Image landmark based positioning in road safety applications using high accurate maps

Abstract

An accurate positioning is a key requirement of an advanced driver assistance system (ADAS). For cooperative systems using car-to-car and car-to-infrastructure communication, an exact lane precise position estimation becomes essential. We present a system which improves a GPS/INS (inertial navigation system) position estimate using image landmarks. This system allows the tracking of multiple lane markings belonging together. For the edge detection in the image a local orientation coding (LOC) operator is used. By means of two different subsequent Hough transforms the detections of this operator are evaluated with respect to their shape. The results of those Hough transforms are tracked with a set of Kalman Filters modelling the lanes, where one filter represents one lane bounding. By means of the most significant track the GPS/INS position estimation is corrected matching the lane marking track to the road elements stored in a digital map.