低秩矩阵和结构化稀疏分解的视频背景差分方法
A Background Subtraction Method Based on Decomposition of Low Rank and Sparsity Matrices

针对基于矩阵分解的视频前景检测传统算法中忽视前景元素之间相关性会导致检测结果容易受噪声干扰和运动目标检测不完整等问题,提出了一个低秩矩阵和结构化稀疏分解的视频背景差分算法。该算法充分考虑到视频前景区域的结构化分布特征,利用结构化稀疏范数对前景进行约束;针对矩阵分解方法中参数选择的难题,采用了一种基于运动显著性判定的两步法来实现动态背景去除和正则化参数的自适应选择,即第一步利用低秩和结构化稀疏分解获得运动候选块,第二步对运动候选块进行显著性分析并利用自适应正则化参数的块稀疏分解进行前景检测。实验结果表明:与现有的基于矩阵分解的前景检测方法相比,该算法能够更加适应复杂多变的视频环境,在I2R测试库中检测出的前景有较高的精确度和召回率。
A background subtraction method based on decomposition of low？？rank and structured sparsity matrices is proposed to solve the problem that detection results are sensitive to noise and incomplete caused by ignoring the relationship between foreground pixels in traditional foreground detection methods based on matrix decomposition. The method takes the structural distribution of the foreground into account, and a structured sparsity constraint is used on the foreground pixels. Moreover, a two？？stage framework based on motion saliency is introduced to address the parameter setting issue in dynamic background videos and to tune regularization parameters adaptively. Motion block candidates are obtained by using the low rank and structured sparsity decomposition in the first step. Then, motion saliency analysis is applied to these candidates and the adapt block sparsity decomposition is used to detect the foreground in the second step. Experimental results show that the performance of the proposed method is more adaptive than the existing foreground detection methods based on matrix decomposition in complex videos, and that the proposed approach outperforms the state？？of？？the？？art methods according to the precision and recall results on dataset I2R