摘要
目前,复杂场景下的目标跟踪仍然是智能视频监控领域的一个难点问题,论文针对遮挡和目标尺度变化这两种复杂场景,提出了一种改进的Camshift算法,该算法采用分块跟踪来处理遮挡问题,同时,考虑到遮挡时不应更新模板,目标尺度变化时应该更新模板,采用自适应模板更新机制来局部更新模板,避免错误的模板更新,最后,引入了对目标尺寸变化具有鲁棒性的几何关系直方图来辅助描述运动目标。实验结果表明,该算法既能很好地适应目标的尺度变化,又具有较好的抗遮挡能力,能够保持良好的跟踪效果。
At present,target tracking under complex scenes is still a difficult problem in the field of computer vision.For occlusion and the scale change of targets,which are two kinds of complex scenes,an improved Camshift algorithm is proposed in this paper.Firstly,the improved algorithm adopts block tracking to handle occlusion problem.Secondly,considering that targets template should update when scale changes but not when occlusion,adaptive template-updating mechanism is proposed.Finally,the geometric histogram,which is robust to scale variation,is introduced to describe the moving targets.The experimental results show that,the improved algorithm can well adapt to scale change of targets,and also has good anti-occlusion ability,can achieve a good tracking effect.
出处
《计算机与数字工程》
2015年第6期1018-1022,共5页
Computer & Digital Engineering
