Final project for ROB501: Computer vision for robotics.
Developed a modularized Python tool to generate orthographic maps (bird’s eye views) using omnidirectional stereo imagery collected by a rover at ground level on a simulated Martian terrain. Tested several methods for generating a realistic composite with minimal visual artifacts from multiple rover runs and cameras. Made use of data from the Canadian Planetary Emulation Terrain Energy-Aware Rover Navigation Dataset.
A high-resolution orthographic map produced using all available data and tuned parameters is shown below.