基于相似度加权的自适应HD算法

Hausdorff距离(Hausdorff distance, HD)是一种点集与点集之间的距离测度, 常用于目标物体的匹配、跟踪和识别等. 本文在分析经典HD及改进算法的基础上, 提出了一种基于相似度加权的自适应HD (Adaptive Hausdarff distance, AHD)算法. AHD算法利用不同点到点集的最小距离的个数作为匹配相似度的测量, 并舍弃对判断匹配几乎没有作用的较大的点到点集的最小距离值; 同时根据点到点集的最小距离自适应选择权值, 从而得到一种基于相似度测量加权系数; 通过利用部分点到点集的最小距离和基于相似度的加权平均, 既增强了算法的鲁棒性, 又尽可能地保证了算法的精度. 实验结果显示, AHD算法在匹配准确性、抵抗噪声和遮挡干扰等方面性能良好.

Adaptive Hausdorff Distance Based on Similarity Weighting

Hausdorff distance (HD) is a popular measure between two sets of points, and has been widely used in object matching, tracking and recognition. Based on a thorough analysis of the traditional HD and its improved variants, a new adaptive Hausdorff distance based on similarity weighting (Adaptive Hausdorff distance, AHD) is proposed. The AHD uses the number of samples which have the minimum distance to a given point in the other set as the similarity measure, and rejects those relatively large minimum distances due to their marginal influence on matching evaluation. In addition, the weighting factor is adaptively adjusted according to its minimum distance of a point to a set. Furthermore, the robustness and accuracy are well balanced by using a subset of minimum distances and weighted averaged similarity measure. Experiments show that our proposed AHD has good performance in terms of matching accuracy, and is robust to random noise and occlusion.