Computer Vision in Autonomous Cars: Algorithms, Approaches and Applications

Before our very eyes a new era of unmanned ground vehicles (UGVs) is born. Request for autonomous transport arose in the mid-1950s, however, being dependent on technology markets, this industry developed only in the early 2000s, and now the trend is growing rapidly, thereat Gartner predicts the introduction of such vehicles by 2020. This paper explores the state in the branch of computer vision, which has the greatest weight in the advancement of such transport. The aim of this project is to test the available methods of the OpenCV library in projects dedicated to ground transportation. It is also crucial to keep in mind all the recent breakthroughs in computer vision algorithms despite the fact that the used platform and infrastructure for competitive and applied robotics limits the progress in computing means, which is implied in commercial projects. Finally, it is required to provide an ability to adopt practical part in further commercial projects or competitions. The structure of this paper is organized as follows: chapter 1 provides the theoretical base related to the topic of technical vision and machine learning evolution, chapter 2 is dedicated to global practice and includes a detailed analysis of cases related to unmanned ground vehicles and other devices based on computer vision, and chapter 3 reveals two projects involving the OpenCV leading library: the project of the loader's orientation at the logistic warehouse and the autonomous 4WD drive robot for the competitions of increased complexity (the project is not completed).