Solar photovoltaic (PV) technology is increasingly popular in metropolitan cities. From 2018 to 2019, the solar energy produced in Hong Kong grew from 47 TJ to 74 TJ, with a 57% increase. To enhance the efficiency of PV equipment deployments, this study aims to determine the solar energy potential of cities by using geographic information system (GIS), remote sensing (RS) and computer vision technology. In addition, in order to obtain the solar reflectance of the façade of buildings in Hong Kong, we obtained a large number of street view images in the urban area of Hong Kong, identified the material types in the street view images by semantic segmentation, and finally projected them into the three-dimensional city model to estimate the reflection of the solar on the city envelopes.
Machine learning-based estimation of solar potential on three-dimensional urban envelopes