Online Detection of Repeated Structures in Point Clouds of Urban Scenes for Compression and Registration

Sam Friedman, Ioannis Stamos

Abstract: Laser range scans of urban areas have a distinctive geometry dominated by facade and ground planes and repetitive regular fenestration. Detection of these ubiquitous features provides profound insights into the scene. We present a novel method for detecting major planes and repetitive architectural features. Armed with this knowledge we illustrate its application in compression and registration of range scans. What is more our algorithm operates online, processing the scan as it is retrieved by the scanner. This realtime approach opens up new possibilities in range data segmentation, compression and registration.