基于ORB和改进RANSAC算法的图像拼接技术  

Image stitching technology based on ORB and improved RANSAC algorithm

作　　者：佘建国[1] 徐仁桐 陈宁[2] 

机构地区：[1]江苏科技大学机械工程学院,江苏镇江212003 [2]江苏科技大学能源与动力工程学院,江苏镇江212003

出　　处：《江苏科技大学学报（自然科学版）》2015年第2期164-169,共6页Journal of Jiangsu University of Science and Technology:Natural Science Edition

基　　金：江苏省高校科研成果产业化推进基金资助项目(JHB2011-41)

摘　　要：图像拼接技术能够很好地解决单张拍摄照片的视野狭窄问题,是数字图像处理的一个重要分支.其一般步骤是先粗匹配,后提纯,然而在用随机抽样一致性算法(random sample consensus,RANSAC)提纯时,将粗匹配的所有特征点对都进行迭代运算,运算量较大,导致拼接速度慢.针对该问题,文中采用了ORB(oriented FAST and rotated BRIEF)算法,并结合空间一致性检测理论来改进RANSAC算法,以提高拼接速度.文中采用ORB算法提取特征点,并进行粗匹配;在RANSAC提纯之前,先进行一次空间一致性检测,从而缩小RANSAC抽样总量,减少迭代次数.分别用ORB+改进RANSAC算法和ORB+RANSAC算法对两组图片进行对比实验,实验表明,ORB和改进RANSAC算法的结合在保证匹配精度的基础上提高了匹配速度.Image stitching technology can better solve confined vision of a single photograph , which is an impor-tant branch in the field of digital image processing .The general image stitching steps are coarse matching and then purification , however , RANSAC （ random sample consensus ） algorithm has large amount of computation , lower efficiency problem when going on point-purification .This paper uses ORB （ oriented FAST and rotated BRIEF） algorithm and spatial consistency check theory for improving RANSAC algorithm in order to improve stitching speed .This paper uses ORB algorithm to extract feature points and carry out coarse matching .Before RANSAC purifying , this paper uses spatial consistency check theory to reduce the amount of RANSAC sampling and decrease the number of iterations .With two pairs of images ,this paper adopts ORB ＋improved RANSAC and ORB＋RANSAC algorithm to have a purification experience .Experiments show that ORB combined with im-proved RANSAC algorithm can increase the matching speed and guarantee matching accuracy .

关 键 词：ORB 空间一致性检测 改进RANSAC 单应矩阵 提纯 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]