RoadShield AI: An Intelligent Road Damage Detection and Maintenance Prioritization System Using YOLOv8

Author

Arul Deepa J, Mrs. P. Jasmine Lois Ebenezer

Abstract

  Road damage such as potholes, cracks, and surface erosion can reduce road safety and affect the quality of transportation infrastructure. Traditional road inspection methods generally depend on manual observation, which can require considerable time and human effort. To address these limitations, this paper presents RoadShield AI, an intelligent road damage detection and maintenance prioritization system based on computer vision and deep learning. The proposed system uses the YOLOv8 object detection model to identify three major categories of road damage: potholes, cracks, and surface erosion. For every detected damage region, the system extracts the damage type, confidence score, bounding-box coordinates, width, height, and approximate damaged area. Based on the estimated damaged area, the detected damage is categorized into Low, Medium, or High severity. The severity level is then converted into a corresponding repair priority: Normal, High, or Urgent. A dashboard is provided to present the detection results, severity levels, repair priorities, and overall maintenance information in an understandable form. Unlike conventional road damage detection approaches that mainly concentrate on identifying defects, RoadShield AI extends the analysis toward severity assessment and maintenance prioritization. The proposed approach can therefore support road inspection and help maintenance teams identify road sections that require greater attention.

Keywords

Road Damage Detection, YOLOv8, Computer Vision, Pothole Detection, Crack Detection, Surface Erosion, Severity Estimation, Repair Prioritization, Deep Learning, Maintenance Dashboard

Full Text:

Download Paper PDF

References

[1] Enhanced Pothole Detection System Using YOLOX Algorithm.

[2] Crowdsensing-based Road Damage Detection Challenge (CRDDC-2022).

[3] OBC-YOLOv8: An Improved Road Damage Detection Model Based on YOLOv8.

[4] RDD2022: A Multi-National Image Dataset for Automatic Road Damage Detection.

[5] J. Redmon et al., You Only Look Once: Unified, Real-Time Object Detection.

[6] Ultralytics, YOLOv8 Documentation and Object Detection Framework.

Share your valuable work from Social Media Buttons