Nowadays a big variety of methods and applications are being designed with the aim to assist robotics in autonomously performing tasks. One kind of these robots are the unmanned aerial vehicles or drones, which recently have attracted the interest of the research community. Drones can perform various applications such as reconnaissance, surveillance, search and rescue and infrastructure inspection. A core task in these applications, which is of great importance to be performed accurately, concerns the visual detection, recognition and tracking of specific desired objects. In this work, our aim is to develop a system that makes drones automatically perform surveillance and efficiently detect and track objects using their own camera in real world conditions. The results from the performance evaluation were very encouraging.