Highway Origin-Destination Vehicle Weaving Analysis Using Field Camera Videos

Highway vehicle weaving analysis aims to determine the percentage and number of vehicles changing lanes from origin to destination on a highway segment. Vehicle weaving often slows traffic, causes jams, and increases safety concerns. Quantitative lane-based highway weaving patterns analysis is crucial for traffic management and road improvement. However, there is currently no cost-effective tool to obtain lane-based vehicle weaving data. This paper describes a method combining machine learning with convenient user verification to achieve accurate weaving analysis using field camera videos. The effectiveness of this approach has been demonstrated in experimental results, and a software tool has been deployed to the Indiana Department of Transportation.