Cylindrical surface image matching method combining with SURF feature extraction and curve fitting

The invention relates to a cylindrical surface image matching method combining with SURF feature extraction and curve fitting, comprising steps of arranging two images (A and B) on an upper position and a lower position in a left-right alignment manner, adopting a SURF characteristic detection algorithm to perform characteristic detection on the two images, finding out a matched pair set, calculating the angle between the straight line defined by each matched pair in the characteristic point matched pair set and the horizontal direction and the angle and an Euclidean distance between the two characteristic points in each matched pair, establishing an angle set K of the image matched pair and an Euclidean distance set D of the image, performing curve fitting on the angle set K of the matched pair, wherein the independent variable of the curve fitting is the abscissa X1i of the matched pair, the ordinate of the curve fitting is angle Theta I, removing the mismatching, performing curve fitting on the Euclidean distance set D of the matched pair, wherein the independent variable of the curve fitting is the abscissa X1i of the characteristic point in the image A, and removing the mismatching. The invention can more accurately remove the mismatching of the cylindrical object.