Non-destructive Detection of Apple Maturity by Constructing Spectral Index based on Reflectance Spectrum

Abstract. Maturity is one of the important indicators closely related to the quality of apples, especially the harvest maturity affects the yield, quality, storage and transportation of apples. In order to realize the application of visible near infrared spectroscopy in apple maturity detection, apples were divided into three types of maturity (M1, M2, M3) by iodine test. Then, by analyzing the spectral morphological differences of different maturity samples, four spectral indexes are established; analysis of variance and correlation analysis show that there are significant differences between spectral indexes of different groups of samples, and there is a significant correlation between spectral index values and maturity; finally, a linear discriminant analysis model is established by using spectral index, and the classification accuracy of prediction set samples is 87.15% 84.09% 80.68% 85.23% respectively. In addition, the conventional chemometrics analysis method is compared with the spectral index method, 11 characteristic wavelengths are extracted by using the continuous projection algorithm (SPA), and the classification accuracy of the prediction set samples of the established RAW-LS-SVM model and SPA-LS-SVM model is 87.36%, 90.11%, respectively. In general, the above results show that both spectral index method and conventional chemometrics methods can realize the non-destructive testing of apple maturity. Although the classification accuracy of LDA model based on spectral index is lower than that of LSSVM model based on full wavelength and characteristic wavelength, it has broad application potential because the construction of spectral index is simpler and easier to realize portable and low-cost apple maturity detection instruments.