Visual tracking using harmony search

Abstract

In this chapter we present a novel method for tracking an arbitrary target through a video sequence using the Harmony Search algorithm called the Harmony Filter. The Harmony Filter models the target using a color histogram and compares potential matches in each video frame using the Bhattacharyya coefficient. Matches are found using the Improved Harmony Search (IHS) algorithm. Experimental results show that the Harmony Filter can robustly track targets in challenging environments while still maintaining real-time performance. We compare the runtime and accuracy performance of the Harmony Filter with other popular methods used in visual tracking including the particle filter and the Kalman Filter. We show that the Harmony filter performs better in both speed and accuracy than similar systems based on the particle filter and the Unscented Kalman Filter (UKF). © 2010 Springer-Verlag Berlin Heidelberg.