Due to the problems of SIFT algorithm in image recognition

the recognition accuracy is not high and the matching speed is slow. In this paper

a fast method for obtaining and matching feature points is proposed. Firstly

based on the feature point acquisition method of image edge

the image edge is extended

and the feature points are judged only in the extended region. Then

the normalized gray difference values of 16 subregions around the feature points are added to the feature point descriptor to enhance the feature point description ability. The dimension of descriptors is reduced by KPCA method

and the cosine similarity is used to preprocess noise feature points to speed up the matching speed. The experimental results show that compared with SIFT algorithm

the image recognition rate is improved by about 8%

and the recognition speed is improved by about 50%.