Aimed at the problem of judging the surface defects of carbon fiber reinforced plastic (CFRP) after machined, a method of judging the proportion of surface defects in grinding of CFRP based on microscopic vision was proposed in this paper. The method uses image processing technology to analyze the micro-imaging of grinding surface of CFRP, which is based on region growing algorithm, extracts and calculates the delamination defects and fiber intermittent damage defects formed on grinding surface of CFRP, and quantitatively determines these two main types of defect. The results show that this method can quickly and preliminarily predict the surface advantages and disadvantages of CFRP workpiece, and provide a strong reference for the next improvement of processing technology.