Qualitative identification of rice actual storage period using olfactory visualization technique combined with chemometrics analysis

Abstract

The actual storage period is one of the important indicators reflecting the storage quality of rice. In this study, an olfactory visualization system was assembled by the colorimetric sensor array made of optimized chemical dyes to identify the actual storage period of rice. First, 15 chemical dyes were selected to prepare a colorimetric gas sensor array with strong specificity to assemble the olfactory visualization system. Then, principal component analysis (PCA) was used to perform feature compression and visual presentation of the sample spatial distribution of the pre-processed sensor color components. Finally, three different linear and non-linear pattern recognition methods, i.e., k-nearest neighbor (KNN), probabilistic neural networks (PNN) and support vector machine (SVM) were compared to build recognition model to realize the qualitative recognition of rice actual storage period with high precision. The experimental results showed that the PNN nonlinear method was the most suitable for the establishment of the actual storage period qualitative model of rice in this study by comparing the recognition results of different optimal recognition models. The correct recognition rates in the training set and prediction set were 98.89% and 94.67%, respectively, showing good stability and generalization performance. The overall results sufficiently demonstrate that the colorimetric sensor array of the optimized olfactory visualization system and the chemometrics analysis can achieve qualitative identification of the actual storage period of rice with high precision.