点击次数：

论文类型：期刊论文

发表时间：2018-11-07

发表刊物：NEUROCOMPUTING

收录刊物：SCIE、SSCI、Scopus

卷号：314

页面范围：132-142

ISSN号：0925-2312

关键字：Robust tracking; Visual saliency; Partial occlusion; Local sensitive Hash

摘要：A robust visual tracking system is expected to track the object accurately and rapidly in complex scenes with clutters, distracters and partial occlusion etc. These challenges remain the focuses of current research. Studies on human vision reveal that the human visual system is able to tackle these challenges efficiently by applying mechanisms including visual saliency and attention. Inspired by this research finding, this paper presents a salience-based tracking method for robust tracking. The major steps include salience computation, feature tracking and model updating. The salience of visual features is defined on their regions of interest. By examining visual saliency in regions of a visual object and its environment, we are able to estimate object salience and environment salience of extracted visual features for robust visual tracking. Experiments indicate that the proposed method can achieve promising performance in an environment with distracters and partial occlusion compared with the state of the art tracking methods. (C) 2018 Elsevier B.V. All rights reserved.