Video surveillance technology has been widely used in Intelligent Transportation System (ITS) to measure traffic flow parameters and detect accidents. Occlusion of moving objects by stationary or other moving foreground objects always causes tracking errors. This paper presents a novel module for vehicle tracking under the condition of occlusion, which can be conveniently added to the existing video surveillance systems as a complementary part. The appearance information of the object (shape and texture) is combined with the spatial-temporal Markov Random Field (MRF) based tracking model through a kind of special structuring elements in Mathematical Morphology (MM) theory. As a result, the sites in MRF are reduced, which saves computational cost greatly, and the assumption on the vehicle shape model in prior is not needed. Experiments on various kinds of image sequences show that the proposed module can effectively track vehicles with severe occlusion.