Abstract: An improved SIFT image feature detection and matching algorithm is proposed.Firstly,this paper is based on the depth map containing the pixel information.Through the corresponding mapping relation,the depth map is transformed into a two-dimensional image.And then the gray value of the two-dimensional image is determined according to the local differential property of each vertex of the depth map to get the two-dimensional gray feature image.Secondly,the SIFT algorithm is used to detect the feature points of the feature images.Thirdly,the K-nearest neighbor algorithm and the bidirectional feature matching algorithm are combined to make the matching result more accurate and fewer mismatching pairs,then restore the matching results to the depth map.Finally,the random sampling consistency RANSAC algorithm is used to eliminate the mismatching points,and the registration of two images is realized.The experimental results demonstrate the robustness and effectiveness of the improved algorithm.