Fused Raised Pavement Marker Detection Using 2D-Lidar and Mono Camera

In this paper, we describe a raised pavement markers detection algorithm using a monocular camera and a 2d lidar that supports automated driving in urban and highways scenes. For each sensor, we present a detection methode based on its sensing properties and range. The results of these approaches are tracked with a multi Kalman filter and are represented as mixtures of Gaussians. Afterwards, we use a Random Sample Consensus (RANSAC) spline fitting algorithm for a robust estimation of the road boundary. To improve the precision of the applied RANSAC we use the detected painted lane markings. Eventually, we fuse the tracked features with the lane markings on the basis of a product approximation method for Gaussian Mixtures. Experimental results show the reliability and the realtime capability of our approach based on the data we collected in California and Nevada.