Automated 3D Quantification of Concrete Pavement Joint Damage

Abstract

Joint damage, including raveling and spalling, poses substantial challenges to the safety and long-term durability of concrete pavements. In response, accurate assessment and quantification of joint damage play a vital role in upholding the quality of construction joints and facilitating timely maintenance. This approach ultimately leads to an extended service life of concrete pavements. In this study, we propose an automated 3D joint damage quantification algorithm (DQA) that utilizes a smartphone camera, 3D point cloud reconstruction, and deep learning (DL) segmentation. The DQA algorithm utilizes a smartphone camera to capture 2D images of the joint. These images are then processed by a trained DL model to detect and color-mask the damaged areas, providing a visual representation of the joint damage. Afterward, a 3D reconstruction of the damaged joint is generated using a series of these 2D images. To ensure accurate quantification of the damage, a color thresholding criterion is applied, which enables precise measurement of the extent of the damage. To validate the effectiveness of the DQA framework, we conducted experiments on contraction joints from four different projects. The results demonstrate that the DQA achieves a joint damage index accuracy of 76% with a 10% error margin, indicating its reliability and robustness in accurately quantifying joint damage.