摘要
基于粒子滤波提出了一种优化的目标跟踪算法——PFTMT算法。PFTMT算法在粒子滤波算法抗遮挡性强基础上结合了模板匹配算法跟踪精度高的优点,同时克服了2种算法各自的缺点,即该算法在保留粒子滤波算法的抗遮挡性强的基础上提高了粒子滤波算法的跟踪精度。PFTMT算法首先使用粒子滤波算法对目标进行大致定位,在该定位结果的基础上小范围进行模板匹配,与目标模板相似度最大的位置就是目标的精确定位。PFTMT算法的时间复杂度虽然略高于粒子滤波算法,但是可以满足目标实时跟踪的要求。实验结果验证了PFTMT算法的有效性。
An optimized object tracking algorithm based on particle filter and template matching is proposed in this paper.The proposed PFTMT （Particle Filter and Template Matching based Tracking ） algorithm integrates both the merit of the anti-occlusion of the particle filter and that of the high accuracy of the template matching .Firstly, standard particle filter algorithm is used to roughly locate the object in current frame .Then, template matching is done in the local region to find the accurate object location which corresponds to the local maxima similarity value between candidate and template .The time complexity of the proposed algorithm is a little higher than standard parti-cle filter , but it can still meet the requirement of real-time tracking application .Experimental results and their anal-ysis have demonstrated preliminarily the effectiveness of our approach .
出处
《西北工业大学学报》
EI
CAS
CSCD
北大核心
2013年第6期967-973,共7页
Journal of Northwestern Polytechnical University
基金
航空科学基金(2011ZC53036)资助
关键词
粒子滤波
模板匹配
视觉跟踪
PFTMT算法
algorithm, efficiency, error analysis, flowcharting, location, Markov processes, probability distribu-tions, optimization, pixels, target tracking, template matching, particle filter, PFI＇MT algorithm, visual tracking
