Targeting on the demand of the market for apple quality detection, a handheld device for apple sugar content detection combined for mobile phone based on visible/near infrared spectroscopy technology was designed to determine the wavelength range and spectral sensor suitable for apple sugar content detection by optimizing the characteristic wavelength. The combination with the mobile phone completed the highefficiency, nondestructive and lowcost detection of apple sugar content. An STS-NIR miniature fiber optic spectrometer (wavelength range 650~1100nm) was selected to collect the spectra of 120 apples by using the spectrum acquisition platform built by the laboratory itself, and the true sugar content of the measured apples was obtained through the sugar refractometer. The partial least square (PLS) algorithm was used to model the fullwavelength data, and variable selection methods such as successive projection algorithm (SPA), genetic algorithm (GA) and competitive adaptive reweighted sampling method (CARS) were used to identify and simplify the characteristic bands of the fullwavelength to select the effective wavelength. Variables of the measured wavelength, and the effective wavelengths were selected according to the regression coefficient. The results of variable selection showed that the three sets of characteristic variables obtained had overlapping terms, and all of them contained wavelength variables related to the apple sugar content. The partial least squares (PLS) algorithm was used to establish a prediction model of apple sugar content based on three sets of characteristic bands variables, and the three sets of results were analyzed, including the comparison of prediction correlation coefficient (Rp), prediction root mean square error (RMSEP) to evaluate the accuracy of the built model. The experimental results showed that the modeling results obtained by using the three groups of characteristic were all good, and the predictive correlation coefficient was above 0.93, among which GA-PLS model had the best predictive effect on apple sacchariness, with the predictive correlation coefficient up to 0.9447. According to the highly overlapping coincidence term of the characteristic variables bands obtained above, the characteristic wavelength bands and their corresponding optical sensor for detecting apple sugar content saccharification were determined, and 40 other apples were tested and verified based on the designed handheld device for testing the sugar content of apples. The correlation coefficient was predicted to be 0.8822 based on the designed a handheld device for apple sugar content detection combined for mobile phone. The results showed that the device designed was of low cost, easy to carry and had high detection accuracy and efficiency, and it was feasible to realize the realtime nondestructive testing of apples sugar content. 