The paper aims to improve the detection accuracy of overprint error in flexographic printing. In this paper, the main causes of overprint error in flexographic printing were analyzed, and a method of overprint error detection based on machine vision was proposed. Firstly, the industrial camera was calibrated to obtain the original "cross-shaped" circle marking image, and the image was preprocessed by fusion filtering algorithm. A second segmentation algorithm was proposed for image color separation when the preprocessed image was analyzed by data analysis. Then template matching algorithm was used for template matching to obtain the deviation data for overprint deviation analysis. In the experiment, the axial and vertical detection errors of the five sets of alignment markers were all controlled within 0.03 mm, which was far in line with the detection standard of the industry within 0.2 mm. The visual-based overprint error detection method proposed in this paper has the characteristics of high recognition accuracy, fast processing speed, reliable and stable operation, and has a certain feasibility.