Marcus Mattsson

Oh boy, this is going to get geeky. Today, me and my group finished our 3rd semester project after three months of very hard work. The semester theme was Human Senses – Digital Perception, and was mainly focused on using computer vision. We ended up creating a cyclist protection system, which can detect a cyclist in a right turn situation and warn the driver. The system is implemented in C++ using Intel’s computer vision library, OpenCV, for various image processing functions. Basically, the system, called Right Vision, estimates what is called optical flow (the apparent motion of objects, surfaces and edges) and performs segmentation based on objects’ and the background’s vanishing points to detect a cyclist from a dynamic video feed mounted on a moving car.

Estimating right-to-left movements.

This semester at Medialogy was the first in which I really felt that I learned something useful. I have always wanted to learn C++ and learning how to implement various image processing techniques such as thresholding, edge detection and compound operations at the same time was very effective.

The entire project ended with a 97 page report, which can be downloaded here.