The positioning process of the feature points of the SURF algorithm is consistent with the SIFT algorithm, where each pixel processed by the Hessian matrix (i.e., the Hessian matrix discrimination value of each pixel) is obtained, and the image domain (same size image) and all neighbors of the scale domain Points (adjacent scale space) are compared. If it is larger or smaller than all adjacent points, this is the extreme point. As you can see in the picture below, the middle detection point should be done with 8 pixels near 3 × 3 of the image, 18 pixels near the adjacent 3 × 3 of the upper and lower layers, and a total of 26 pixels. compare. When a feature point is found for the first time, the main point, which is weak in energy and misplaced, is filtered through filtering, so you can finally select the desired feature point that is stable.