Image Recognition Of Crop Diseases Based On Deep Learning

Crop diseases will affect the quality and yield of crops,and bring huge economic losses to growers.It is of great practical significance to identify crop diseases in time.The image recognition method of crop diseases based on deep learning overcomes the shortage of artificial recognition method and combines the advantages of both image recognition technology and deep learning technology,which has good practical value and application prospect.In order to improve the accuracy of crop disease image recognition and the generalization of the model,this paper studies the image recognition of crop disease based on deep learning,focusing on the image recognition of corn disease and tomato disease based on transfer learning,and the application of generative adversarial networks to corn disease data set.The main contents of this paper are as follows:(1)Based on transfer learning,a method for recognizing crop disease image in small sample data set was proposed,and experiments were carried out on maize disease data set and tomato disease data set respectively.In order to compare and analyze the effect of the same pre training model and training method on the accuracy of disease identification of different crops,this paper compares and analyzes the recognition results of 4 pre-training models and 3 training methods on 4 corn diseases and 9 tomato diseases.(2)In view of the difficulty of sample data collection and labeling of crop disease data set and the small amount of sample data,the model of generative adversarial networks in image translation is applied to crop disease data set.Based on the original data set,the pix2 pix model is used to generate new disease pictures,so as to expand the original data set and achieve the effect of data enhancement.(3)The experiment results of the original corn disease data set and the new corn disease data set composed of the corn disease pictures generated by the generative adversarial networks under different pre-training models and training methods are compared and analyzed,so as to evaluate the feasibility of using the generative adversarial networks to expand the data set and its impact on the recognition results.