YOLOv8-Based Road Defect Detection and Tracking Using Video Analysis

Abstract

Road surface defects, including cracks, potholes, and edge breaks, pose serious challenges to transportation safety and infrastructure maintenance. Conventional inspection methods are labour-intensive, time-consuming, and often fail to provide timely assessment of road conditions. This paper presents an automated road defect detection and tracking system based on the YOLOv8 object detection model for efficient road surface monitoring. The model was trained on a road defect dataset containing nine defect categories representing different severity levels of cracks, potholes, and edge breaks. OpenCV was employed for video processing, while a centroid-based tracking algorithm was integrated to assign unique IDs to detected defects across consecutive video frames. The proposed system processes road videos by extracting frames, detecting and classifying defects, tracking identified objects, and generating an annotated output video. Performance evaluation was conducted using manually annotated video frames as ground truth and standard object detection metrics, including Precision, Recall, F1-Score, mAP@0.5, mAP@0.5:0.95, and Frames per Second (FPS). Experimental results demonstrated high detection accuracy, reliable localization, and efficient processing speed, highlighting the effectiveness of the proposed approach for automated road inspection. The developed system reduces dependence on manual surveys and provides a scalable solution that can be extended for intelligent transportation systems and smart road maintenance applications.

Introduction

This research presents a YOLOv8-based Road Defect Detection and Tracking System that automatically identifies, classifies, and tracks road surface defects from video footage. The system addresses the limitations of manual road inspections, which are labor-intensive, time-consuming, and prone to human error, by providing an efficient AI-based solution for intelligent road monitoring.

Motivation

Road defects such as cracks, potholes, and edge breaks can lead to vehicle damage, traffic disruptions, and safety hazards if left unrepaired. While existing deep learning methods can detect defects, many are limited to image-based analysis, support only a few defect categories, or lack continuous object tracking across video frames.

Proposed Methodology

The proposed system processes road videos through the following stages:

1. Video Processing: OpenCV extracts individual frames from road inspection videos.
2. Road Defect Detection: A trained YOLOv8 model detects and classifies road defects into nine categories, covering different severity levels of cracks, potholes, and edge breaks.
3. Object Tracking: A centroid-based tracking algorithm assigns unique IDs to detected defects and tracks them across consecutive frames, reducing duplicate detections.
4. Visualization: The final output video displays bounding boxes, class labels, confidence scores, and tracking IDs for each detected defect.

System Architecture

The architecture consists of:

• Video acquisition and frame extraction.
• YOLOv8-based defect detection and classification.
• Defect verification and centroid-based tracking.
• Annotated video generation with tracking information.

Performance Evaluation

The system was evaluated using manually annotated video frames and standard object detection metrics:

Results

• The model accurately detected and localized multiple road defect categories.
• High Precision and Recall indicate reliable detection with few false positives and missed defects.
• The confusion matrix showed only minor misclassifications between visually similar defects.
• Precision–Recall curves demonstrated strong performance across all nine defect categories.
• The centroid tracker successfully maintained consistent object identities across video frames, enabling continuous monitoring.

Conclusion

This paper presented a YOLOv8-based road defect detection and tracking system for automated road condition monitoring using video analysis. The proposed framework integrates YOLOv8 object detection with a centroid-based tracking algorithm to accurately detect, classify, and track road defects across consecutive video frames. The system was trained to recognize nine categories of road defects, including different severity levels of cracks, potholes, and edge breaks, and was evaluated using manually annotated video frames. Experimental results demonstrated that the proposed approach achieved high detection accuracy and reliable localization, as reflected by strong Precision, Recall, F1-Score, and mAP values. The integration of object tracking enabled consistent identification of detected defects throughout the video sequence, improving the interpretability of the results while minimizing duplicate detections. These findings indicate that the proposed system provides an efficient and practical solution for automated road inspection and infrastructure monitoring. Future work will focus on extending the system to real-time deployment using live camera feeds, incorporating GPS-based defect localization, and optimizing the model for edge devices. Further improvements can also be achieved by training on larger and more diverse datasets to enhance robustness under varying environmental and road conditions.

References

[1] Liu Y, Duan M, Ding G, Ding H, Hu P, Zhao H. HE-YOLOv5s: Efficient Road Defect Detection Network. Entropy (Basel). 2023 Aug 31;25(9):1280. doi: 10.3390/e25091280. PMID: 37761579; PMCID: PMC10527674. [2] Shuangxi Zhou, Dan Yang, Ziyu Zhang, Jinwen Zhang, Fulin Qu, Piyush Punetha, Wengui Li, Ning Li, Enhancing autonomous pavement crack detection: Optimizing YOLOv5s algorithm with advanced deep learning techniques, Measurement, Volume 240, 2025, 115603, ISSN 0263-2241. https://doi.org/10.1016/j.measurement.2024.115603. [3] Li Y, Yin C, Lei Y, Zhang J, Yan Y. RDD-YOLO: Road Damage Detection Algorithm Based on Improved You Only Look Once Version 8. Applied Sciences. 2024; 14(8):3360. https://doi.org/10.3390/app14083360 [4] Jiang, Y. Road damage detection and classification using deep neural networks. Discov Appl Sci 6, 421 (2024). https://doi.org/10.1007/s42452-024-06129-0 [5] Wang, D., Xu, A., Zhu, Y., Yang, D., Xu, Y. (2025). YOLOv8-GRW: A YOLOv8-based Algorithm for Road Defect Detection. Information Technology and Control, 54(1), 290-306. https:// doi.org/10.5755/j01.itc.54.1.3751 [6] Peter Klco, Dusan Koniar, Libor Hargas, Marek Paskala, Automated Detection Of Potholes Using YOLOv5 Neural Network, Transportation Research Procedia, Volume 74, 2023, Pages 1150-1155, ISSN 2352-1465, https://doi.org/10.1016/j.trpro.2023.11.255. [7] Jiang, Y. Road damage detection and classification using deep neural networks. Discov Appl Sci 6, 421 (2024). https://doi.org/10.1007/s42452-024-06129-0 [8] ZhiZun Zeng, RongJie Liu, YongHang Huang, YuJie Li, and BenYing Tan. 2025. GA-YOLOv8: A Road Defect Detection Model Based on Improved YOLOv8. In Proceedings of the 2024 6th International Conference on Video, Signal and Image Processing (VSIP \'24). Association for Computing Machinery, New York, NY, USA, 36–41. https://doi.org/10.1145/3708568.3708574 [9] Wan, F., Sun, C., He, H. et al. YOLO-LRDD: a lightweight method for road damage detection based on improved YOLOv5s. EURASIP J. Adv. Signal Process. 2022, 98 (2022). https://doi.org/10.1186/s13634-022-00931-x [10] Zeng, J., Zhong, H. YOLOv8-PD: an improved road damage detection algorithm based on YOLOv8n model. Sci Rep 14, 12052 (2024). https://doi.org/10.1038/s41598-024-62933-z [11] Zaiyan Zhang, Jiachen Wu, Weidong Song, Yangyang Zhuang, Yanli Xu, Xin Ye, Guoli Shi, Hongyue Zhang, ARDs-YOLO: Intelligent detection of asphalt road damages and evaluation of pavement condition in complex scenarios, Measurement, Volume 242, Part C, 2025, 115946, ISSN 0263-2241, https://doi.org/10.1016/j.measurement.2024.115946. [12] Ma, L., Chen, M. Road damage detection based on improved YOLO algorithm. Sci Rep 15, 28506 (2025). https://doi.org/10.1038/s41598-025-14461-7

Copyright

Copyright © 2026 M P Princy Veronica, N Naveen Kumar. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.