基于颜色和轮廓特征的目标视觉跟踪算法研究  

A Visual Object Tracking Algorithm Based on Features of Color and Profile

作　　者：连红森[1] 乔兵[1] 胡鹏[1] 

机构地区：[1]南京航空航天大学航天学院,南京210016

出　　处：《弹箭与制导学报》2011年第3期76-80,共5页Journal of Projectiles,Rockets,Missiles and Guidance

摘　　要：针对目前大多数视觉跟踪算法仅采用颜色特征进行跟踪,易导致跟踪错误的问题,提出了一种基于颜色和轮廓特征的改进粒子滤波算法。算法利用粒子滤波(PF)对目标进行预测,通过Camshift搜索对粒子进行优化,根据搜索窗口面积信息将目标的颜色和轮廓信息自适应的结合起来。实验表明,该算法在处理遮挡和相似色干扰方面优于传统的PF+Camshift跟踪算法,且具有良好的鲁棒性和实时性。To solve the problem of tracking error of the most conventional visual tracking algorithms by only using color feature, an improved tracking algorithm with particle filter based on both color and profile features of the target was proposed, in which the particle filter was applied to predict the state of the target and the Camshift was used to optimize the particles. Meanwhile, the color information and the profile feature of the target were adaptively fused through the area of the tracking window. The tracking results show that the algorithm has a good real-time performance and robustness and is more effective than the traditional PF＋Camshift algorithm in dealing with shielding and similar color interfering.

关 键 词：视觉跟踪 轮廓信息 颜色信息 粒子滤波 CAMSHIFT 

分 类 号：TN911.73[电子电信—通信与信息系统] V443.3[电子电信—信息与通信工程]