A machine vision-based FPC defect detection method is designed to address the problems of low efficiency of FPC defect detection. The Otsu algorithm is applied for image segmentation to decompose the image into two parts: foreground and background; the SURF algorithm is applied to achieve image alignment, and then an experimental platform is built for detecting FPC defects. For the foreign colors, foreign objects, and pressure/scratch/scratch defects in the appearance of FPC boards, detection algorithms are designed to achieve automatic recognition and classification of FPC board defects according to their defect characteristics. Finally, based on the above, by calculating the detection of 10 groups of FPC boards containing defects, the average over-inspection rate is 5.66% and the average leakage rate is 8.74%, which are better than the values of manual visual inspection in enterprises.