The Research And System Implementation Of Plant Disease Detection Algorithm Based On Deep Learning

Plant disease problems have been limited the healthy growth of plants for a long time.In order to solve the problem of plant diseases,a lot of research work have been done.In recent years,tasks(such as image classification,detection,and recognition)based on deep learning techniques have achieved good results.This paper has carried out research on plant disease detection algorithms and corresponding system development.In this paper,based on the detection of plant diseases,taking the detection of cucumber sick leaf as an example,the collected 1500 images of cucumber leaves including healthy leaves and Cucumber_Scorch as data sets,Aiming at the problem of classification and detection of Cucumber_Scorch under complex background,this paper proposes a dual network detection model based on deep learning,Through the model optimization and parameter training,the classification detection of the Cucumber_Scorch is realized.In order to compare with the dual network detection model,the classification method of VGG16 model and support vector machine are applied to the classification detection of cucumber death picture.The experiment results show that the detection accuracy of the proposed dual network model can reach 98.32%in the picture detection task of cucumber dead disease.Setting the configuration optimizer and the optimal parameters,the detection accuracy can be as high as 98.79%.Based on the above algorithm research,the plant disease detection system is developed.The system function is mainly composed of system interface module,system registration login module,disease detection module,visualization module and administrator module.The system interface module mainly provides the user with a disease detection button and measures for preventing and controlling common diseases;The system registration and login module mainly provides the administrator with registration and login entry;The disease detection module mainly performs disease detection on the uploaded pictures of the user;The visualization module is responsible for visualizing the disease detection results;The administrator module mainly completes the functions of updating the interface,collecting images,and viewing images.The main work and contributions of this paper are as follows:1?This paper constructs a dual network detection model based on deep learning.This model is an organic combination of two VGG16 network models.First,the network first initializes the network model parameters with the pre-trained VGG16 model on Imagenet,then extracts the convolution features from the training set and fine-tunes the model parameters,and then saves the parameters of the network one and initializes the parameters of the network two,and then The parameters of the network 2 are fine-tuned,and finally the trained model is saved,and the image is classified and predicted.2?The research of VGG16 model and SVM algorithm was carried out.The two algorithms are applied to classify and predict the cucumber death picture in complex background,and compared with the dual network detection model.The dual network detection model constructed in this paper has higher classification detection accuracy.3?The system development uses windows as the development platform,java as the development language,eclipse3.8 as the development tool,and open source MySQL to build the relational database.The plant disease detection system is independently designed and developed using Java web technology.