Omnidirectional vision system developed for a home service robot

An omnidirectional vision system developed for a home service robot is described. The system is built in low cost using off-the-shelf components. The vision system is mounted on a mobile home robot, which is controlled by a PC through wireless communication. Two applications of the vision system are developed; intruder and fire detection. An adaptive background subtraction technique is used to extract intruders in image sequence. A novel vision processing algorithm for fire alarm is also proposed. This algorithm processes images in three steps; pixel-level, block-level, and global level.