Fig. 1. Spectral data, data processing, and index analysis of four rice varieties. A, Original spectral curves of grinding sample of four rice varieties. B, Classification accuracy of different preprocessing methods under different classification models. C, Four kinds of rice grinding sample spectral (PE) values obtained based on the ASW-PE algorithm. D, Four kinds of rice grinding sample spectral PE ratios obtained based on the ASW-PE algorithm. E, Rice variety classification accuracy based on full spectral data and six characteristic wavelength selection algorithms. F, Classification ability of rice variety under different classification modeling methods. WC, XS, YS, and YG represent rice varieties Wuchang, Xiangshui, Yinshui, and Yueguang, respectively. SNV, Standard normal variate; PLS, Partial least squares; PSO-SVM, Particle swarm optimization support vector machine; RF, Random forest; R2, Coefficient of determination; RMSE, Root mean square error; PE, Permutation entropy; PCA, Principal component analysis; ASW-PE, Adaptive sliding window permutation entropy; SW-PE, Sliding window permutation entropy; ANOVA, Analysis of variance; CARS, Competitive adaptive reweighted sampling; SPA, Successive projection algorithm.