基于线性动态系统稀疏编码的异常事件检测

Abnormal Event Detection Using Linear Dynamical System Combined with Sparse Coding

线性动态系统模型结合稀疏编码实现异常事件检测。线性动态系统可有效地捕捉动态纹理在时间和空间的转移信息,描述视频的时空小块。然而,线性动态系统属于非欧氏空间,无法直接用传统的稀疏编码进行异常检测。基于约束凸优化公式,将相似性变换与稀疏编码结合,可实现线性动态系统稀疏编码的优化求解。实验表明,所提出的方法具有更好的性能。

Linear dynamical system model combined with sparse coding is used to achieve abnormal event detection. Linear Dynamical System（LDS） as a description for dynamic texture can capture the transition of appearance and motion effectively. LDS is applied to describe spatio-temporal cuboids. Since LDS does not belong to Euclidean space, traditional sparse coding techniques can not be applied. Similarity transformation combined with sparse coding based on a principled convex optimization formulation can deal with the optimization of spare coding. The results show that the proposed algorithm has better performance and outperforms the earlier approaches.