Publications / 2004 Proceedings of the 21st ISARC, Jeju, South Korea
In this paper, an algorithm for door detection using a cheap PC-camera is developed based on the Bayesian network in a hallway where doors are located at various depth. The open/close status of the door is also important in developing a GVG-SLAM and we extracted five important door features from the image and distance data. We applied the Bayesian network algorithm and an entropy concept to increase the robustness. Experiment result shows that the proposed algorithm can help the GVG-SLAM of mobile robot by successfully identifying the status of the door.