Object Tracking for ‘Car Platooning’ Using a Single Area-Scan Camera
Fidelis Theinert, Tony Covena, Joris Lamers, Daan van Oppenraaij · 2018
Depth determination and object tracking with computer vision can be difficult tasks. A reliable setup is often costly, requiring multiple cameras and advanced processing algorithms. However, full depth determination is not always required. For certain applications the tracking of a simple object is sufficient. In this paper a method is proposed for retrieving steering information by using a single low-cost camera. This setup can be used in an embedded system to track an object in 3D space to realize `car platooning'. A test setup was made to develop the required algorithms and to prove the concept. Once the test setup proved successful, the concept was tested in a real environment with multiple vehicles to achieve follow-the-leader behavior. The 3D location of the object is used to determine the distance and the angle to the vehicle in front. This way it is possible to achieve autonomous following behavior with a single area-scan camera. The proposed method can be useful for applications where budget, space or processing power is limited. It can be a very handy solution for projects considering computer vision for following and tracking.