• 
• 
• 
• 
• 
• 

Wheelchair Detection Using Stereo Vision

The goal of this project is to automatically discriminate between ambulatory people and wheelchair users waiting at an elevator, using computer vision. A trinocular stereo system is used to (a) detect newly-appeared objects in front of the elevator, (b) recover 3D shape of the objects, (c) use 3D shape to identify wheelchair users.

Background & Objective:  Many elevators already provide modified service when a wheelchair is present. But the presence of the wheelchair is signaled by manually pressing a special button - we wish to replace this button with automatic wheelchair detection.

Technical Discussion:  A trinocular stereo rig is mounted at the top of an elevator frame looking toward the exterior environment. From a single image, there  are no strong visual cues to discriminate between an ambulatory person and a wheelchair user (the wheelchair itself is only partly visible given this type of frontal view). But there are valuable cues with multiple images, and we use stereo for measurements of height and 3D shape, plus temporal analysis to detect the foot motion of a walking person (versus the rigid feet of a wheelchair user).

Technology Area:  Computer Vision

Modification Date:  June 13, 2008