Road boundary detection for run-off road prevention based on the fusion of video and radar

An approach for detecting the road boundary on different types of roads without any preliminary knowledge is presented. We fuse information obtained from an algorithm which detects road markings and road edges in images acquired by a video camera as well as data from a radar sensor. Each road marking, each road edge and each road barrier is tracked individually. Hence we can even capture exits or laybys. We use an edge image for road marking detection and texture information for road edge detection. Additional data provided by a radar sensor is used to measure targets referring to static barriers along the road side such as guardrails. The output of each processing unit is fused into a Kalman filter framework, where the confidence of each subsystem influences the innovation of the overall system. The underlying geometric road model comprises parameters for multiple lanes, the flanking road edge as well as the vehicle's relative pose. The work is part of the project Interactive.