UAV aerial image-based apparent defect detection method for embankments using an integrated framework of improved YOLOv8 and SAHI

Dongyang Yuan, Changwei Qiu, Bowen Wei, Jiawei Li · 2026

Read UAV aerial image-based apparent defect detection method for embankments using an integrated framework of improved YOLOv8 and SAHI by Dongyang Yuan, Changwei Qiu, Bowen Wei, Jiawei Li (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13500846.txt.

Description

Dongyang Yuan leads a four-author team presenting a 2026 drone-based method for finding surface defects, such as cracks and collapse pits, on large embankments. Manual inspection of long flood-control embankments is slow and misses defects. So the team photographed embankments from drones and trained a detection model on the images. Because real defect photos were scarce, a CycleGAN network generated extra synthetic defect images under varied lighting to enlarge the training set.

The detector builds on the YOLOv8 network, strengthened with an attention module called CBAM and a localization loss called SIoU to sharpen accuracy. A slicing aid named SAHI was added at inference time so small defects remain visible in high-resolution aerial frames taken from far away. In an engineering case study the improved network beat plain YOLOv8 on accuracy and missed fewer defects, and both the synthetic images and the slicing aid measurably helped.

Catalog record

CREATORDongyang Yuan, Changwei Qiu, Bowen Wei, Jiawei Li
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE40163
LICENSEhttps://creativecommons.org/licenses/by/4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/article/MED/PMC13500846

Browse the full library · About the Tower of Babel