Comparative analysis of rice seed viability detection based on different spectral bands

Rice seed vitality is one of the very important determining factors of high yield. Therefore, it is vital to test and determine seed vigor before sowing or sale for farmers as well as seed companies. It is particularly important to develop a method to test the viability of rice seeds quickly and non-destructively so far. Spectroscopy is a science and technology developed from experiments. In this study, two spectral bands of 400-1100nm (spectral range I) and 980-1700nm (spectral range II) which contains visible and near-infrared spectroscopy combined with multiple data preprocessing methods and classification model were compared to identify the seed vigor. 100 normal rice seeds and 100 heat-damaged seeds were detected. The standard germination test was performed on the detected seeds to determine the vigor of seeds. The spectrograms were analyzed, and the two groups of experimental data were pre-processed separately, and the partial least squares discriminant analysis was used to discriminate the seed vigor. The data results showed that the accuracy of the seed vitality discrimination model based on spectral range II without pretreatment was higher than that of spectral range I. The accuracy of the first derivative classification were the highest in the two bands, and the accuracy of the two sets data were approximate. The experimental results showed that the spectral technique of the two bands had the potential to identify vigor of rice seeds, rapid and non-destructive detecting of rice seed vigor can be performed combined with a seed single granulation device.