Robust FPGA-centric CNN visual localisation for GPS-denied UAVs

This project targets real time, on board, absolute localisation from a single image on low power embedded hardware. A convolutional neural network (CNN) maps a monocular frame directly to coordinates in a pre surveyed reference image, avoiding continuous feature tracking and cloud offload. The method is intended to complement inertial devices such as an inertial measurement unit (IMU) and a barometer by providing periodic absolute fixes that bound drift in an inertial navigation system. The goal is to enable reliable autonomy when satellite navigation cannot be trusted, while keeping latency and energy use within the strict budgets of small UAV platforms.