项目名称: 基于自适应分级稀疏模型的鲁棒目标跟踪研究
项目编号: No.61202323
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 计算机科学学科
项目作者: 韩振军
作者单位: 中国科学院大学
项目金额: 25万元
中文摘要: 目标跟踪技术是计算机视觉、图像理解领域的核心研究课题之一,且目标跟踪技术在视频会议、视频内容分析、视频检索、运动分析和合成等领域发挥了重要作用。由于在跟踪过程中,目标形态及运动状态可能发生明显变化等情况的存在使得对运动目标的跟踪变得更加困难。在本研究中,申请人拟通过信息融合技术以及自适应的分级稀疏模型对目标进行鲁棒的跟踪。多源信息融合的目标描述可以获取目标更完备的信息。然后,在跟踪过程中基于滤波框架将目标的自适应稀疏表示与稀疏重构结合起来,提出一种基于自适应分级稀疏模型的目标跟踪新框架,其不仅考虑紧致、自适应的目标模型,同时融合目标鲁棒匹配与定位新方法,即将目标跟踪中的关键问题(目标表观变化、目标遮挡)都统一到一个新的跟踪框架中,实现目标的自适应鲁棒跟踪。
中文关键词: 目标跟踪;稀疏表示;系数重构;信息融合;分层图模型
英文摘要: Object tracking plays an essential role in many applications and remains challenging open problems. Some of the previous research tends to fail when the object subjects to dynamic background,partial occlusion or image deterioration. To address these problems, we propose a new approach for robust visual object tracking via data fusion based adaptive two-stage sparse representation and reconstruction, where two main contributions are devoted in terms of object representation and location respectively.Firstly, we extract a new combined feature. Secondly, we design a new sparse feature to compactly represent the object appearance via an online learning algorithm. The designed feature is capable of ensuring the discrimination of the object representation against various appearance patterns. Thirdly, a reconstruction strategy using the sparse representation is proposed for the sake of obtaining accurate object location spatially. The sparse reconstruction facilitates the effective location of the object local parts even in the cases of partial occlusion or image deterioration. Finally, the sparse representation and reconstruction are integrated into a Kalman filter (KF) framework to develop a robust object tracker.
英文关键词: object tracking;sparse representation;sparse reconstruction;information fusion;layered graph model