Adaptive tracking window updating algorithm based on particle filtering

Practical tracking system must be able to adjust the tracking windows adaptively according to the size-changes of the tracked objects; otherwise it can not track the objects with obvious size-changes accurately. Based on the visual theory, and combined with the primal sketch of the objects extracted by the Otsu method as well as the changes of the elements-number as the measure information, this paper proposed a new automatic tracking window scale updating algorithm, which was then used to improve the particle filtering algorithm based on color histogram. Experimental results demonstrated that the improved tracking algorithm can adjust the tracking window scale adaptively to obtain a stable tracking for the objects with obvious size-changes, increasing or decreasing.