Enhancing color-based particle filter algorithm with ORB feature for real-time video tracking

In this paper, we propose an enhancement of color-based particle filter algorithm with oriented FAST and rotated BRIEF (ORB) feature detector (ORBPF). By carefully defining the size and position of the search window, ORB will generate interesting key-point close to the object being tracked. The location of matched key-point with high color similarity is then selected to replace the particles from the original filter. This matched key-point can be used to manage the particle spread and minimize the disadvantages of a random selection of the particle set such as degeneracy and sample impoverishment problem. The number of particles can be significantly reduced by using this method while maintaining the accuracy of the tracker. We test our tracker on numerous video sequences and using real-time video input from CCD camera in an indoor and outdoor application that involves some challenging situations including variations in illumination, scale, rotation, and occlusions. From the experiment, we show that our tracker performs more efficiently than standard color-based particle filter without ORB feature detector (CPF).