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.
出处
《小型微型计算机系统》
CSCD
北大核心
2009年第9期1864-1868,共5页
Journal of Chinese Computer Systems
