Identification of hybrid rice varieties of the same series using near-infrared spectrum

Selection of high purity rice seed is a crucial part of guaranteeing high rice yields, and carrying out rigorous rice seed identification is of great significance in enhancing rice yields and maintaining food security. Identification of hybrid rice varieties of the same series using near infrared spectrum was carried out. Five hybrid rice varieties, namely, Bai Xiang You 005, Han Liang You 786, Lv Liang You 9873, Zhen Xiang You, and Zhi Liang You 6, were selected for the study. Near-infrared spectral preprocessing method, radial basis function neural network classification model and its parameter optimization algorithm were investigated, and finally a rice seed identification model was constructed. Furthermore, Kernel Principal Component Analysis was used for data dimensionality reduction, and the accuracy of the identification model using the downscaled data was 96.77%.