Real‐time human tracking system with multiple omnidirectional vision sensors
Takushi Sogo, Hiroshi Ishiguro, Mohan Manubhai Trivedi · Systems and Computers in Japan · 2004
Abstract We propose a real‐time human tracking system that uses multiple omnidirectional vision sensors. This system estimates and tracks the locations of multiple persons from omnidirectional 360‐degree images captured by multiple omnidirectional vision sensors using N‐ocular stereo, which is obtained by extending conventional trinocular stereo, and without using image characteristics. In N‐ocular stereo, robust real‐time target location estimation can be achieved by taking into account human model errors and human azimuth observation errors, as well as the problem of measurement precision of target locations in stereo based on omnidirectional images; and by further simplifying these factors and correcting them. Using this method, we created a prototype system and verified its effectiveness based on a human tracking experiment. © 2004 Wiley Periodicals, Inc. Syst Comp Jpn, 35(2): 79–90, 2004; Published online in Wiley InterScience ( www.interscience.wiley.com ). DOI 10.1002/scj.1228