Real-time floor recognition in indoor environments using TOF Camera

Navigation in public spaces is an essential technique for pedestrian guidance and autonomous robot assistance. Navigation systems mainly consist of positioning systems, maps, route finding, and a user interface. Image acquisition or 3D sensing is also required to recognize the unknown environment around a pedestrian or robot. Therefore, we aim to develop a sense-and-avoid application for pedestrians and autonomous robots. We propose a real-time methodology to recognize the mobility area in an indoor environment using an active depth imager. In our experiments, we used a handheld time-of-flight (TOF) camera. We selected corridors, stairs, large rooms, and our laboratory in our campus as study areas. These areas consist of walls, glass walls, steps, and gaping holes. These objects were recognized with online processing, and notification for obstacle avoidance was issued after object recognition. Our experiments confirmed that our approach can process 3D measurements, classify objects, and notify for obstacle avoidance in real time.