摘要： 为解决非重叠视域监控系统中的目标匹配问题,提出一种基于行人主颜色统计信息和空间分布信息的特征提取模型。使用最近邻聚类法对HSV颜色空间的目标进行聚类,得到行人主颜色分布特征,在此基础上,对行人各个部分进行加权,并通过设置阈值确定目标是否匹配。实验结果表明,该方法能实现合理的相似度值并排除匹配时不必要颜色的干扰,减少拥有相似颜色信息的不同行人的错误匹配,同时对行人形态和姿势的变化具有较好的鲁棒性。

关键词: 非重叠, 特征, 聚类, 目标匹配, 相似度

Abstract: In order to solve the problem of object matching based on non-overlapping camera views,a feature extraction model based on pedestrian’s main color statistical information and spatial distribution information is put forward.The main color feature is obtained by clustering the target through the nearest clustering method in HSV color space.Each part of the pedestrian is weighted based on the pedestrian’s main color feature.And it determines whether the two targets’ are matched by setting the threshold.Experimental results show that the method can achieve reasonable similarity value and eliminate undesired color matching.It is capable of reducing false matching of different pedestrians with similar information and has good robustness to the change of shape and gesture.

Key words: non-overlapping, feature, clustering, object matching, similarity

中图分类号: 

• TP391