采用圆投影和序贯相似检测的图像匹配技术  

Fast image matching algorithm based on circular projection and sequential similarity detection

作　　者：贾晓芬[1] 赵佰亭[1] 周孟然[1] 杨芝权 

机构地区：[1]安徽理工大学电气与信息工程学院,安徽淮南232001

出　　处：《哈尔滨商业大学学报（自然科学版）》2015年第2期232-236,241,共6页Journal of Harbin University of Commerce：Natural Sciences Edition

基　　金：国家自然科学基金(51174258);安徽高校省级自然科学研究项目(KJ2013B087);淮南市科技计划项目(2013A4017;2011B31);安徽理工大学青年教师科研基金(2012QNZ06);国家创新创业项目(201210361066)

摘　　要：针对现有的高精度匹配方法复杂度较大,现有的圆投影匹配方法在实际应用中计算量高、速度慢的缺点,提出了一种新的图像匹配算法.该算法基于序贯相似检测原理,采用由粗到精的匹配策略,应用步长自适应的跳跃式搜索方法,在不影响匹配精度的前提下,跳跃了大量匹配点,减小了在非匹配点处的计算量.实验结果表明,与已有的基于不变矩的匹配算法、圆投影匹配算法相比,该算法具有较高的匹配精度,且匹配速度快.Fast image matching algorithm is the important research content of guidance, re-connaissance and other applications.In the process of template matching, the existing circu-lar projection matching algorithm of arbitrary angle rotation adopts full search, and poses sig-nificant computational overhead.In order to reduce the computation, coarse-to-fine matc-hing and adaptive skipping were employed by pruning the search space.In this paper, the existing image matching algorithms based on circular projection were analyzed, and a new method based on distances between inner-subtemplate’ s circular projection vectors and se-quential similarity detecting was proposed.Experimental results showed that the proposed method not only has won the advantages of already existing circular projection matching algo-rithm, and had a faster matching speed and the same precision as existing circular projection matching algorithm.

关 键 词：模板匹配 圆投影匹配 旋转不变性 自适应像素跳跃 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]