Designing autonomous driving algorithm with image processing techniques

This thesis is a synthesis of techniques in the field of autonomous driving and is written to provide a foundation for future research. Every year tens of thousands of people lose their lives in traffic accidents. Although there are many reasons for these accidents, the biggest reason is human error. In this thesis, the autonomous robot created using image processing techniques aims to add a new step to the steps previously taken for the transition to autonomous driving, which is necessary to reduce human errors. It has been studied to gain the software and hardware knowledge required for autonomous driving and to bring innovations to this sector. Lane detection, sign detection and traffic light detection studies were carried out with image processing. Autonomous vehicles need to make very fast decisions to avoid accidents. They published a new study showing that humans need about 390 to 600 milliseconds to perceive and react to hazards on the road. Younger drivers detect hazards almost twice as fast as older drivers. Driverless cars equipped with radar or lidar sensors and a camera system have a reaction time of 150-200 milliseconds. A driverless vehicle reacts much faster than humans. In this study, it is aimed to reach the response times of autonomous vehicles available in the market and to design an autonomous vehicle that reacts earlier. The response time of the autonomous vehicle used in the study is 75-100 milliseconds. A vehicle that reacts twice faster than the autonomous vehicles available in the market is designed.