Extraction of Roof Planes from Different Point Clouds Using RANSAC Algorithm

Abstract

Solar energy is a renewable energy source directly from sunlight and its production depends on roof characteristics such as roof type and size. In solar potential analysis, the main purpose is to determine the suitable roofs for the placement of solar panels. Hence, roof plane detection plays a crucial role in solar energy assessment. In this study, a detailed comparison was presented between aerial photogrammetry data and LIDAR data for roof plane recognition applying RANSAC (Random Sample Consensus) algorithm. RANSAC algorithm was performed to 3D-point clouds obtained by both LIDAR (Laser Ranging and Detection) and aerial photogrammetric survey. In this regard, solar energy assessment from the results can be applied. It is shown that, the RANSAC algorithm detects building roofs better on the point cloud data acquired from airborne LIDAR regarding completeness within model, since aerial photogrammetric survey provides noisy data in spite of its high-density data. This noise in the source data leads to deformations in roof plane detection. The study area of the project is the campus of Istanbul Technical University. Accuracy information of the roof extraction of three different buildings are presented in tables.