Applications of Augmented Reality and mobile virtual reality (VR) need mobile pupil tracking within Head-mounted displays (HMD). The ultimate goal is an increase of immersion. However, there are no feasible solutions for this available. In particular, they often come with increased costs for the underlying system, the cameras and its surrounding setup. This thesis provides an implementation of a pupil detection algorithm ported to a mobile consumer platform. Attached to an affordable camera, the system is able to simulate the functionality of a pupil tracker. Using this is a step to a self-built cheap alternative to expensive HMD setups. The algorithm was ported to limited computing constraints of the target platform. A function to perform remotely over network was introduced, sending results and calibration data via Wi-Fi. To further enhance the robustness of the system, parts of the algorithm were modified and new approaches were tested. All results have been extensively evaluated and compared with the non-optimized algorithm. In the discussion, open problems are analyzed for future work.
種類: Project thesis
日付: August 2015