Research On Corn Disease Recognition System Based On RESNET Network

In China,corn,as the most widely planted and the highest yield crop,plays an important role in the national economy and is an important food crop and feed crop.However,in recent years,due to the change of cultivation system,imperfect crop health care measures,variation of original pathogen varieties and many other reasons,the varieties of maize leaf diseases increased and the disease degree increased.These problems have become a difficult problem of restricting corn production,but also brought great economic losses to farmers.Therefore,more and more attention has been paid to the identification and treatment of maize leaf diseases.In the past,people used the visual observation method in the process of corn disease identification,which mainly depends on the experience accumulated by agricultural technicians for a long time.This method requires agricultural technicians to judge one by one,which is inefficient and low accuracy,and human subjective judgment plays a major role in the process of recognition.Therefore,for the growing agricultural economy,this method has been unable to meet the demand.In addition to the traditional recognition methods,there are many machine vision based technologies,which first need to recognize the image.However,the discrimination method based on image recognition has its inherent disadvantages.First,the process is cumbersome,and all images need to be normalized to ensure the consistency of the size and size of the sample data.Second,the error is large.Because relying on subjective judgment will inevitably affect the accuracy of the results.In this paper,the recognition of maize leaf diseases is realized based on RESNET network of deep learning.It mainly includes the following work:1.Convolution neural network can effectively extract image features through convolution and pooling.In this paper,three groups of convolution neural network experiments are set up to construct corn disease model through deep convolution neural network.In order to fully verify the effectiveness of transfer learning and data enhancement in corn disease recognition task,the first group of experiments used to build a common CNN model to explore the accuracy of corn disease recognition under the shallow common neural network;the second group of experiments used the transfer learning model based on resnet50 to explore whether the deeper network can achieve better results in corn disease recognition The third group of experiments uses the transfer learning model based on resnet101,and finally obtains the final results by comprehensively comparing the accuracy of the three models in the test set and the change of loss curve in the process of model training,and constructs the best corn disease recognition model.2.In view of the lack of experience of disease identification for ordinary users,in order to reduce the wrong judgment caused by professional short board,weaken the economic loss caused by identification stage to the greatest extent,and reduce the error caused by human judgment.This paper designs and develops a corn disease identification system based on resnet50.Through this system,ordinary farmers can quickly and accurately get different kinds of corn diseases,improve the accuracy of disease prevention and identification of ordinary agricultural personnel in the planting process,and it is of great significance to help relevant technical personnel to carry out complex research work.