Photovoltaic (PV) technology is one of two modes of energy generation which utilize solar energy as its source. The rooftops of buildings can be utilized for solar power generation using this technology, and are considered to be highly promising sites for urban PV installations due to land space limitations. However, to properly plan such installations decision makers would need to have detailed information about the amount of rooftop area that is available, as well as the distribution of individual rooftop sizes. In this paper a machine learning based approach for detecting rooftops is proposed and its utility for planning rooftop PV installations is demonstrated via a simple pilot study on two different residential areas in Abu Dhabi, UAE. The proposed method uses a two-stage classification model to estimate the rooftop area that is available for solar panel installation. Next, a comparative study of three different types of PV technologies is conducted in terms of energy generation and economic viability. The results obtained from these experiments suggest that thin-film panels may have a distinct advantage over other PV technologies. Even though the cost of PV panels is still quite high, this could be balanced by the potential benefits to the environment. If reasonable subsidies and feed-in tariffs are implemented, PV technology can become a cost effective option for the UAE.