Research On Classification Of Crop Leaf Diseases Based On Convolutional Neural Network

Crop disease is one of the most important challenges in agricultural development.Quick and accurate identification of crop diseases is conducive to early development of treatment programs,and can greatly reduce agricultural economic losses.Nowadays,the use of computer technology to accurately and quickly identify crop diseases is a key factor to ensure the yield and quality of agricultural products,and it is also an important means to promote agricultural modernization.On the basis of highly developed information technology,big data technology has opened up new possibilities for agricultural development.The essence of big data technology is to collect data provided by agricultural production on a large scale,uses the corresponding data to analyze the model,and ultimately serves for increasing agricultural production.Deep learning technology is to automatically mine favorable information from a large amount of data and apply it to actual production needs.Applying deep learning models to crop leaf disease recognition greatly reduces the workload of hand-designed features,making intelligent identification of crop leaf diseases possible.This dissertation proposes a classification method for crop leaf diseases recognition based on convolutional neural network,taking leaf disease images in apple,corn,grape,peach and other common crops as research objects,and further applying convolutional neural network in crop leaf diseases recognition.The main research contents are as follows:(1)An identification method for apple leaf diseases based on convolutional neural network is proposed,with improved and optimized classic VGG16 model.The model adopts a batch normalization layer after each convolutional layer to speed up the training speed.Moreover,a loss function combining Center loss and Softmax loss is used to make the learned deep features more distinguishable.The improved model has great superiority in accuracy,and the recognition rate is as high as 97.58%.(2)A method for identifying crop leaf disease levels based on improved convolutional neural networks is proposed,in order to take better control measures for crop diseases according to the degree of diseases.First,the selected leaf image data set is pre-processed,including the steps of illumination processing,image segmentation,and extraction of diseased pixels.Then the pretreated diseased leaves are divided into different levels according to the severity.Moreover,Focal loss function and Adam optimization algorithm are used to optimize the model and improve the recognition accuracy.The accuracy rate of crop leaf disease grade identification based on deep learning reaches up to 95.61%.