Using the shadow as a single feature for real-time monocular vehicle pose determination

Abstract

In this work we propose a way to detect vehicles in monocular camera images and determine their position and orientation on the ground plane relative to the camera. The camera does not need to be stationary which allows the method to be used in mobile applications. Its results can therefore serve as an input to advanced driver assistance systems (ADAS). The single feature used is the shadow beneath the vehicles. We implemented a real-time applicable method to detect these shadows under strongly varying conditions and determine the corresponding vehicle pose. Finally we evaluate our results by comparing them to ground truth data. © 2011 Springer-Verlag.