Have you driven an automobile that has an active Lane Departure Warning (LDW) system? These vehicles scan the roadway for lane markings and provide feedback to the driver when they detect the vehicle crossing into a different lane. For such systems to work effectively, the vehicles must have a highly reliable lane marking detection package. This concept needs great on-vehicle technology and high-quality roadway lane markings on the road.
State and local transportation agencies maintain these lane markings and, yes, there are many, many miles of them. Keeping track of the quality of the striping is a daunting task. Can a system similar to LDW be used to evaluate the quality of lane markings and provide information to agencies about poor lane markings?
This project developed a proof-of-concept for lane marking assessment using a combination of machine vision (MV) algorithms and artificial intelligence (AI). We collected lane marking images from the existing third-party archives and using Go-Pro mounted on agency vehicles. We then trained an AI to detect lane marking locations and MV to identify deterioration in lane marking.
The proposed proof-of-concept can supplement the existing manual survey-based framework for identifying deteriorating lane markings. Detecting deteriorating lane marking faster, at lower cost, and over a wider road network. This can improve safety by allowing for timely detection and repairs of deteriorated marking at locations where markings are critical for successfully navigating the roadways.
