38. A Wearable Walking-Aid System For Visual-Impaired People Based on Monocular Top-View Transform

This paper presents a wearable system which can provide walking-aids for visual-impaired people in an outdoor environment. Unlike many existing systems that rely on stereo cameras or combination of other sensors, the proposed system aims to do the job by using just single camera mounted at user’s belly. One of the main difficulties of using single camera in outdoor navigation task is the discrimination of obstacles with cluttered background. To solve this problem, this paper makes use of the inhomogeneous re-sampling property of top-view transform. By mapping the original image to a top-view virtual plane using top-view transform, background edges in the near-field are sub-sampled while obstacle edges in the far-field are oversampled. Morphology filters with connected component analysis are used to enhance obstacle edges as edge-blobs with larger size, whereas sparse edges from background are filtered out. Based on the identified obstacles, safe path can be estimated by tracking a polar edge-blob histogram on the top-view domain. To deliver the safe direction to the user, an audio message interface is designed. The system is tested in different outdoor scenes with complex road conditions, and its efficiency has been confirmed.