Bridge Pier Surface Defect Detection Based on Improved YOLOV9

This research introduces an innovative surface defect detection methodology specifically designed for bridge piers, which integrates state-of-the-art image enhancement algorithms with sophisticated target detection frameworks. This hybrid approach effectively addresses some of the inherent limitations observed in existing deep learning-based defect detection methodologies, particularly under conditions of suboptimal image quality and challenges related to the detection of minute targets. Comparative results demonstrate that this novel technique achieves a 3.9% increase in the mean Average Precision (mAP50) over the baseline model. Furthermore, this is accomplished with a reduction in model complexity, as evidenced by a 9.8% decrease in the number of parameters and a substantial reduction in computational demand, quantified as a 7.5 GFLOPS decrease. This study not only advances the field of structural health monitoring but also enhances the operational efficiency of automated defect detection systems.

Name dc:type.qualificationname

msc

Level dc:type.qualificationlevel

masters

Grantor dc:publisher.institution

University of Wales Trinity Saint David

Year dc:date.issued

2024

Author dc:creator

• Cai, Hanzhe

• TK Peirianneg drydanol. Peirianneg Niwclear Electronig

Dc Identifier Grantnumber

UWTSD

OAI identifier oai:identifier

oai:repository.uwtsd.ac.uk:3308