SIFT 由特征提取，特征描述符描述和特征匹配 3 部分构成，该算子特征提取数目庞大，建立特征描述符运算

量高，导致算法效率低。提出了一种 SEC( SIFT-Edge-Corner) 算法，在图像尺度空间提取角点代替 SIFT 特征点，并根

据角点是边缘曲率极值理论，预先采用 Canny 算子得到高斯边缘图像金字塔，再提取角点并进行尺度选择。实验结

果表明: 该算法在保障高准确率的前提下大幅度提高特征提取效率-By the SIFT feature extraction, feature descr iptions and feature matching descr iptors 3 parts, the large number of feature extraction operator established feature descr iptor computation high, resulting in low efficiency of the algorithm. Presents a SEC (SIFT-Edge-Corner) algorithm, the image scale space instead of SIFT feature extraction corner points and corner points based on extreme value theory is an edge curvature in advance using Canny operator edge image obtained Gaussian pyramid, and then extract corner point and scale selection. Experimental results show that: the algorithm protect high accuracy under the premise of feature extraction efficiency greatly improved