Real-Time Indoor Staircase Detection on Mobile Devices

摘要

Staircases are among the most common architectural features in urban environments. In this paper, we describe a method for detecting indoor staircases from depth images. The objective of this development is to aid visually impaired people in perceiving the environment, especially in unfamiliar places. This method is based on the detection and clustering of patches that have the surface normal vectors oriented towards the "up" direction. A series of filters is applied in order to eliminate false-positive patches. In addition to the depth information, we use the camera's absolute orientation information, provided by an IMU sensor. This additional information is used to correct the orientation of the surface normal vectors in such a way that no matter how the camera moves, the surface normal vectors have the correct orientation. This method provides real-time performance on a mobile device with low computational capabilities and it was tested on real-case scenarios, showing promising results.