Aiming at the problems of labor intensity, low automation level, and low efficiency in manual inspection of production date defects, the machine vision-based dairy production date coding defect detection system was developed. The system uses a CCD camera, computer, light source, etc. to build a hardware platform, and image processing techniques, such as median filtering, binarization, image segmentation, and template matching were used to achieve the goal of defective detection, including production date missing and fuzzy codes. The experimental results showed that the system could accurately identify products with defective production date, and achieve the expected results, showing practical significance.