Journal:Geomatics & Spatial Information Technology

Key Words:SIFT feature points, GPU, parallel computing, fast matching

Abstract:lmage matching is the core problem of computer vision and digital photogrammetry. Because of strong ability of matchingand good rbustness, SlF'T matching algorithm based on feature has beeome the research hot spot. But the characteristics of largquantity and the high dimension of the extracted feature points direet elfeet the matching speed of SlfT operator. So In this paper, using the poweriul parallel procesing capacity of GPU, the design and implement of Slf'T feature points matching on the GPU was introdued in detail. And the eontrast experiment further shows the advantage of the matching of Slf'T feature points brought by the GPU

Co-author:TIAN Yihan,DUAN Yansong

First Author:KE Tao

Volume:37

Issue:12

Page Number:14-16

Translation or Not:no

Date of Publication:2014-12-01