基于AOS的运动目标检测算法

从能量极小化的观点出发，对运动目标检测问题定义了新颖的能量模型，给出了相应的曲线进化方程．结合水平集算法，所提出的曲线进化模型能够自动地适应拓扑变化，检测出多个运动目标区域及轮廓．定义了基于对数差图像的能量项，使得新模型对于场景照明不敏感．采用加性算子分裂(AOS)策略作为无条件稳定的数值方案来加速算法的快速实现。

Moving Object Detection Algorithm Based on AOS Scheme

A novel energy model is defined for moving object detection according to energy-minimizing conception and the corresponding curve evolution equation is given. Combined with level set method, the proposed model can adapt to topological change automatically, and detect multiple moving targets with their contours. Based on the difference of logarithmic images, a novel image energy term is designed to make it adaptable to large variation of scene illumination over time domain. An additive operator splitting (AOS) based scheme is introduced as unconditionally stable numerical method to speed up the proposed algorithm.