| Popis: |
Potholes in flexible asphalt pavement systems are one of the major distresses for fatal accidents. Ingress of water through the pothole disturbs the integrity of the pavement system. Delayed maintenance of potholes will adversely affect safety of road users and health of the road pavements. Therefore, detection, quantification, and maintenance of potholes are three indispensable tasks in pavement asset management. Manual collection of pothole data is time-consuming and laborious. Hence, the use of cutting-edge artificial intelligence techniques has become popular in the recent times. The major objective of this study was to develop a framework for pothole detection, quantification, and maintenance system (PDQMS) to detect and quantify potholes using pavement images collected by an automated survey vehicle; the system was also incorporated with a mechanism that calculates the amount of patching material required for maintenance. The state-of-the-art multiple-object detection algorithm, You Only Look Once version 3 (YOLOv3) was selected to detect potholes from the images. One of the salient characteristic features of the PDQMS developed in this study was to use severity-based pothole classification approach, a first-of-its-kind novel framework, which helped group the pavement sections based on severity of potholes for maintenance operations. The proposed framework is envisioned to assist the agencies in making decisions to patch potholes and reduce fatal accidents, if not maintained. |
