Vision-based cloth sewing track calibration method

The invention provides a cloth sewing track calibration method based on vision. The method comprises the following steps: intercepting an ROI (Region of Interest) containing cloth; using an image processing algorithm to search inflection points of an image contour in the ROI area; the method comprises the following steps: acquiring a data set of to-be-processed cloth, acquiring some pictures of each type of cloth for training a deep learning model, marking inflection points of the edges of the cloth after the cloth is acquired, training a deep learning key point detection model after the marking is completed, and finding out inflection points of image contours; a curvature search method and a deep learning method are adopted to jointly search inflection points, after the inflection points are searched respectively, the number and the distance of the inflection points searched by the two methods are obtained, and whether the piece of cloth needs to be abandoned or not is judged according to the number and the