Application Study Of Electronic Nose In Seed Viability Testing And Varieties Discrimination

The seed material is the basis of human survival and development. Seeds conservation at low-temperature genebank is the most important manner for germplasm preservation. But even in the low-temperature genebank, seed deteriorates gradually, which will result in seed death ultimately. Therefore, viability monitoring is one of the most important activity at genebanks. Another issue for genebank curators is to identify homonymous seed accessions. The classical methods used for detecting seed viability and identifing homonymous seed accessions are time consuming and seeds consuming. Therefore, to establish a rapid, nondestructive method for detecting seeds viability and identifing homonymous seed accessions is particularly important. In this study, we detected seed volatile gases using electronic nose, in order to test its possible utilization in seed viability detection and homonymous seed accessions identification. The materials used in this study were wheat, soybean, oilseed rape seeds with different viabilities and 10 soybean varieties with the same name "Mancangjin". Seeds were put in vials and their volatile gas were detected with PEN3 electronic nose. PC A, LDA, Loading and BP neural network were used to analyze the data collected by PEM3. The main results obtained are as follows: (1) Seed viability test using PEN3. The materials were wheat (4 varieties), soybean (3 varieties), oilseed rape (5 varieties). We used PCA and LDA to separate seeds with different viabilities, loading to analyze the main sensors, and BP neural network analysis method to predict seed viability. The results showed that:â‘  Electronic nose combined with appropriate analytical methods (PCA, LDA or Loading) can separate seeds with different viabilities. LDA worked better than PCA. â‘¡ Electronic nose combined with BP neural network can predict seed viability. The BP neural network training accuracy rate reached more than 95.8%, the predicted accuracy rate was higher than 98.3%, indicating that using electronic nose combined with BP neural network analysis to predict seed viability are reliable.â‘¢ Using Loading method to analysis the main sensors contribute to seed viability detection and prediction. The results showed that the first principal component accounted the dominant position in separating seeds with different viabilities. The same crop or same type of germplasm, share similar main sensors, indicating the volatile gas composition contributing to viability detection are in common.(2) Homonymous seed accessions identification using PEN3. In the aspect of identify the same name germplasm, the materials used were 10 soybean varieties with the same name "Mancangjin" selected from the National Genebank of China. Discrimination them with PCA, LDA and also agro-traits method, compare the effectiveness of two methods.2 varieties were selected randomly to verify the validity of electronic nose technology for identification of the same name germplasm. The results show that:the electronic nose combined with PCA analysis can distinguish all these 10 varieties clearly, while LDA analysis and agro-traits method can not. The accuracy of the results that PC A analysis of random samples was high, declaring that the use of electronic nose-PCA technique is reliable in the identification of homonymous soybean varieties.Taken all these results together, we believe that the combine of electronic nose and appropriate analysis methods could be a quick, non-destructive, and low-cost technique for detecting seed viability and identifing homonymous seed accessions.