Abstract: The article maintains that the current popular corner set detection method has three drawbacks:First, in the digital image space, curvature exists calculation errors, thus it is inappropriate to locate corner set with curvature; two, the corner edges of an image are generated not only by a single point, but also by point set. Therefore not depending on corner point set is inappropriate; three, large curvature points is not equivalent to corner points. In view of this, the article presents a sudden change degree based on the corner set detection method. The idea is that in the binary image, the image edge can be seen as a one-dimensional manifold. The sudden change point set divides the one-dimensional manifold into smooth manifold segments with various sizes. If the smooth manifold undergoes drastic changes in direction after going through the sudden change point set, the sudden change point set is a corner set. Through experiments, the author proposed that the algorithm results in the article are superior to the currently popular algorithm test results.