摘要
针对基于反稀疏表示跟踪方法存在的跟踪准确性不高的问题,提出了一种结合时空上下文和反稀疏表示的目标跟踪方法。首先,使用模板对目标进行表示,并利用粒子滤波的方法生成相应的候选目标。然后,利用时空上下文置信图的方法对候选目标进行优化。最终,再由选出的候选目标和目标模板利用反稀疏表示方法得到最终的跟踪结果。实验结果验证了论文方法在跟踪准确性方面的有效性。
Aiming at the low accuracy problem in tracking methods based on the inverse sparse representation,an improved tracking method combining the spatio-temporal context with the inverse sparse representation is proposed.First of all,the partial filter method is used to generate candidates with the object template.Then,spatio-temporal context confidence is adopted to optimize the candidates.Finally,tracking result is gotten by the inverse sparse representation with the candidates and target template.The experiment results validate the proposed method’s performance.
作者
赵俊齐
伍海龙
刘婕
刘朝荣
ZHAO Junqi;WU Hailong;LIU Jie;LIU Chaorong(College of Electrical and Information Engineering,Lanzhou University of Technology,Lanzhou 730050;National Demonstration Center for Experimental Electrical and Control Engineering Education,Lanzhou University of Technology,Lanzhou 730050;Key Laboratory of Gansu Advanced for Industrial Processes,Lanzhou University of Technology,Lanzhou 730050)
出处
《计算机与数字工程》
2019年第12期3015-3019,3024,共6页
Computer & Digital Engineering
基金
国家自然科学基金项目(编号：61461028)
甘肃省自然科学基金计划(编号：1508RJZA092)资助
关键词
目标跟踪
反稀疏模型
时空上下文
粒子滤波
object tracking
inverse sparse model
spatio-temporal context
particle filter
