Title :
Viewpoint-oriented human activity recognition in a cluttered outdoor environment
Author :
Eftakhar, S. M Ashik ; Tan, Joo Kooi ; Kim, Hyongseop ; Ishikawa, Seiji
Author_Institution :
Dept. of Control Eng., Kyushu Inst. of Technol., Fukuoka, Japan
Abstract :
Motion analysis in a cluttered outdoor scenario is a real challenging task. In this paper, a suitable human activity recognition approach is proposed that adapts with the noisy environment and with variable viewpoints. The recognition approach encompasses some motion processing methodologies having high precision for specific analysis, among those are: motion segmentation, optical flow computation, Motion History Image (MHI) generation, structured database development, and so on. Estimating the subject´s moving body region, 4-directional MHIs of each activity are generated. An efficient pre-developed motion database aids in recognition of the activities. The proposed approach proves its significant precision in recognition for various activities from varying viewpoints.
Keywords :
image motion analysis; image segmentation; image sequences; pattern recognition; MHI generation; cluttered outdoor environment; cluttered outdoor scenario; human activity recognition approach; motion analysis; motion database; motion history image generation; motion processing methodology; motion segmentation; noisy environment; optical flow computation; structured database development; viewpoint-oriented human activity recognition; Computer vision; Databases; Humans; Image motion analysis; Motion segmentation; Optical computing; Optical imaging; cluttered outdoor; human activity recognition; motion history image; motion segmentation; optical flow;
Conference_Titel :
SICE Annual Conference 2010, Proceedings of
Conference_Location :
Taipei
Print_ISBN :
978-1-4244-7642-8
