Pothole detection and tracking in asphalt pavement using monocular depth estimation

Changhwan Jang; Byeong-Hun Woo; Jin-Young Seok; Hong Gi Kim

doi:10.22702/jkai.2025.15.2.24

All Issue

2025 Vol.15, Issue 2 Preview Page Next Page

Scientific Paper

Pothole detection and tracking in asphalt pavement using monocular depth estimation 단안깊이추정을 활용한 아스팔트 포장의 포트홀 감지 및 추적

31 December 2025. pp. 284-292

PDF XML

Abstract

Recent advances in artificial intelligence (AI) have enabled rapid progress in image-based object detection, with applications extending into infrastructure monitoring. This study investigates the feasibility of applying Monocular Depth Estimation (MDE) to detect and track potholes in asphalt pavement. Unlike traditional approaches that require absolute depth data and extensive training, MDE provides relative depth information directly from single images, offering a lightweight and efficient alternative. Using the Zoe.D library, one of the most advanced MDE modules, we evaluated pothole detection performance on publicly available datasets. Experimental results demonstrate that MDE can successfully highlight pothole regions under favorable conditions, particularly when images are captured at close range with minimal background noise. However, failures occurred when distant objects such as trees, vehicles, or fences dominated the depth perception, or when reflective surfaces like water introduced noise. In some cases, post-processing adjustments improved the visibility of pothole contours, suggesting that MDE can complement conventional object detection models in multimodal frameworks. Overall, while MDE alone is insufficient for robust pothole detection, its integration with existing detection architectures shows promise for cost-effective, non-intrusive pavement monitoring. This research highlights both the limitations and potential of MDE in construction engineering applications.

Keywords

Monocular depth estimation

Pothole

Pavement

Detection

Machine learning

References

서민수･홍승혜･이정명 (2018). “반자율주행 맥락에서 AI 에이전트의 멀티모달 인터랙션이 운전자 경험에 미치는 효과: 시각적 캐릭터 유무를 중심으로: 시각적 캐릭터 유무를 중심으로”, 한국콘텐츠학회논문지, 18(8), pp. 92-101.

10.5392/JKCA.2018.18.08.092

엄태영･배기덕･최영호･조성훈･문용선 (2020). “멀티모달 센서모듈 기반 지능형 로봇을 이용한 실외무인경비 플랫폼 및 시스템에 관한 연구”, 한국통신학회지(정보와통신), 38(1), pp. 27-33.

이경배･김종호･유준상･송재경･조근철･손덕수 (2024). “U-net 기반 균열 자동검출을 위한 라인스캔 노면영상의 처리기법 개발”, 한국도로학회 학술대회 발표논문 초록집, 한국도로학회, pp. 251-252.

한유라･신복균 (2023). “드론-로봇 연계 도심지 고중량 화물 멀티모달 배송기술 개발”, 항공우주시스템공학회 학술행사 논문집, 항공우주시스템공학회, pp. 148-149.

Afshar, M.F., Shirmohammadi, Z., Ghahramani, S.A.A.G., Noorparvar, A. and Hemmatyar, A.M.A. (2023). “An efficient approach to monocular depth estimation for autonomous vehicle perception systems”, Sustainability, 15(11), 8897.

10.3390/su15118897

Bhat, S.F., Birkl, R., Wofk, D., Wonka, P. and Müller, M. (2023). ZoeDepth [Computer software]. GitHub. https://github.com/isl-org/ZoeDepth

Chaiyasarn, K., Buatik, A., Mohamad, H., Zhou, M., Kongsilp, S. and Poovarodom, N. (2022). “Integrated pixel-level CNN-FCN crack detection via photogrammetric 3D texture mapping of concrete structures”, Automation in Construction, 140, 104388.

10.1016/j.autcon.2022.104388

Chen, X., Yu, Y., Wang, Y. and Hu, Z. Z. (2025). “Multimodal data fusion for ergonomic assessment of construction workers in visually obstructed environments”, Automation in Construction, 179, 106495.

10.1016/j.autcon.2025.106495

Dow, H., Perry, M., Pennada, S., Lunn, R. and Pytharouli, S. (2024). “3D reconstruction and measurement of concrete spalling using near-field Photometric stereo and YOLOv8”, Automation in Construction, 166, 105633.

10.1016/j.autcon.2024.105633

Howard, A.G., Zhu, M., Chen, B., Kalenichenko, D., Wang, W., Weyand, T. Andreetto, M. and Adam, H. (2017). “Mobilenets: Efficient convolutional neural networks for mobile vision applications”, arXiv preprint arXiv:1704.04861.

Huang, Z., Ali, M.A., Nukman, Y., Xu, H.L., Zhang, S., Chen, H. and Alkhedher, M. (2025). “A Systematic Review of Monocular Depth Estimation for Autonomous Driving: Methods and Dataset Benchmarking”, Results in Engineering, 26, 105359.

10.1016/j.rineng.2025.105359

Ibrahem, H., Salem, A. and Kang, H.S. (2022). “DTS-depth: Real-time single-image depth estimation using depth-to-space image construction”, Sensors, 22(5), 1914.

10.3390/s2205191435271061PMC8914965

Li, J., Song, X., Gao, R. and Tao, D. (2023). “Monocular Depth Estimation for 3D Map Construction at Underground Parking Structures”, Electronics, 12(11), 2390.

10.3390/electronics12112390

Li, S. and Zhao, X. (2019). “Image‐based concrete crack detection using convolutional neural network and exhaustive search technique”, Advances in civil engineering, 2019(1), 6520620.

10.1155/2019/6520620

Liu, Y., Alias, A.H.B., Haron, N.A., Bakar, N.A. and Wang, H. (2024). “Exploring three pillars of construction robotics via dual-track quantitative analysis”, Automation in Construction, 162, 105391.

10.1016/j.autcon.2024.105391

Roboflow (2024, June). Pothole dataset [Open source dataset]. Roboflow Universe. Retrieved November 27, 2025, from https://universe.roboflow.com/pothole-detector-dmdod/pothole-clzln

Swain, S. and Tripathy, A.K. (2024). “Automatic detection of potholes using VGG-16 pre-trained network and Convolutional Neural Network”, Heliyon, 10(10), e30957.

10.1016/j.heliyon.2024.e3095738803954PMC11128863

Yu, J. and Choi, H. (2021). “YOLO MDE: Object detection with monocular depth estimation”, Electronics, 11(1), 76.

10.3390/electronics11010076

Information

Publisher :KOREAN ASPHALT INSTITUTE
Publisher(Ko) :한국아스팔트학회
Journal Title :Journal of the Korean Asphalt Institute
Journal Title(Ko) :한국아스팔트학회지
Volume : 15
No :2
Pages :284-292
Received Date : 2025-12-05
Revised Date : 2025-12-19
Accepted Date : 2025-12-19
DOI :https://doi.org/10.22702/jkai.2025.15.2.24

Journal of the Korean Asphalt Institute ISSN:2234-0785(Print) 2635-9553(Online) 한국아스팔트학회지

All Issue