基于不变矩景象匹配算法  

Scene Matching Algorithm Based on Moments Invariant

作　　者：周振环[1] 赵明[1] 

机构地区：[1]深圳职业技术学院软件工程系,深圳518055

出　　处：《计算机工程》2007年第22期203-205,222,共4页Computer Engineering

基　　金：深圳市科技计划基金资助项目(05KJcd004);2005年粤港关键领域重点突破基金资助项目(2005A11304003)

摘　　要：分析基于灰度匹配方法存在着亮度和裁剪误差,提出基于不变矩景象匹配算法,并对矩不变量作如下改进:用原点矩代替中心矩,克服平移不变性给不变矩带来的影响;用圆形测量窗代替方形测量窗消除因旋转而产生的裁剪误差;引入亮度因子,克服基准图和匹配图之间因亮度差异带来的影响;采用金字塔算法减少搜索次数,提高匹配速度。实验表明,基于不变矩下视影像匹配算法具有较好稳定性,匹配准确率较高。Since there are errors of brightness and clipping in gray-based paring algorithm, a scene matching algorithm based on moments invariant is developed and improved, in which, error of translation invariance is reduced by replacing raw moment with central moment, clipping error of rotation is eliminated by replacing rectangle measurement image with round one, affect of brightness difference between reference map and real time image is avoided by adding brightness factor and pyramidal search algorithm is adopted to reduce search times and speed up matching rate. Experiments show that matching algorithm based on moments invariant is very stable with high accurate rate.

关 键 词：景象匹配 矩不变量 亮度因子 金字塔算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]