In this article we present a new method for detecting textured moving objects. Based on a known background estimation and a fixed camera, the algorithm is able to detect moving objects and locates them at video rate, moreover this method is used for object tracking purposes. Our method is multi-step: First, we use level lines to detect pixels of the background which are occluded by moving object. Then, we use an a contrario model as general framework to make an automatic clustering. Thus the moving objects are detected as regions and not only as pixels, eventually we correct this region to better fit the moving object. Experimental results show that the algorithm is very robust to noise and to the quality of the background estimation (e.g. ghosts). The algorithm has been successfully tested in video sequences coming from different databases, including indoor and outdoor sequences.