Sky View Factor Estimation Based on Deep Learning and Big Data

In urban climate studies, the radiative exchange component is essential, hence urban geometry has been investigated in a view factor framework for decades. The sky view factor (SVF) is a commonly used 3-d indicator and represents the fraction of the overlying hemisphere occupied by the sky. It can be calculated by hemispheric pictures taken by cameras with fish-eye lenses. In this thesis a new code has been developed to define SVF based on 360-degree street view through produced the coordinate processing of the panoramic image which is an effect that can be comparable to the output obtained from traditional fish-eye camera. At the same time, deep learning is used to segment the panoramic image to automatically select SVF and other objects to overlap resulting solar paths for the given location and analyze solar radiation hours. Each step is defined and verified in this thesis. Additionally, initial sample application results are shown.