One of the most important tasks for Driver Assistance Systems is improving road safety. With a simple notification to the driver, accidents and injuries may be prevented. Until cars with high level autonomous driving capabilities become widespread, drivers’ task of keeping a keen eye on the surroundings and ensuring safety of the vehicle could be made easier.
That is why it is important for today’s car manufacturers to push their limits developing more advanced features. Built in cameras and ultrasound sensors start to appear in more and more cars nowadays, while LIDAR continues to struggle to overcome its significant disadvantage due to its high cost.
The idea behind my thesis is to explore the concept behind these current technologies. Combining the fields of computer vision and graphics, I plan to implement a soft-ware capable of detecting certain objects in the car’s environment using a camera, and providing real-time visual feedback based on the detection’s results. While it is not possible for me to use a high-end video camera or a screen built directly into the windshield and testing my work in a car, I plan to implement the software keeping this use-case in mind.
This document summarises the two semesters long project. The first consisted mostly of getting familiar with the programming language (C++) and the libraries (OpenGL, OpenCV) that are necessary for implementing certain functionalities, and creating the baseline of the software. In the second semester I expanded its visual recognition capabilities by switching to a more complex detection algorithm and calculating depth in-formation.