基于马氏距离加权图转换的点模式匹配  

Point pattern matching based on Mahalanobis distance weighted graph transformation

作　　者：张官亮[1,2] 邹焕新[1] 孙浩[1] 刘志波[1] 

机构地区：[1]国防科技大学电子科学与工程学院,湖南长沙410073 [2]武警乌鲁木齐指挥学院,新疆乌鲁木齐830049

出　　处：《中南大学学报（自然科学版）》2013年第S2期323-328,共6页Journal of Central South University:Science and Technology

摘　　要：针对大部分点模式匹配算法对噪声和出格点稳健性较差的问题,对提出的基于马氏距离的点模式匹配方法进行了理论分析,然后针对马氏距离及加权图转换匹配方法的不足,将马氏距离融入到加权图转换匹配算法中,提出基于马氏距离加权图转换的点模式匹配算法。该算法利用中值和角度距离约束不断剔除出格点,获得了更加精确的匹配结果。实验对比结果表明该方法的可行性和鲁棒性。As most point pattern matching algorithms perform poorly when the outliers and noises exist,firstly,an algorithm based on Mahalanobis distance was proposed and its performance was analyzed.Secondly,to overcome the limitation of Mahalanobis distance and weighted graph transformation,a point pattern matching algorithm based on weighted graph transformation using Mahalanobis distance was proposed.The similarity evaluated by Mahalanobis distance was embedded into WGTM algorithm under the constraint of median distance and angular distance.Then point pairs were obtained through iteratively eliminating the outliers.The results demonstrate that the proposed algorithm is effective and robust.

关 键 词：点模式匹配 马氏距离 加权图转换 K近邻 角度距离 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]