Estimating the height of buildings and vegetation in single aerial images is a challenging problem. A task-focused Deep Learning (DL) model that combines architectural features from successful DL models (U-NET and Residual Networks) and learns the mapping from a single aerial imagery to a normalized Digital Surface Model (nDSM) was proposed. The model was trained on aerial images whose corresponding DSM and Digital Terrain Models (DTM) were available and was then used to infer the nDSM of images with no elevation information. The model was evaluated with a dataset covering a large area of Manchester, UK, as well as the 2018 IEEE GRSS Data Fusion Contest LiDAR dataset. The results suggest that the proposed DL architecture is suitable for the task and surpasses other state-of-the-art DL approaches by a large margin.

Knowledge of roof geometry and physical features is essential for evaluation of the impact of multiple rooftop solar photovoltaic (PV) system installations on local electricity networks. The paper starts by listing current methods used and stating their strengths and weaknesses. No current method is capable of delivering accurate results with publicly available input data. Hence a different approach is developed, based on slope and aspect using aircraft-based Light Detection and Ranging (LiDAR) data, building footprint data, GIS (Geographical Information Systems) tools, and aerial photographs. It assesses each roof’s suitability for PV deployment. That is, the characteristics of each roof are examined for fitting of at least a minimum size solar power system. In this way the minimum potential solar yield for region or city may be obtained. Accuracy is determined by ground-truthing against a database of 886 household systems. This is the largest validation of a rooftop assessment method to date. The method is flexible with few prior assumptions. It can generate data for various PV scenarios and future analyses.