摘要
为了提高概率图模型点模式匹配的精度,本文提出了改进的动态图模型点模式匹配算法。首先,在动态图模型点模式匹配的相似性度量中应用混合高斯分布,以提高模型利用多特征的能力,使匹配方法对噪声更加稳健。其次,在目标点集中引入了虚拟的哑点并给出了包含哑点的相似性度量。当模板中的点和哑点相匹配的相似性度量更大时模板点将和哑点匹配,以减少由异常点所导致的误配。实验结果表明所提出的匹配方法对噪声和异常点更加稳健,匹配的精度也优于传统方法。
Aiming to improve the point pattern matching accuracy with graphical models, an improved point pattern matching algorithm is proposed using dynamic generating graphical model. First, mixed Gaussian distribution is applied in similarity measure of dynamic generating graphical model to improve the multi-feature ability of model, which make the matching results more robust to noise. Second, a dummy point is introduced in the target point set and the similarity measure including the dummy point is provided. A point in the template would match the du.mmy point when the similarity measure of a template point with a target point is less than that of the template point and the dummy point, which can reduce the mismatching rate caused by outliers. Experimental results with simulated and real images show that the proposed algorithm is more robust to noise and outlier, and compared with the traditional methods, matching accuracy is improved.
出处
《光电工程》
CAS
CSCD
北大核心
2013年第1期132-138,共7页
Opto-Electronic Engineering
基金
国家自然科学基金资助项目(10926197
11126312
61201323)
陕西省教育厅自然科学基金(12JK0744)
关键词
图模型
点模式匹配
异常点
graphical model
point pattern matching
outlier