Study On Automatic Identification Of Cotton Leaf Diseases And Insect Pests Based On Convolution Neural Network

China is both a big cotton producer and a big cotton consumer,Cotton production is related to national economy and people’s livelihood,The whole growth cycle of cotton was attacked by many diseases and insect pests,Seriously affect its output,Therefore,it is very important to identify the species of cotton diseases and insect pests quickly and accurately,and to control them timely and accurately,so as to avoid the further spread of diseases and improve the yield of cotton.Because traditional cotton pest detection relies on plant protection experts or experienced farmers,It is timeconsuming and laborious,and the recognition accuracy of small lesions that are difficult to observe by human eyes is low,In addition,the traditional detection method needs complex artificial feature segmentation and extraction,which increases the labor cost,and the high computational complexity also causes the recognition effect is not ideal.In view of this,in this paper,the convolutional neural network(CNN)algorithm is applied to the automatic recognition of cotton leaf diseases and insect pests.On the basis of CNN algorithm,this paper studies Cotton Fusarium wilt,Verticillium wilt,cotton boll blight,brown spot,angular spot and cotton spider mite.In order to solve the problem that the image data set of cotton diseases and insect pests is small and easy to produce over fitting,the migration learning algorithm and data enhancement technology are introduced to solve the problem.Finally,an app is developed for automatic detection of cotton diseases and insect pests.The main research contents are as follows:(1)According to the data collected in this study,six CNN models with different convolution depths are designed to automatically extract the image features of cotton pests,and the Softmax classifier is used to identify the types of pests,and the depth network model suitable for the data set of cotton pests is explored.The experimental results of traditional support vector machine(SVM)and back propagation(BP)neural network are compared.(2)The network structure of 5 convolution layers,2 full connection layers and 1 Softmax classification layer was built based on the Alexnet model.The model was used to classify 6 diseases and insect pests in cotton leaves.This part of the experiment is divided into two parts.The first part uses the PlantVillage big data set to learn the pretraining model from the built model as the feature extractor and save the model,and then uses the migration learning method of model migration and fine-tuning parameters to train our model on the original cotton disease data set;the second part uses the data enhancement technology to advance the original cotton disease data set The first part of the experiment is repeated with the new data set instead of the original data set to get the final model.The experimental results of SVM and BP neural network and deep convolution neural network model(vgg-19 and Google perception V2)are compared.(3)In this paper,the final model of the research is transferred to Android device to make an app to realize the identification of cotton leaf pests.