Abstract
Crack percentage is established as an index for quantitatively assessing asphalt pavement crack damage. Currently, calculating the crack percentage involves sketching cracks on the road surface and then counting the number of cracks within a partition, and this manual labor further requires significant amounts of work and time and does not collect important information such as crack opening width. This study combines machine learning with decision tree and image analysis to establish a method for automatically detecting cracks from digital images. We verified the high crack detection performance of the developed method by analyzing the images of dense graded asphalt and porous asphalt pavement.
