3D contours of Las Vegas

These are 3D contours from LiDAR point cloud of Las Vegas. The QL-1 datasets (≤10cm vertical/≤35cm horizontal accuracy, ≥8 points/m²) required preprocessing due to excessive data volume (142GB for Santa Clara alone). Our method reduces data while preserving structurally critical line features for satellite image-LiDAR point cloud registration, focusing on building contours rather than less prominent road edges. First, building footprints were extracted using Google's 2D shape vectors instead of raw segmentation or classification. Misaligned vectors were corrected via cross-modal registration. Next, probabilistic plane extraction isolated roof geometries across varying building sizes and point densities, followed by alpha-shape-based contour derivation. No post-regularization was applied to maintain native accuracy (±10cm vertical/±35cm horizontal).