Classification of residential building architectural typologies using LiDAR

The availability of light detection and ranging (LiDAR) datasets over urban areas has significant potential to facilitate the automatic parameterization of sophisticated building energy models. In this paper we present an approach to building architectural typology classification using LiDAR data and decision tree regression. By integrating a suite of LiDAR derived building morphological characteristics with field training data we accurately classified (84%, Kappa = 0.76) of the modelled residential building types. Furthermore, our analysis suggests that building characteristics related to height, volume and roof slope provide the most important predictor variables for classifying building typologies in the examined study area.