A hybrid human machine system for the detection and management of potholes on asphalt road surfaces

Abstract

Road distresses such as potholes can have a negative economic and social impact. The timeous detection and identification of potholes could expedite the maintenance and repair of potholes. The research team previously investigated and reported on the Visual Surveying Platform, which is a system that automatically detects and geo-tags potholes, with a detection accuracy of approximately 82%. At this level of accuracy, errors consisting largely of false positives could result in repair teams responding to non-existent potholes. In order to incorporate the detection system into the existing workflow of one of the municipalities in the Gauteng area, the detection accuracy needed to be improved. The research team modified the system to include a ?human-in-the-loop? mode of operation, where the detection system performs a more suggestive function. The mobile detection system automatically detects potholes in real-time and presents the detections to an operator for validation. The validated detections are then introduced into the operational workflow of the maintenance and repair teams. The ?human-in-the-loop? system and the operational workflow are described in detail in this paper.