Scale Invariant Kernel-Based Object Tracking

摘要：

　　Traditional kernel-based object tracking methods are useful for estimating the position of objects,but inadequate for estimating the scale of objects.In this paper,we propose a novel scale invariant kernel-based object tracking (SIKBOT) algorithm for tracking fast scaling objects through image sequences.We exploit the set property of regions and propose a new method to estimate the potential of the intersection of the object and the kernel.Regarding robustness,we iteratively estimate the scale of the object by means of basic set analysis.The scale and position of objects are simultaneously estimated by mean shift procedures in parallel.The proposed SIKBOT algorithm is demonstrated by extensive experiments on challenging real-world image sequences.

关键词： kernel tracking mean shift set analysis

作者: Peng Li Zhipeng Cai Hanyun Wang Zhuo Sun Yunhui Yi Cheng Wang Jonathan Li

作者单位: School of Electronic Science and Engineering, National University of Defense Technology, Changsha, C Department of Computer Science, School of Information Science and Technology, Xiamen University, Xia Department of Computer Science, School of Information Science and Technology, Xiamen University, Xia

会议类型: 国际会议

会议名称: 2012年遥感中的计算机视觉国际会议

会议地点: 厦门

会议语种:英文

页码: 1-4

在线出版日期: 2012-12-16（万方平台首次上网日期，不代表论文的发表时间）