Development of AutoMated Image CoLLection AND ANALYSIS Systems For Infrastructure asset management

Hosin “David” Lee[1] and J. Kim[2]

Automated pavement image collection systems have been developed over the years to evaluate the condition of pavements.  The simplest method is to have a person ride or walk along the road and visually evaluate it.  However, there are a number of problems with this approach, including the boredom attendant on such a mundane task and the subjective and arbitrary nature of the evaluations due to human nature.  Because visual evaluations of pavements are by nature subjective and arbitrary, automated systems have been developed in an attempt to make the evaluation procedure more objective and consistent.  The main objectives of the research are to develop: 1) an automated image collection system (AICS), 2) a manual image analysis system (MIAS), and 3) an automated image analysis system (AIAS).  The MIAS was designed to measure cracks by processing digital images interactively through a computer screen in the most efficient manner. A robust tile-based AIAS was developed to determine the crack type based on Unified Crack Index (UCI) and Crack Type Index (CTI). When there is a significant amount of noises in the pavement image, a pixel-based approach would produce very unreliable results. The tile-based AIAS approach significantly reduces computational complexity over pixel-based computation. The tile-based AIAS based on UCI and CTI is very stable because isolated crack pixels will be ignored as background noises. To further improve the accuracy of the AIAS, a variable optimum threshold level was implemented for classifying a tile as a cracked tile or not.