Abstract
In order to maintain the safety of elevators in buildings, it is important to find out if passengers have become trapped inside an elevator car. The purpose of our work is to develop a system that can recognize a trapped status by using only a single camera installed for surveillance. In this paper, we propose a new method which consists of two modules. The first module detects door motion. By utilizing a technique of spatio-temporal image analysis, reliable door motion detection was achieved. The second module detects passengers by the difference between two sequential images. We also proved the effectiveness of this method using real image sequences.
