一种增量符号相关的景象匹配方法  

Scene matching approach by increment sign correlation

作　　者：刘小军[1] 杨杰[1] 凌建国[1] 沈红斌[1] 

机构地区：[1]上海交通大学图像处理与模式识别研究所,上海200240

出　　处：《红外与激光工程》2006年第6期732-737,共6页Infrared and Laser Engineering

基　　金：航空科学基金资助项目(04D57004)

摘　　要：针对图像灰度变化和遮挡问题,提出一种增量符号相关的景象匹配方法。介绍了增量符号相关算法的理论基础,基本原理为将二维图像变换为一维向量,再将一维向量按照某种编码规律进行编码,得到由0、1组成的特征字符串,对由实时图和基准图上同样大小的基准子图变换而来的特征字符串求相关,相关系数最大者即为最佳匹配区。将此算法与绝对差算法和归一化积相关算法进行实验对比,在对算法进行定量分析的基础上,证明增量符号算法对灰度变化和遮挡有较好的鲁棒性,为图像匹配导航与制导提供了新的思路。A scene matching approach based on increment sign correlation is proposed to solve the problem of gray change or occlusion in image. The increment sign correlation approach is theoretically introduced, which transforms the 2D image to 1D vector and then codes the 1D vector to special character strings made up of 0 and 1. By calculating correlation coefficient between two special character strings, which are transformed from the same size sub-image in the reference image and the sensed image, the sub-image corresponding to the maximum correlation coefficient is determined to be the best matching area. Based on quantitative analysis and comparison with the sum of absolute difference algorithm and the normalized cross-correlation algorithm through fundamental experiments, this approach is proved to be robust for gray change or occlusion. It is a novel idear for scene matching navigation.

关 键 词：景象匹配 增量符号相关 灰度变化 遮挡 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]