Thesis (M.Sc.) - Cairo University - Faculty of Engineering - Department of Computer Engineering

Recently, a huge attention is given to vision-based solutions in a wide range of applications including robotics, smart driving among others. One category of such solutions is visual odometry (VO), which is the process of estimating the complete body/camera pose using the images captured by the camera. In this thesis, previous e orts exerted in VO and its tuning are reviewed and criticized qualitatively. Moreover, a generic framework for testing, evaluating, and rapid prototyping of such systems is provided. Consequently, this framework is used to analyze the impact VO parameters on the pose estimation performance both from the accuracy point of view as well as the time complexity. Finally, a novel method for enhancing the performance of monocular VO is proposed with the intent of reaching a non-aided vision-based solution

Issued also as CD