A novel approach for automatic asphalt pavement crack length estimation

Abstract

Estimation of the actual size of an object from its digital image can help to automate processes and increase productivity. One example of object-size measurement is evaluation of the pavement condition by estimating crack length. As input images taken by different cameras are often inconsistent, a robust method of automatically measuring objects is necessary. Here, the crack size has been estimated by combining the proposed ‘asphalt crack length extractor’ (ACLE) method and an object-size conversion method called ‘pixel to actual size conversion engine’ (PACE). In ACLE, image segmentation is performed to detect the object and extract its pixel-wise length from the image. In PACE, the pixel-wise measurements are converted to metric units. Several experiments were conducted to evaluate the proposed methods. The combined PACE/ACLE module showed an average error of less than 3%. Thus, it is proposed for use in crack length determination and condition assessment of asphalt pavements.