diseases and pests and control information.Opening the map,the user can view the geographical distribution of pests and diseases.With mobile phones,the user can take pictures or select images in the album.Then the system can call the corresponding pest/disease deep learning network model to achieve the automatic identification of diseases and pests.If the user does not agree with the result,the expert can be requested to carry out remote identification(3)A Tomcat application server is set up.The function of data transmission and reception is implemented by writing Servlet.The MySQL database is used and the function of database operation and management is achieved by JDBC.Caffe deep learning framework and recognition model are deployed on the server side.The local logical business code is exported as a war file,which is packaged and deployed on the Ali cloud server.The user can get the diagnosis result within 1.5s after uploading the rice fields diseases and pests images.The intelligent diagnosis system for rice fields diseases and pests based on Android and deep learning consists of several functions.It includes information inquiry of 49 kinds of pests and 47 kinds of diseases in rice fields,automatic identification,image recognition history of diseases and pests,diseases and pests location information,and remote expert identification.The system provides a tool for farmers and pest forecasting technicians.On the one hand,it can identify agricultural pests conveniently and accurately.On the other hand,it can provide users with customized pest control information.Furthermore,experts do not need to go to the field to identify pests and diseases,which greatly saves money and time.