Analysis And Research On Rice Disease Recognition Method Based On Deep Learning

Rice is one of the most important food crops in China and even the world.However,with the continuous changes in human activities,the quality of climate,soil,and water sources have also changed,and the disease of rice has become more and more serious.Traditional artificial pests and diseases identification methods can no longer adapt to the emergence of a large number of diseases.Artificial visual recognition increases the uncertainty of the identification results and cannot solve the problem in a targeted manner.This will not only not cure the disease,but even achieve a half-effective result.In the incidence of rice diseases,rice blast,rice smut,and bacterial blight have the highest incidence,the most damage,and the most representative.Therefore,the research in this article focuses on the above three categories.In recent years,machine learning has made great progress in crop applications,and image recognition technology based on convolutional neural networks has matured.Using this method for disease identification not only accelerated the integration rate of the agricultural industry and scientific and technological intelligence,but also improved the accuracy of rice disease identification.This experiment conducted further research on the identification of rice diseases.First,we collected sample pictures of rice blast,rice smut,and bacterial blight.Due to the difference in the distance and light of the sample photos,the size and angle will be biased.Therefore,it is necessary to unify the specifications of these images by some means to improve the efficiency of network model recognition.Neural network recognition needs to absorb many sample pictures to classify and learn features.The main research objects of this article are rice blast,rice smut,and bacterial blight.Therefore,this article also expands the data set for such diseases,and unifies the specifications through size cutting,angle change and vertical symmetric mirroring.A new network model based on deep learning was built to realize parameter initialization design.The accuracy of the rice disease recognition model built at the beginning cannot meet the actual requirements.In order to upgrade the model in-depth,this experiment increased the entry point for analysis and research,integrated four parameters of iteration number,batch size,learning rate and optimization algorithm,and strived to optimize the experimental results.In this study,the confusion matrix is selected as the evaluation standard,and the two highly reference network models,VGG and Res Net,are used for horizontal comparison,and more objective and reference value experimental results are obtained.The research results show that the accuracy rate of the optimized model identification is 98.64%,which has reached the goal of accurately identifying the disease.After data integration and model optimization,a rice disease intelligent identification platform was established.This platform designed and realized the function of uploading rice disease samples and identifying functions.When a disease occurs,the grower uploads a picture through the platform to get the diagnosis result in the first time,so that it can more quickly carry out the next step of the treatment of the disease,and it also reflects the technical characteristics of disease identification.