Using the computer image processing technology to analyze three geometric parameters of the flat peach (i.e. the diameter of flat peach in the suture midline, the diameter perpendicular to the suture midline and the thickness). The prediction model of flat peach-mass was established based on external characteristic information. And through the analysis of multiple linear fitting, found the prediction model of the measured values and flat peach-mass. By comparing goodness of fit of different models based on different parameters, the optimal quality prediction model was established. The MATLAB R2014a software acquired the pixel value of the corresponding geometric parameter. The pixel-mass prediction model was founded by fitting with it. The results suggest that predicting flat peach-mass based on external characteristic information is practicable. The study provides the basis for the use of machine vision method to classify flat peach-mass.