Image Matching Research Based on Improved SIFT Algorithm

Abstract

Image matching technology is the key technology for indoor vision robot positioning and map construction. The Scale Invariant Feature Transform(SIFT) algorithm has scale and rotation invariance, but there is a problem of long registration time and poor real-time performance. Aiming at this problem, a dual-threshold Featurnes From Accelerated Segment Test(FAST) algorithm is proposed to transform the feature extraction method of SIFT algorithm. At the same time, the 128-dimensional feature point descriptor of the traditional SIFT algorithm is reduced to 32-dimensional, so that local features can be extracted quickly. The experimental results show that the improved algorithm is twice as fast in registration time, which is suitable for some occasions with high real-time requirements.