Abstract— This paper presents results of the project in the course Applied Estimation at KTH, Stockholm. The project presents a method for visual color based object tracking using particle filter. By pre-processing the data with help of a color based RGB computer vision technique the filter tracks objects through noisy scenes. Colors will be compared from frame to frame for tracking. By using an arbitrary scene the motion model is set to be Gaussian. Two sets of re-sampling methods has been studied, multinomial and systematic. The result verifies that the tracking algorithm has fulfilled its mission to track an object in a video.
Keywords: computer vision; object tracking; particle filter; weighting.