Registration and Tracking of Objects with Computer Vision for Autonomous Vehicles

Abstract

For this research, an autonomous leader-follower vehicle caravan is being designed using computer vision. The focus of this research is to investigate computer vision methods of identifying objects detected by the leader vehicle and determining the follower vehicle' s position relative to the objects. Many object registration and tracking methods are presented in this paper such as Hough transforms, optical flow, and normalized cross correlation. A novel statistical-gradient method is presented to combine feature based methods and area based methods for object registration and tracking. These methods are evaluated against both synthetic and captured video data. This registration method is shown to be effective with both synthetic data and acquired data.