The Image Recognition Method Of Major Diseases And Pests Of Mihu Cherry In Yiliang

In Yiliang,Kunming city,the Mihu Cherry Festival has become a name card of Yiliang County.However,the growth cycle of Daihong cherry tree is slow and the fruiting time is short.There are more than 40 kinds of diseases and pests of Midudai red cherry.Due to poor management and negligence of major diseases and pests before and after fruiting,many farmers suffer heavy losses every year.At present,the identification of diseases and insect pests of Midudai red cherry mainly relies on experience,and the error rate is high,which can no longer meet the demand of agricultural market.With the wide application of image processing technology,this paper,based on machine learning,takes Yiliang rice Hodai red cherry plant as the research object to carry out the identification research of unique diseases and pests.Convolutional neural network was used to realize intelligent recognition of cherry diseases and insect pests,so as to quickly diagnose diseases and timely control diseases and insect pests.The main research contents,methods and results of this paper include:(1)To establish the data set of diseases and pests of Yiliang Mi Husai red cherry.By going to cherry base to collect data,the detailed overview,category,main symptoms,harmful sites,pathogenesis,pathological data and other information of diseases and insect pests of Midai red cherry in Yiliang were collected through data analysis,data review and field investigation.(2)Study on image recognition algorithm of diseases and pests of Yiliang rice Husai red cherry.The identification research was carried out on 4 kinds of Daihong cherry diseases(brown spot,trunk mildew,anthracnose,bacterial perforation)and 4 kinds of insect pests(yellow moth,grylatoria,red necked longicorn,Chrysophora chrysophora)with high incidence in Mihu area of Yiliang.Through machine learning,VGGNet-16,Inception Net-V3 and Mobile Net-V2 lightweight models were used to train the collected data sets.(3)We need to ensure high identification accuracy and prevent data overfitting,the image data set is enhanced and the model is improved using Dropout.Finally,the accuracy rate of VGGNet-16 and Inception Net-V3 reached 93.7% and 94.8% respectively,which provided algorithm support for the disease and pest image recognition function in the subsequent system development.In order to facilitate outdoor use,mobile device mounting can be realized by transferring it to lightweight model Mobile Net.By changing Re LU function and reducing the inverted residual structure,the final accuracy rate reaches 93.9%,and finally the image recognition of diseases and insect pests can be realized.By comparison,Inception Net-V3 has the highest accuracy.In this paper,with the help of machine learning technology,the main diseases and insect pests of Yiliang Mizudai red cherry were identified,which is of great significance to the development of automation and informatization of Yiliang cherry industry.Although the main research object of this paper is Yiliang Mihudai red cherry,the model method used can be extended to the identification of many crops,which has a certain application reference value in improving agricultural efficiency.