On the accuracy of particle filter-based object tracking

In particle filter framework, there are many parameters need to be set to have the object tracking work best. On the other hand, the success of an object tracking is also determined by the characteristics of the video being used. Different videos might require different values of parameters to obtain best performance. This paper investigates thoroughly the effect of parameters in object tracking within particle filter framework for videos with different characteristics. We also propose a new method in selecting particle number and matrix value that used in propagation model. This work finds best parameters for video with particular characteristics and contributes in providing guideline of parameter setting in particle filter-based object tracking.