Visual object tracking serves a fundamental role in many vision systems such as bio-metric analysis, medical imaging, smart traffic systems and video surveillance. The first part of this thesis presents two new methods for single object visual tracking. The first method tackles the occlusion problem in visual tracking and the second method formulates the visual tracking problem in a framework of maximum entropy reinforcement learning. The rest of the thesis presents two applications of visual object tracking in transportation where multiple object tracking and vehicle re-identification are investigated. This thesis also covers an exploration of the scenario of multiple camera and multiple object tracking.