ROBUST OBJECT TRACKING METHOD BY USING 3D SPATIAL-TEMPORAL MARKOV RANDOM FIELD

摘要：

In this paper a novel tracking method based on 3D Spatial-Temporal Markov Random Field (3D S-T MRF) is proposed. By taking temporal axis into account the traditional spatial MRF is extended to 3-D. Object tracking can be regarded as a labeling problem, i.e. assigning every pixel a label 0 or 1, of which 1 stands for tracking region and vice versa. Through defining proper 3D MRF structure, such as the nodes, neighbor system, data energy and smooth energy function, color cue and motion cue can be fused naturally. The labeling problem comes down to an energy minimization problem. Considering the simplicity and efficiency, Iterated Conditional Method (ICM) algorithm is used to minimize the energy function. The experiments show that this method can get promising results in challenging background and to some extent is robust against occlusions owing to the fusion of motion and color information through energy function.

关键词： Tracking Markov Random Field Energy Function Information Integration.

作者: Huijun He Hong Liu

作者单位: Key Laboratory on Machine Perception and Intelligence,Peking University,P.R.China

会议类型: 国际会议

会议名称: 2008高等智能国际会议(2008 International Conference on Advanced Intelligence)

会议地点: 北京

会议语种:英文

在线出版日期: 2008-10-18（万方平台首次上网日期，不代表论文的发表时间）