The Application Of Recognition And Classification Of Agricultural Diseases And Pests Image Based On The Depth Of The Convolution Network

With the development of modern digital technology,as an agricultural country,these traditional agricultures should be changed to the direction of digital agriculture in China.Such as,in the process of farming,some useful information of crop diseases should be achieved and can be analyzed,so some accurate and effective actions can be done.Farmer workers can spray pesticide with dose precision in a strict region,and ensure the healthy parts of crops do not accept pesticide spraying.To ensure the pesticides ca n be sprayed efficiently,and the whole growth process of crops planting can be detected digitally,high production and quality can be achieved,and realize digital management of agriculture planting.In the thesis,the soybean leaves as the research objec t.Because these soybean leaves are diseases easily by all kinds of diseases and insect pests in the soybean planting,especially mould and bacterial spot mildew are the most causes to decrease the production of soybean.How to find these two kinds of plan t diseases and insect pests,and how to make accurate and effective measurements quickly is the most important way.First,these relevant agricultural information technologies and algorithms were introduced in the first chapter.And a lot of information ab out agricultural information technologies and algorithms in China and abroad were analyzed in the article.Second,with the rapid development of deep learning technology,the convolution neural network has certain advantages in the field of image classification and recognition.The convolution neural network technology can be used in the application of agricultural plant diseases and insect pests.The development of the convolution neural network was introduced in the paper,including the history of convolution of the neural network,study significance,working principle and these important technologies of it.Thirdly,how to distinguish crop health region segmentation and plant diseases and insect pests were introduced in the pattern recognition field.Som e the basic algorithms and theory were introduced in the paper,including image preprocessing,image enhancement algorithm,feature vector extraction method(principal component analysis and independent principal component analysis),the relevant optimization algorithm,feature wavelength extraction method and support vector machine(SVM)algorithm.These experiments of the related algorithm can be done by the actual samples,some results can be compared and analyzed.Finally,based on traditional machine learning algorithms,the theory of depth of the convolution neural network can be used in soybean identification of plant diseases and insect pests.Because the different sizes of the input data,the different network structure can be designed,so the segment ation and recognition results are obtained.And compared the different of results can be compared between the traditional classification method and depth of the convolution neural network,found that deep learning algorithm is more effective.