摘要
为了克服灰度目标跟踪中目标和背景的对比度低、界线模糊的情况,提出了融合梯度特征的灰度目标跟踪方法.该算法采用mean-shift迭代过程,是一种用新的概率密度向量代替以亮度为特征建立概率密度向量的跟踪方法.目标的梯度特征描述了目标的纹理等细节信息,增加了目标和背景的对比度,因此通过提取目标的梯度特征,建立梯度概率密度向量,与灰度概率密度向量融合构成新的向量,用新概率密度向量描述目标模板的特征跟踪目标.实验结果表明,该算法与mean-shift算法相比,跟踪稳定得到了较大改进,取得了理想的跟踪效果.
In order to deal with low contrast and boundary ambiguous between object and background in object tracking,a new gray object tracking method is presented.This method adopts mean-shift procedure and using new PDF(probability density function)replaces traditional PDF which only selects gray feature to represent targets.Grads of targets shows much texture information of targets and increases contrast between targets and background,but it is unstable.So we use new PDF vectors to describe objects which combine g...
出处
《微电子学与计算机》
CSCD
北大核心
2009年第2期69-71,共3页
Microelectronics & Computer
基金
国家“八六三”计划项目(2005AA778032)