Research And Application Of Fruit Leaf Disease Identification Based On Convolution Neural Network

Crop diseases constitute the major threat to the agricultural production of orchards.Reduced yields due to illness can result in immeasurable economic losses.This makes it very important to quickly identify crop diseases.Among all kinds of detection methods,the detection method based on computer vision has the advantages of rapidness,accuracy,portability,and has gradually become the focus of crop disease detection research.However,traditional machine learning algorithms must manually extract the image characteristics of plant leaves,making it difficult to meet the needs of modern intelligent agricultural production.In this paper,by studying image recognition and Convolutional Neural Network model,an algorithm of plant leaf disease recognition based on Convolutional Neural Network is proposed,which can be used to identify many different types of plant diseases.The main methods and innovation of the thesis are as follows.Firstly,we propose a two-stage method based on Convolutional Neural Network to realize multi-class leaf disease recognition.This method uses ensemble learning to improve recognition performance,which effectively improves the recognition accuracy compared with the single network structure.This paper also proposes a new dynamic activation function,which is an improvement of Re LU(Rectified Linear Unit)and can make Convolutional Neural Network more robust.Then,to realize plant disease identification on mobile devices,a lightweight Convolutional Neural Network is designed in this paper,which can greatly reduce the computational resources of disease identification while maintaining relatively reliable accuracy.Experimental results show that compared to the current advanced Convolutional Neural Network,the lightweight Convolutional Neural Network is at a state-of-the-art level in terms of precision and recognition speed.Finally,in order to make the research more in line with the actual production needs,this paper built a plant leaf disease recognition platform based on in-depth learning,and developed the corresponding image preprocessing system.In this paper,the plant disease identification method based on Convolutional Neural Network was tested on 60 kinds of data sets of 10 plants,and the average accuracy rate was93.66%.In addition,the lightweight Convolutional Neural Network designed to improve the speed of disease recognition improves the recognition speed to 5 milliseconds.In practical application,the identification platform of plant leaf disease constructed in this paper is simple and easy to use,and it realizes the expected identification and prevention of diseases in orchard area.