This paper presents a method for tracking moving objects in video sequences. The tracking algorithm is based on a graph representation of the problem, where the solution is found by the minimization of a matching cost function. Several cost functions have been proposed in the literature, but it is experimentally shown that none of them, when used alone, is sufficiently robust to cope with the variety of situations that may occur in real applications. We propose an approach based on the combination of cost functions, showing that it enables our system to overcome the critical situations in which a single function can show its weakness, especially when the frame rate becomes low. Experimental results presented for video sequences obtained from a traffic monitoring application, confirm the performance improvement of our approach.