Research On Rice Seed Quality Detection Combined With Deep Learning And Hyperspectral Imaging Technology

Rice is the most important food crop in China.The variety and vigor of rice seeds can affect the yield and quality of the grain.Existing methods for testing the quality of rice seeds have disadvantages such as cumbersome operation,low efficiency,and damage to the seeds.Near-infrared hyperspectral imaging reflects the internal chemical composition of rice seeds.The internal chemical composition of rice seeds differs between qualities.Therefore,the use of NIR hyperspectral imaging technology can detect the quality of rice seeds quickly and non-destructively.This study investigates the nondestructive detection of two important quality indicators of rice seeds,variety and vigor,and proposes a nondestructive detection method for rice seed vigor and variety combining near-infrared hyperspectral imaging technology and deep learning methods.The main research of this paper is as follows:(1)This study used near-infrared(NIR)hyperspectral technology with conventional machine learning methods(support vector machine(SVM),logistic regression(LR),and random forest(RF))and deep learning methods(Le Net,Goog Le Net,and residual network(Res Net))to establish variety identification models for five common types of rice seeds.Among the deep learning methods,the classification accuracies of most models were higher than 95%.This study further used the deep learning methods to establish variety identification models for 10 varieties of rice seeds without considering their types.Among them,the Res Net model had the best classification results.The classification accuracy on the test set was 86.08%.This study used the saliency map method to visualize each convolutional neural network(CNN)model to find the band region that contributed the most to the data.The results showed that the bands with the largest data contribution were mainly concentrated at approximately 1300-1400 nm and secondarily concentrated at approximately1050-1250 nm.The overall results showed that NIR hyperspectral imaging technology combined with deep learning could effectively distinguish rice seeds of different varieties.This method provided an effective way to identify rice seed varieties in a quick and nondestructive manner.(2)In this study,near-infrared hyperspectral imaging was used to determine the viability and vigor of naturally-aged rice seeds.Standard germination test was conducted to determine the reference values of the viability and vigor.Convolutional neural network(CNN)and conventional machine learning methods(support vector machine(SVM)and logistic regression(LR))were built using full range spectra and characteristic wavelengths selected by principal component analysis(PCA)to predict the viability and vigor of different varieties of rice seeds under natural aging conditions.The overall results showed that deep learning methods and conventional machine learning methods could predict the viability and vigor of different varieties of rice seeds well,and the accuracy of most models was over 85%.Models using full spectra and the characteristic wavelengths showed close results.Models on all varieties performed closely to those on single variety.This study provided an effective method for fast,nondestructive,and efficient prediction of rice seed viability and vigor.