In the International Aerial Robotics Competition: Mission 8 the goal of the vision team was to take images of four separate pieces of a single QR code, isolate then combine them and finally read the QR code as a whole. In order to accomplish this, we based our approach on a paper, “fast detection and recognition of QR codes”. Based on this paper, I implemented a linear algebra program called parallel coordinates which would identify the borders of the QR code as being the points in parralel coordinate space with the most line intersections. This program was written in OpenGL and runs on the GPU in order to process each image in less than a second, whereas cpu bound python code took about 10 seconds per image. The output of this was passed back to python where we would crop the images accordingly and permute the isolated pieces until the QR code as a whole could be read, thus solving the competition.

Mission 8 Vision Code.