esc

Type at least two characters to start searching.

↑↓ select↵ openesc close

Share

Export Citation

APA
MLA
Chicago
Harvard
Vancouver
BIBTEX
RIS
Universitas Hasanuddin
Research output:Contribution to journal›Article›peer-review

Multi-Altitude UAV-Based Pothole Area Estimation Using Dynamic GSD Correction

Khafifah M.

Proceedings 2026 International Conference on Informationtechnology and Computing Icitcom 2026

Published: 2026

Abstract

Variations in UAV flight altitude change image scale and can cause inconsistent physical area estimates for the same pothole. This study evaluates pothole area estimation by comparing a fixed ground sampling distance (GSD) approach with a dynamic GSD approach computed from flight altitude and camera geometry. YOLOv11m-seg was used to generate pothole masks, and the resulting pixel-domain areas were converted into physical areas using both scale-conversion methods. The evaluation included five physical potholes observed at flight altitudes of 10, 15, and 20 m, resulting in 15 hold-out observations with field-measured areas used as reference values. Dynamic GSD reduced the mean pothole-level absolute percentage error from 45.03 percent to 22.13 percent, the mean pothole-level MAE from 0.1596 to 0.0806 square meters, and the mean pothole-level RMSE from 0.1800 to 0.0867 square meters. The mean cross-altitude coefficient of variation decreased from 46.44 percent to 9.51 percent. Dynamic GSD produced lower pothole-level error metrics for all five evaluated potholes and improved measurement consistency across flight altitudes. Residual errors remained because scale correction does not eliminate uncertainty associated with segmentation boundaries, altitude metadata, camera orientation, analytical camera assumptions, and fieldreference measurements. These findings should be considered preliminary because the evaluation was limited to five physical potholes and 15 observations collected from one road section.

Other files and links

Fingerprint

Pothole (geology)Sciences
EstimationSciences
Remote sensingSciences
Environmental scienceSciences
Estimation theorySciences
Noise (video)Sciences
Computer scienceSciences
MathematicsSciences
Stability (learning theory)Sciences
CalibrationSciences