An Automotive Perception System using Distance Cameras and Polar Occupancy Grids

Abstract: One of the basic tasks of automotive collision avoidance and collision mitigation systems is the robust and reliable detection of objects as well as the prediction of future trajectories. This paper presents a system which uses a distance camera as range sensor to solve this task. The underlying algorithms uses an innovative Polar occupancy grid, which supports the segmentation of the distance image of the range sensor. An efficient sampling approach for the ego-motion compensation of the Polar occupancy grid is shown. Furthermore, we present a tracking system based on Unscented Kalman filters, which uses the segmented measurements of the distance image. Results of practical tests of the system are presented for two use cases: A front view application and a blind spot observation application.

• Type: Journal
• Year: 2009
• Where: GI
• Authors: Norman Mattern, Robin Schubert, Christian Adam, Gerd Wanielik
• Keywords: Distance Image, GI 2009, Polar Occupancy, Range Sensor, Theoretical Computer Science