Weighted Neighborhood Pixels Segmentation Method for Automated Detection of Cracks on Pavement Surface Images

Abstract

A new method is designed to detect and segment a crack on a pavement surface image from its background. Gray images of pavement surface have been collected from asphalt concrete pavement on an interstate highway in Maryland using charge-coupled device (CCD) digital cameras. These pavement surface images contain different types of pavement surface distresses and pavement markings. The first step of the new algorithm is preparing a uniform background by applying the new average brightness level of each column. The weighted neighborhood pixels method is proposed, which is based on the intensities of all pixels in three surrounding loops. Seven different patterns are studied and compared, which leads to the best performance eight-direction pattern in terms of accuracy and robustness for feature extraction of pavement images with cracking; then, a local threshold approach and shape filtering using eccentricity value parameters are applied to enhance the candidate cracks. Finally, crack fragments are connected by using a dilation operator. The performance of the new method is evaluated against the ground truth data using manual detection and segmentation. The results show that the developed automated detection and segmentation method is accurate, fast, robust, and suitable for online pavement condition assessment.