Identification Of Corn Grains Varieties Based On Deep Learning

As one of the two golden belts of corn cultivation in the world,China has always been the world’s major corn producer and consumer.The research of corn grains variety identification is of great significance to corn breeding research,cultivation management and market circulation.Previous researchers have done a lot of research on corn grains variety identification and integrity recognition by using traditional machine learning methods,and achieved some results.However,traditional machine learning methods need to summarize laws and design features quantity manually,which will cost a lot of time and energy in image pretreatment and evaluating the effectiveness of extracted features.With the continuous development of science and technology,deep learning can automatically learn the characteristics of images,and it has excellent performance in image recognition tasks,making it widely used in various fields.In this study,Denghai 518,Jundan 20 and Zhengdan 958 were selected as the research objects,and a corn variety recognition method based on deep learning was proposed.The main research is as follows:(1)Introduced the research status at home and abroad,mainly including:corn grains variety identification and deep learning technology research.The theoretical basis of deep learning is expounded,and the relevant knowledge of Convolutional Neural Network is introduced in detail.(2)In view of the lack of open source data sets in the agricultural field,the production of corn grains data sets was completed.Use the scanning function of the multifunctional all-in-one to obtain digital images of corn grains,expand the data set using data enhancement technology,and mark the three types of Denghai 518,Jundan 20 and Zhengdan 958 as 0,1,2 respectively.(3)In view of the shortcomings of traditional machine learning in manual feature extraction,a method of corn variety recognition based on convolution neural network is proposed.This method refers to Alex Net and builds a Convolutional Neural Network model under the Keras deep learning framework.In order to verify the effectiveness of the convolution pooling structure to extract local features,a Multilayer perceptron network model is designed as a comparative experiment.The recognition rate of the two models is 95%and 74.26%respectively,which proves that the Convolution Neural Network can effectively learn the characteristics of corn grains and achieve better recognition effect.(4)Aiming at the problems of complex depth model structure,large parameter scale and difficult to fit small data set,a method of corn variety identification based on Transfer learning was proposed.Based on the structure and weights of the migration VGG-19,Inception-V3,and Rest Net-50 Convolutional neural network models,a Transfer learning model is constructed,and the training set of the corn grains data set is used for training.The test set is used to test the fully trained model.The results show that the average recognition rate of the three migration models is higher than 99%,which is more than 4 percentage points higher than the Convolutional neural network model.There is a higher recognition accuracy.