Tracking of human head with particle filter

GUO Chao



Author NameAffiliation

GUO Chao Dept. of Information Science and Electronic Engineering, Zhejiang University, Hangzhou 310027, China 



Abstract:

To cope with the problem of tracking a human head in a complicated scene, we propose a method that adopts human skin color and hair color integrated with a kind of particle filter named condensation algorithm. Firstly, a novel method is presented to set up human head color model using skin color and hair color separately based on region growing. Compared with traditional human face model, this method is more precise and works well when human turns around and the face disappears in the image. Then a novel method is presented to use color model in condensation algorithm more effectively. In this method, a combination of edge detection result, color segmentation result and color edge detection result in an Omega window is used to measure the scale and position of human head in condensation. Experiments show that this approach can track human head in complicated scene even when human turns around or the distance of tracking a human head changes quickly.

Key words:  human head tracking  condensation  particle filter  Hausdorff distance

DOI£º10.11916/j.issn.1005-9113.2009.02.025

Clc Number:TP391

Fund: