Vision-guided robot for automated pixel-level pavement crack sealing
Document Type
Article
Publication Date
12-1-2024
Department
Department of Civil, Environmental, and Geospatial Engineering
Abstract
Automated pavement crack sealing plays a crucial role in road maintenance. However, challenges remain in refining crack segmentation and sealing control accuracy. This article proposes an automated pavement crack sealing robot(APCSbot), which employs a crack refinement network (CrackSegRefiner) and a crack sealing controller(LQR). Specifically, CrackSegRefiner is based on a denoising diffusion model to refine the coarse mask of crack through a diffusion process. Additionally, the LQR controller integrates weight matrices Q and R to ensure control and state of APCSbot based on visual servo, facilitating the delivery of emulsified asphalt for sealing through the end effector. Extensive experiments conducted on the DeepCrack, CFD, and S2TCrack datasets confirm the effectiveness of APCSbot, which achieved a segmentation precision of 84.48% and mIoU of 79.28%. Furthermore, the system demonstrated a sealing error of 6.22 mm and speed of 0.0456 m/s when addressing discontinuous cracks, showcasing its excellence and robustness in crack sealing.
Publication Title
Automation in Construction
Recommended Citation
Zhang, J.,
Yang, X.,
Wang, W.,
Wang, H.,
Ding, L.,
El-Badawy, S.,
&
You, Z.
(2024).
Vision-guided robot for automated pixel-level pavement crack sealing.
Automation in Construction,
168.
http://doi.org/10.1016/j.autcon.2024.105783
Retrieved from: https://digitalcommons.mtu.edu/michigantech-p2/1073