Distance-invariant sign detection in high-definition video

Abstract

High-Definition video can provide useful information regarding distant objects. For example, when trying to detect traffic signs. However, current detection methods have not been thoroughly tested in terms of their ability to detect signs over changing distance. In this work, we present such an experimental comparison of detection methods. We also present a novel object detection method based on colour cooccurrence data, specifically designed to provide robust detection with varying sign distance. Experimental results show that our new method performs competitively with existing methods-providing consistent detection accuracy despite distance between observer and sign.