The invention discloses a tire fetal-membrane surface character defect detection method based on machine vision, which comprises the following steps: scanning tire fetal membranes to be detected and acquiring a set of images, respectively processing the acquired image and then obtaining the external arc-shaped contour of the tire; fitting circle center and radius of the external arc-shaped contour of the tire, conversing the external arc-shaped image to be detected of the tire to a straight type image to be detected through polar coordinates transformation, performing threshold segmentation, positioning a ROI image to be detected; obtaining the straight type image corresponding to a CAD design image of the tire fetal membrane to be detected, intercepting the straight type image to obtain an image block matched with each ROI image to be detected; performing character identification on each ROI image to be detected and the matched image block, and performing defect determination according to the character identification result. The method can automatically detect the character defect of the tire fetal membrane, has the advantages of high detection stability, low detection cost and fast detection speed, and can be widely used in the tire fetal membrane detection field.