Feature-aware surface interpolation of rooftops using low-density lidar data for photovoltaic applications

Abstract

Digital surface models (DSM) are used to estimate the solar irradiation on rooftops. Estimates are more accurate when the precise geometrical characteristics of roofs are well represented in the DSM. The existing DSM covering Switzerland has a low accuracy for buildings. It was derived from a low density Lidar dataset with an average point density of 0.5 points per square meter. In this paper, we present a method to interpolate a DSM from point cloud data focusing on geometric modelling of rooftops. The method uses a combination of robustly fitted planes to local point clouds and inverse distance weighting interpolation. It was applied to roughly 3 million buildings, and compared to a reference DSM from a high density point cloud, which revealed a significant reduction of error compared to the existing DSM.