作者: Bogdan Kwolek
DOI: 10.15388/INFORMATICA.2004.055
关键词:
摘要: This paper proposes the use of a particle filter combined with color, depth information and shape features as an efficient effective way to deal tracking head on basis image stream coming from mobile stereovision camera. The is modeled in 2D domain by ellipse. color distribution within interior ellipse represented histogram. histogram dynamically updated over time. length ellipse's minor axis determined information. particles representing candidate ellipses are weighted each time step respect intensity gradient near edge matching score histograms surrounding tracked object currently analyzed one. proposed algorithm can track reliably cases temporal occlusions well varying illumination conditions dealing multiple hypotheses for pose. Experimental results obtained long sequences show feasibility our approach perform undergoing complex changes appearance against background. tracker has been evaluated experiments consisting face real agent.