Generating lane level road data from vehicle trajectories using Kernel Density Estimation

An informative digital map is a prerequisite for future Intelligent Transportation System (ITS) applications such as lane level navigation systems. Preparing and updating such detailed digital maps by using existing methods such as surveying and image digitization is not practical due to the time and cost involved. We propose a method that statistically mines Global Positioning System (GPS) trajectory data obtained from vehicles to generate a lane level digital map of the road. The proposed method is capable of generating a digital map of the road which contains lane centerlines. The proposed method is independent of the lane width, lane parallelism and can handle lane splits and merges. In addition, a method that can be used to generate a map of lane boundaries is also presented. The proposed methods have been tested using the GPS data collected using vehicles equipped with GPS enabled mobile phones. Results show that the proposed method for lane centerline generation is successful in different road geometries