Automated Detection and Classification of Airport Pavement Cracks
TL;DR
A machine-learning algorithm is introduced to automate the identification and classification of airport pavement cracks that enhances the Tennessee Department of Transportation's (TDOT) ability to monitor pavement conditions efficiently, reducing inspection times and supporting more timely, cost-effective maintenance planning.
Abstract
This research introduces a machine-learning algorithm to automate the identification and classification of airport pavement cracks. Utilizing drones equipped with high-definition cameras, the study implements a deep learning neural network based on the U-net framework, along with a specialized module for precise quantitative analysis of crack length and width. The algorithm categorizes cracks into distinct types and incorporates a color-coding system for quick identification. To overcome the lack of publicly available datasets on airport pavements, we conducted field investigations at several General Aviation Airports in Tennessee to develop a new dataset, CrackAirport, which is now accessible on IEEE DataPort. The proposed algorithm effectively distinguishes cracks from similar non-crack features at the pixel level, ensuring high accuracy in complex environments. This advancement enhances the Tennessee Department of Transportation's (TDOT) ability to monitor pavement conditions efficiently, reducing inspection times and supporting more timely, cost-effective maintenance planning.