To study on internal quality of the fruit, in order to improve the detection speed and robustness, near infrared spectral diffuse transmission is applied to on-line detection research, using partial least squares regression (PLSR) algorithm combined with different pre-processing methods to establish the quantitative models of soluble solids content (SSC) of internal quality of the apple. The results of final model were achieved as follow: the model was stability which was used by first derivative combined with multiplicative signal correction (MSC), rootmean square errors of calibration and rootmean square errors of prediction are 0.17 and 0.39, respectively, the correlation coefficient of calibration is 0.988 3. The results indicated that the diffuse transmission of NIRS could be applied successful determine the SSC in apple as a fast and nondestructive method.