摘要
针对单一颜色特征跟踪性能差的缺点,提出一种基于联合特征直方图的均值漂移目标跟踪算法.采用颜色、边缘、运动3个具有一定互补性的信息进行联合概率跟踪,以提高目标模型的描述能力;利用运动信息修正颜色和边缘模型核函数,用以克服背景颜色、轮廓对跟踪效果的影响;根据各特征所占权重,运用自适应融合法生成关联特征直方图,并将所构建的目标模型融合到mean shift跟踪框架中,实现目标跟踪.实验结果表明,所提出的算法具有较高的准确性.
For improving the performance of traditional tracking algorithm based on the single color feature, a target tracking algorithm is proposed based on the joint-feature histogram and mean shift. The description capacity of the target character is enhanced by combining the advantage of color, edge and movement. To overcome the tracking impact of environment＇s color and outline, the kernel function of color and edge model are corrected by using movement information, and the joint-feature histogram is gained according to the adaptive fusion method of multi-feature weights. Target tracking is realized by integrate the model into the mean shift tracking frame. Experimental results show the higher accuracy of the proposed method.
出处
《控制与决策》
EI
CSCD
北大核心
2016年第10期1845-1852,共8页
Control and Decision
基金
国家863计划项目(2009AA11Z207)
高等学校博士学科点专项科研基金项目(20110009110011)
关键词
目标跟踪
均值漂移
直方图分布
联合特征
target tracking
mean shift
histogram distribution
joint-feature
