Improving of GPS localization by 3D vision

点击翻译按钮获取中文摘要

GPS-based positioning is often poor near high buildings or interferences such as some electronic devices. This research aims to enhance GPS position accuracy by analyzing a video captured from a camera bound to the GPS sensor. This research is particularly useful for unmanned aerial vehicles (UAVs), which mainly rely on GPS for positioning and also often equipped with digital cameras. 3D computer vision algorithms can recover the 3D motion of a video camera from its recorded video frames. These algorithms provide a position measurement independent of that from GPS. By fusing these two position measurements, we can improve the accuracy and coherency of GPS readings. When a satellite image is available, computer vision algorithms allow us to register the motion trajectory to the satellite map to further enhance position reliability. This research will be carried out with MDA, a local company dedicated in UAV applications, to strengthen their products and services.