Monocular Video serves RADAR-based Emergency Braking

Much work was carried out recently for emergency braking based on radar signals. The key step for emergency braking is the reliable detection of obstacles. Moving objects are verified as such by tracking the radar signal. However, discarding so-called phantom objects remains a challenge for stationary objects. This leads to the question of sensor fusion for more reliable verification of obstacles. In this paper we propose a novel method using a monocular camera, such as the night view camera in the Mercedes S class. Our two goals in this paper are the verification of obstacles and the detection of obstacle boundaries. This allows analysis of the situation for carrying out emergency braking. The verification of obstacles is done by analyzing the scaling of obstacles as they get closer to the camera. The perspective image motion of the ground plane serves as a counter hypothesis to detect phantom objects. Obstacle boundaries are found by graph cut segmentation on these two motion fields.