Moving Object Detection in Satellite Videos via Spatial–Temporal Tensor Model and Weighted Schatten <i>p</i>-Norm Minimization

Low-rank matrix decomposition approaches have achieved significant progress in small and dim object detection satellite videos. However, it is still challenging to achieve robust performance fast processing under complex highly heterogeneous backgrounds since video data can neither adequately fit the foreground structure nor background model existing models. In this letter, we propose a novel method based on spatial–temporal tensor structure. First, construct exploit inner spatial temporal correlation within video. Second, extend formulation with bounded noise backgrounds. This integrates low-rank background, structured sparse foreground, their noises into problem. For separation, weighted Schatten $p$ -norm incorporated provide adaptive threshold obtain singular value of tensor. Finally, proposed solved using alternative direction multipliers (ADMM) scheme. Experimental results various real scenes demonstrate superiority against compared approaches.