A Top-down Building Area Extraction Scheme from Aerial LiDAR Data

초록

A top-down building area extraction scheme is proposed which is able to support real-time disaster managements for detection of changes and damages on buildings from very large volume of aerial LiDAR data. To overcome the heavy processing time problem of the previous point-based schemes, the proposed scheme adapts a top-down approach instead of bottom-up approach in which iterative point-based computations are required. In the proposed scheme, LiDAR data is classified by the tile-based topographical classification using data mining technique. Then, the tiles are grouped by the tile-based region growing to form same building cluster, and the point-based calculation extracts the building edge candidate points, which are located on the outer edge of building and terrain area, only from the exterior tiles of each building cluster. Finally, the building area is constructed by a convex-hull algorithm with the building edge candidate points. From the experimental result with the aerial data of Daejeon, the proposed scheme can extract correctly building areas with the reduced processing time of up to 66%.