摘要
为了提高相似物体干扰、光照变化等复杂环境下目标跟踪的稳定性,提出利用Fisher准则的在线选择鉴别性特征,将在线特征选择嵌入到跟踪算法中。但是,在目标被严重遮挡时仍在线选择鉴别性特征,会使目标模板偏移,导致跟踪失败。为了在目标被严重遮挡的情况下仍能在线选择鉴别性目标特征,选择一个可靠的子区域来推测被遮挡目标的颜色特征,再选择鉴别性特征。实验结果表明该方法在相似物体干扰、被跟踪目标被遮挡等复杂环境下极大地改善了跟踪的稳定性。
In order to improve the robustness of tracking system under complex surrounding such as similar object interfe-rence,illumination change,this paper proposed that employing Fisher criteria online selecting color feature mechanism,which was embedded in target tracking algorithm.However,when the target was severely occluded,still online selecting discriminative feature,would make target template offset,and lead to tracking failure.In order to select discriminative feature in the case of tracked target was severely occluded,selected a reliable sub-region of the occluded target and utilized reliable sub-region infer the color feature of the occluded target area,then online selected color feature.Experimental results show that this paper proposed method can greatly improve the robustness of target tracking method under complex surrounding such as similar to object interference,tracked target being occluded.
出处
《计算机应用研究》
CSCD
北大核心
2011年第7期2797-2800,共4页
Application Research of Computers
基金
海南省自然科学基金资助项目(60897)
海南省教育厅资助项目(Hj2009-135)
湖南省教育厅资助项目(10C1263)
关键词
在线选择鉴别性颜色特征
遮挡处理
目标跟踪
online selecting discriminative color feature
occlusion processing
target tracking