You want to ask next, the teacher gave a task requires a photographic image of books, the teacher want to use the least squares fitting a straight line, but both hoff straight line detection and least squares fitting effect is bad, with OpenCV find outline four vertices and minimum circumscribed rectangle or convex hull effect is very good, want to ask everybody a great god, the rectangular, linear fitting detected with direct use of OpenCV findcounter find outline and drawcounter contours to centroid is there a difference? No words don't straight line fitting with findcounter directly, the follow-up to do robot localization grab, so here I don't know way