计算机应用与软件Issue(5)：244-246,263,4.DOI:10.3969/j.issn.1000-386x.2014.05.062
基于SIFT自然特征的AR系统研究和实现
STUDY AND IMPLEMENTATION OF SIFT NATURAL FEATURES-BASED AUGMENTED REALITY SYSTEM
肖慧 1陆奎1
作者信息
- 1. 安徽理工大学计算机科学与工程学院 安徽 淮南232001
- 折叠
摘要
Abstract
In view of the low efficiency of tracking and registration in current augmented reality system,we put forward an improved SIFT operator-based natural features matching method,it makes pros and cons two-way matching after using DOG pyramid to detect and extract the feature points on key frame image and benchmark image,which well enhances matching efficiency and stability of key features.Meanwhile,it combines L-K light flow tracking algorithm to optimise the processing means when failure in tracking the key frames,so as to reduce the search area needed for key points detection during relocation.Experiments show that,since the improved SIFT matching feature points are lesser and better,the positioning accuracy is higher as well,the overall time cost of tracking and matching decreases nearly 50%,therefore this AR system has good effect of real-time virtual and actual fusion in special environments such as insufficient light,occlusion and jittering.关键词
尺度不变特征/特征匹配/增强现实Key words
Scale invariant feature transform (SIFT)/Feature matching/Augmented reality分类
信息技术与安全科学引用本文复制引用
肖慧,陆奎..基于SIFT自然特征的AR系统研究和实现[J].计算机应用与软件,2014,(5):244-246,263,4.
