Patch distress detection in asphalt pavement images

Abstract

Pavement condition assessment is an indispensable constituent in maintaining roadway infrastructure. Current practices for pavement condition assessment are labor intensive which introduce subjectivity in pavement rating and are time consuming. Automated methods rely on full 3D reconstruction of the pavement surface which introduces high equipment and computation costs. This paper presents a methodology for detection of patch type distresses in asphalt pavement images. This method uses filtering and histogram equalization for clearing and enhancing the image accordingly. It continues by applying the morphological process of closing, along with some rules based on the characteristics of a patch when captured in an image, to finally detect the area it occupies. Criteria for area, length, and width of a patch as seen in an image are taken into account to decide whether a probable patch is actually a patch. The method has been implemented in C#. The preliminary experiments demonstrate it produces promising recognition results.