A machine vision system for detecting bruise on apple surface was developed. The system was consisted of single-channel transfer unit, image acquisition unit and classification unit. A CCD camera with a 635nm filter and two plane mirrors were used to acquire the image of overall surface of apple sample. The system was designed to rotate apple at three different positions in the conveyor running, acquiring nine images of same apple. Digital image processing method was used to analyze sample defects. Ratio between black pixels and all pixels was calculated and compared with a fit value to judge the bruise area. The detection accuracy of the system for sample was 92.5%.