GCD-L: A Novel Method for Geometric Change Detection in HD Maps Using Low-Cost Sensors

Abstract

Updating high-definition maps is imperative for the safety of autonomous vehicles. However, positional changes in lane lines are hard to be detected in a timely manner due to a limited number of expensive surveying vehicles over a large geographic area. Herein, a novel method is proposed to detect the geometric changes of lane lines using low-cost sensors, such as consumer-grade global navigation satellite system (GNSS) hardware receivers and cameras. The proposed framework geometric change detection using low-cost sensors (GCD-L) and algorithm change segment compare (CSC), which are based on the lane width between the curb line and the adjacent leftmost lane line, can perceive the positional changes of the leftmost lane line on highway and expressway roads. The effectiveness of the proposed method is verified by evaluating it on a real-world typical urban ring road dataset. The experimental results show that 71% detected change segments are valid with only two round crowdsourced maps.