CROWD MOTION ESTIMATION IN SUBWAY CORRIDORS USING IMAGE PROCESSING
Samia Bouchafa, Larbi Beheim, Andreas Jahja Sadji, D. Aubert · 1997
This work provides a motion estimation method which takes into account two difficulties: real time constraint and deformable objects. Developed methods deal with crowd images. For this purpose, three techniques were chosen, each of them belonging to a specific class of method such as block matching (matching technique), optical flow (differential technique) and Gabor filter (frequential technique). Attempts to improve both the speed and the performance of each algorithm resulted in a very efficient optical flow technique suitable for various applications in the field of crowd monitoring and traffic management. For the covering abstract, see IRRD 490001.