We present a mobile vehicle classification technique achieved by tracking two vehicle based Points of Interest (PoI) in multiple filter configurations to compose a vehicle specific 3D geometry. Using high fidelity physics based simulation we demonstrate the capability to classify the 3D geometries in the presence of noise by extracting vector magnitudes and angles as features. Additionally, we investigate the classification advantages presented by representing the features in multiple linear transform domains and fusing the information from those different domains into a single ensemble classifier.