Pilot of Biometric Vehicle Occupancy Detection Technology on Managed Lanes

Abstract

Commuter assistance programs have relied on the honor system when commuters log carpool trips to earn rewards for choosing to travel using high-occupancy vehicle (HOV) modes. Where HOV/high-occupancy toll lane incentivization policies presently exist, levels of HOV misdeclaration can reach 50% or higher. In coordination with the technology developer, evaluators piloted a biometric-based vehicle occupancy detection mobile application. Researchers were assisted by the participation of volunteer carpoolers recruited from commuter assistance registrant databases. Volunteers downloaded the mobile app on their personal devices. They logged carpool trips on I-95 in South Florida, I-275 in Tampa, Florida, and I-15 in Salt Lake City, Utah. Carpoolers used the app to record their facial image data for two occupants at the beginning of the carpool trip, then again at the end of the trip. The app calculated scores for facial realness and similarity. Evaluators reviewed the carpool data log for the app’s determination of True Positives (carpool with two occupants) and True Negatives (single-occupant vehicle trip). These were compared with photos recorded by the carpoolers concurrently with the logged trips. There were a total of 837 logged trips. There was one false positive among trips validated by the app and 13 false negatives among trips invalidated by the app. The app is configurable, thus requiring a policy decision for selecting the thresholds for realness and similarity. These represent the balance between reducing the seconds required for trip validation, thus increasing user convenience, versus heightened accuracy.