Thesis authored by Karim Malhis. Completed in 2019.
Thesis abstract: The exponential growth in the implementation and investments in contemporary technology of self-driving cars has created an environment consists of multiple systems working together. One of the systems that were recently implemented in self-driving cars development is vision systems technology. The implantation of vision system technology meant to support other systems that are already existed, which set the vision system to have a secondary role or less priority in development compared to other systems such as the radar system.
Developing self-driving cars oriented around the vision system is expected to bring benefits such as reduce complexity and costs in development. It will reduce the number of required equipment and the need to configure them to be compatible with each other and allows for more utilization in the usage of vision systems as it will be the center of the development and have more resources to work with.
This report covers the potential role of a vision system in an AI Formula Car. It explains how vision systems handle and process images to allow useful information to be extracted, what are the requirements for building a vision system and provides an implementation of AI Formula car vision system prototype design which was conducted to simulate self-driving car. In this prototype the vision system is the primary system, which will detect the lane track and provide the car with the required directions to stay within the specified track which allows to generate data to be analyzed and used in system training to improve the efficiency and reliability of the vision system.