低秩–稀疏与全变分表示的运动目标检测方法
Moving object detection method based on low rank-sparse and total variational representation
摘要点击 233  全文点击 227  投稿时间:2018-07-23  修订日期:2019-04-30
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2019.80547
  2020,37(1):81-88
中文关键词  鲁棒主成分分析  低秩-稀疏  全变分  目标检测
英文关键词  robust principal component analysis (RPCA)  low rank-sparse  total variation  object detection
基金项目  国家科技部重点研发计划项目(2018YFC1312903)
作者单位E-mail
杨磊 上海大学 yanglei130@shu.edu.cn 
庞芳 上海大学 pangfang2018@163.com 
胡豁生 埃塞克斯大学-计算机科学与电气工程学院  
中文摘要
      针对含有动态背景的运动目标检测问题,本文提出了一种低秩-稀疏与全变分表示的运动目标检测方法.提出方法以鲁棒主成分分析(robust principal component analysis, RPCA)为基础,利用三维全变分对运动目标的约束,去除动态背景的干扰;同时利用低秩矩阵在正交子空间下系数的群稀疏性来加速低秩矩阵的秩最小化,弥补全变分的计算量大的问题,平衡整体运行速度.实验结果表明,该方法不仅能较好的把复杂背景下的运动目标检测出来,而且还保持了较快的运行速度.
英文摘要
      Moving object detection with dynamic background is addressed, a moving object detection method with low rank-sparse and total variation representation is proposed. The proposed method is based on robust principal component analysis (RPCA), and the three-dimensional total variation is constrained to the moving object, and the interference of the dynamic background is removed. At the same time, the group sparsity of the coefficients of the low rank matrix in the orthogonal subspace is used to accelerate the rank minimization of the low rank matrix, and the calculation of the total variation is large. The problem is to balance the overall running speed. The experimental results show that the method can not only detect the moving objects in complex background, but also maintain the faster running speed.