Novel Aggregated Solutions for Robust Visual Tracking in Traf?c Scenarios

This work proposes novel approaches for object tracking in challenging scenarios like severe occlusion, deteriorated vision and long range multi-object reidenti?cation. All these solutions are only based on image sequence captured by a monocular camera and do not require additional sensors. Experiments on standard benchmarks demonstrate an improved state-of-the-art performance of these approaches. Since all the presented approaches are smartly designed, they can run at a real-time speed.