Design and Implementation of Computer Vision Based Autonomous Vehicle Prototype

Abstract

In recent years, since automotive makers, IT providers, commercial electronic chip manufacturers have entered a rapid investment race for autonomous vehicles (AVs), they have started to be used in many infrastructures. AVs are automobiles that can drive themselves using the automatic control systems while cruising by sensing the road, traffic flow and surroundings without the need for a driver. AVs can detect objects around them using technologies such as radar, lidar, GPS, audiometry and computer vision. In this study, the design and implementation of a computer vision-based autonomous vehicle prototype is proposed. The developed prototype can perform lane tracking and traffic sign control by processing images obtained from the camera, and can detect objects around it with the distance sensor. In the AV prototype, the Raspberry Pi 3B+ module is used to process the images and control the motors, and the cascade classifier is used to recognize the traffic signs. In the performed tests, traffic signs are recognized in 7 different scenarios and the performances are compared. According to the results, the accuracy rate is 94,8% in the tests performed with only one traffic sign. As a result, the autonomous vehicle prototype developed in this study can successfully recognize traffic signs and move on the determined route in different scenarios.