Research On Apple Leaf Disease Recognition Based On Deep Learning

Apple has become one of the four major fruits in China due to its rich nutritional and medicinal value.The quality of apple leaves is one of the keys to ensuring the growth of apples.However,apple leaf diseases will have a negative impact on the yield and quality of apples,while cause significant production and economic losses to fruit farmers.Therefore,rapid and accurate identification of apple leaf diseases is of great significance.Traditional leaf disease identification methods are mainly based on manual detection,which has problems such as high cost,low efficiency and inability to achieve automation.The methods based on image feature engineering also have many problems,such as large workload,relying on human experience,and low recognition accuracy.For these problems,this paper has carried out research on apple leaf disease recognition based on deep learning,which not only reduces the image pre-processing process,but also automatically extracts image features and has a good recognition effect.The main research work of this article has the following two parts:(1)In order to solve the problems of small training set that cannot meet the training needs,huge amount of training parameters and long training time,this paper proposes an apple leaf disease identification method based on the improved VGG16 network and migration learning.In order to reduce the complexity of the model,this work removed the three fully connected layers in the classic VGG16 model and replaced them with a global average pooling layer,batch normalization layer and fully connected output layer.In order to reduce training time and improve recognition accuracy,this paper uses the idea of migration learning to pre-train the model using the Image Net data set and initialize the improved VGG16 network parameters.Finally,an identification experiment was carried out on the apple leaf disease data set and compared with other methods.The experimental results show that the method used in this paper has achieved the best results in the disease identification of apple leaves.(2)In order to reduce the dependence of the deep network model on the amount of data,in view of the case where the number of training samples is only single digits,this paper further studies the method of using very few sample data for effective disease identification.In this paper,a new network model combining convolutional neural network and graph convolutional network is used,and it has been successfully applied to small sample apple leaf disease recognition.The model first uses a convolutional neural network for feature extraction,and each sample feature and corresponding code is input to the graph convolutional neural network as a graph node,while the graph nodes are guaranteed to be connected in pairs.This article uses graph convolutional neural network to learn the structure embodied in the sample set.The final experimental results show that when there are 10 training samples,the method proposed in this paper can achieve a recognition accuracy of 87.88%.Even if there is only one training sample,the recognition accuracy rate of 65.32% can still be obtained,which has high practical application value for the prediction and recognition problem in the case of very small samples.