Development of a Crack Type Index

Abstract

Many automated systems for crack analysis have been developed to measure the extent and severity of pavement cracking objectively. However, the accuracy of such an automated crack analysis system has not been satisfactory. This paper presents a crack type index (CTI) that can be easily adopted to determine the crack type objectively as longitudinal, transverse, and alligator cracking. The CTI is based on the spatial distribution of the image tiles rather than image pixels, where a tile is defined as a subimage of a whole digital image. The spatial distribution of image tiles is analyzed vertically and horizontally, with a resulting single index, which can be used to identify a spatial orientation of cracking. To determine the accurate CTI threshold values for longitudinal, transverse, and alligator cracks, 150 pavement images were captured with a digital video camera mounted on a sport-utility vehicle: 50 images for each of three types of cracking. These 150 images were analyzed automatically to compute the CTI values that correlate with crack types. To validate the CTI system, another 150 pavement images were captured. The CTI system identified 150 images as proper crack types with an 86% accuracy for alligator cracking, 92% accuracy for transverse cracking, and 94% accuracy for longitudinal cracking. The CTI system is further validated against images of block cracking and multiple cracks. The validation result against block cracking and multiple cracks indicates that the proposed CTI system in conjunction with UCI is robust and can be extended to identify block cracking and multiple cracks. The CTI method can be used to determine crack types from the digital images automatically without any human intervention.