运动目标跟踪中“选择性”颜色直方图的研究  

Research on the "Selective Color Histogram" for Moving Object Tracking

作　　者：张谢华[1,2] 张申[1] 田敏[3] 

机构地区：[1]中国矿业大学信电学院,江苏徐州221008 [2]徐州师范大学现代教育技术中心,江苏徐州221116 [3]同济大学电子与信息工程学院宽带无线通信与多媒体实验室,上海201804

出　　处：《小型微型计算机系统》2009年第9期1864-1868,共5页Journal of Chinese Computer Systems

摘　　要：分析运动目标跟踪系统中基于传统颜色直方图进行颜色特征描述的不足,提出一种使用"选择性"颜色直方图的新方法.该方法通过建立前景与背景的颜色分布模型,提取"选择性"颜色,利用它调整目标区域内每个像素点的统计权重,建立具有更强区分能力的"选择性"颜色直方图,放大目标与周围环境在颜色分布上差异最大的部分,因而在运动目标跟踪中能更好地把握目标特征.实验数据表明,与传统颜色直方图相比算法具有较强的匹配稳定性,在精确性和适应性上有较大提高.Based on the shortcomings of the traditional color histogram, a novel way is presented to use color information, where both foreground and background color distributions are considered in the framework for choosing the ＂Selective Colors （SCs）＂, which is the uppermost difference between the object and its surroundings. And then the importance of these chosen ＂Selective Colors＂ is consciously improved by a non-linear function to form the final Selective Color Histogram （SCH）. This color pattern, because of considering both foreground＇s and background＇s information, can describe object＇s color character better than the traditional ones. Experimental results show that when the SCH is used for moving object tracking, it is more discriminative and more reliable than the formal methods.

关 键 词：颜色直方图 目标跟踪 高斯混合模型 粒子滤波器 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]