Harvest maturity is a key factor affecting apple storage performance. Thus, in order to achieve batch harvesting and suitable picking maturity of apple, rapid and non-destructive discrimination model was studied based on visible/near infrared spectral technology. Apple samples were classified into three maturity levels (mid-ripe, ripe and over-ripe) according to different fruit development times after flowering. Spectral information of all samples was acquired by using the visible/near-infrared spectral system with a wavelength range from 200nm to 1100nm in the laboratory, and spectral differences of samples at different mature stages were analyzed. It was found that the spectral reflectance of the mid-ripe samples was significantly higher than that of the ripe and over-ripe samples. However, the spectral bands behind the ripe and over-ripe samples had similar spectral characteristics, resulting in overlapping. Then, after the preprocessing of SG and multivariate scatter correction method, a method using callback arithmetic operator named twice-detect was used to detect outlier sample for different quality parameters. Among them, five outlier samplers of the soluble solids content, two outlier samples of firmness and no outlier sampler in color factor were removed. Finally, the remaining 235 samples were participated in the modeling analysis. The characteristic variables of soluble solids, firmness and the parameters of L*, C*, h*, a andb were extracted by the random algorithm which can be modeled with a small number of variables. A maturity index coupling internal quality named simplified internal quality index (SIQI) and another maturity index coupling with factor analysis named factor quality index (FQI) were applied to evaluate the maturity of apple. Using the characteristic variable as input, the partial least squares prediction model of SIQI index and FQI index was established. The prediction correlation coefficient of SIQI index was 0.938, and the root mean square error was 0.216. Similarly, the prediction correlation coefficient of FQI index was 0.917 and the root mean square error was 1.152. Therefore, it was feasible to use spectral information to predict the maturity evaluation index of coupled multiple indexes. At the same time, the classification model using spectral information was directly established by the extreme learning machine (ELM) algorithm and the support vector regression (SVR) algorithm. By comparing the results of the four classification models, it was found that the classification results based on SIQI index combined with SVR algorithm were the best, which was better than the direct classification model. The classification accuracy was 85.71%. The results stated that the classification of apple harvest maturity can be achieved by using visible/near-infrared spectroscopy information and the maturity evaluation index coupling with related internal quality indicators. The research result can provide a theoretical reference for the development of nondestructive testing equipment for subsequent apple harvesting maturity.