As a nonparametric gradient climbing technique, Mean Shift algorithm finds local mode of any arbitrary feature space by doing an optimistic search in the neighborhood, which is defined by the target location found in the last frame.

Mean Shift estimation is an efficient tracking method when it is hard to define the proper parameters for the problem model and it is a fast algorithm which can be implemented in real time. A mean-shift based vehicle tracking system has been developed and is illustrated as follows.

• Feature Space Selection

• Probabilistic Modeling for Model and Candidates

• Mode Seeking on Weighted Image