基于轮廓特征多层描述和评价的部分遮挡目标匹配  

Partially occluded object matching via multi-level description and evaluation of contour features

作　　者：史思琦[1] 石光明[1] 李甫[1] 

机构地区：[1]西安电子科技大学智能感知与图像理解教育部重点实验室,陕西西安710071

出　　处：《光学精密工程》2012年第12期2804-2811,共8页Optics and Precision Engineering

基　　金：国家自然科学基金资助项目(No.61100155;61070138;61033004)

摘　　要：针对传统目标匹配算法难以实现部分遮挡目标精确匹配的问题,本文基于轮廓特征的描述和评价提出了一种有效的部分遮挡目标匹配算法。首先,利用曲率划分目标轮廓得到描述局部特征的轮廓分段,并根据目标的骨架对轮廓分段进行合并和分类,实现了目标特征的多层次描述。然后,提出了评价轮廓分段的两个参数:重要性和局部性。前者用于评价轮廓分段所描述目标特征的重要性,后者用于评价轮廓分段相对目标整体轮廓的比例。最后,将两个评价参数与轮廓分段之间的相似度联合起来,得到衡量目标相似程度的加权部分相似度,从而获得部分遮挡目标的最佳匹配结果。与现有遮挡目标匹配算法相比,在不同遮挡情况下本文算法的平均识别率提高了1.5%左右。As traditional object matching algorithms can not match precisely a partially occuluded object, this paper proposes a novel partially occluded object matching algorithm based on the description and evaluation of contour features. Firstly, the contour fragments to describe the local object feature are obtained by splitting the object contour with contour curvature, and those contour fragments are merged and classified according to the object skeleton to describe the object features at multi-levels. Then, two evaluation parameters, importance and partiality, are defined for those contour fragments. The former evaluates the importance of the local feature, and the latter evaluates the proportion of contour fragment to the whole contour. Finally, the two evaluation parameters of contour fragment and the similarity between contour fragments are derived to obtain the weighted partial similarity to measure the matching degree of the partially occluded object reasonably and to obtain the optimal matching result. Compared with some current matching algorithms, the proposed algorithm improves the average recognition rate about 1.5 ~ under various occluded cases.

关 键 词：部分遮挡目标 轮廓分段 多层特征描述 特征匹配 加权部分相似度 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]