Automatic Recognition of Signs for Enhanced Navigation Awareness, and Localization

GPS technology provides a warfighter in unfamiliar territory with up-to-date location data. However, in dense urban and heavily canopied environments, or in the presence of external interference, the effectiveness of GPS-based navigational systems can be severely compromised. Warfighters require a navigational system capable of providing localization data and improving situational awareness even under GPS-suppressed conditions. We propose a system called Automatic Recognition of Signs for Enhanced Navigation Awareness, and Localization (ARSENAL) capable of automatically detecting and recognizing road signage that would directly address this problem by providing precise geographic location (e.g. intersection of two roads) as well as general awareness (e.g. speed limits, “road closed/dead end”). ARSENAL identifies regions in an image that exhibit sign-like properties based on cues common to most signs like the presence of text, vivid color, planar surfaces, compact shape, and retroreflectivity. These cues are fused to produce sign hypotheses which are accumulated over time to classify sign regions with high confidence. Finally, symbolic information (e.g. embedded text) is segmented and recognized from the detected signs, improving situational awareness and providing accurate navigational data.