Change detection based on graph cuts

In this paper we propose a moving object detection system based on Graph Cut. Our method relies on motion modelling using an optical flow algorithm and a classical background subtraction module based on Mixture of Gaussians. The main contribution in our approach is the fusion of the two mask models, as well as the particular cost function used by the Graph Cut algorithm. The experiments as performed on CDnet 2014 benchmark showed that our system has very good results in scenarios such as Bad Weather or PTZ, but is less robust in detecting small changes in the scene.