3-D building reconstruction with ARUBA: a qualitative and quantitative evaluation

Abstract

Reliable and accurate 3-D reconstruction of man-made objects is essential for many applications using digital 3-D city models. Manual reconstruction of buildings from aerial images is time consuming and requires skilled personnel, hence large efforts are being directed towards the automation of building detection and reconstruction. In this paper we present ARUBA — a framework for automated 3-D building reconstruction. After highlighting our strategy and concisely describing the framework and its modules, we evaluate the reconstructed roofs relative to accurate reference data based on three criteria: completeness, geometric accuracy and shape similarity. Finally, we interpret the results of the performance evaluation and make suggestions for improvements.