In order to improve the accuracy of online detecting soluble solids content (SSC) of apples by the method of near infrared spectroscopy, the combination of moving window partial least squares (MWPLS) and genetic algorithm (GA), successive projections algorithm (SPA) was used to select the characteristic variables, and then the partial least squares regression model was developed. The MW-GA model with the 36 selected characteristic variables obtained the best result with correlation coefficient of prediction (Rp) of 0.90 and root mean square error of prediction (RMSEP) of 0.70°Brix, which indicated that the combination of MWPLS and GA could select the characteristic variables of near infrared spectroscopy effectively.