Artificial Dataset Generation for Automated Aircraft Visual Inspection

Authors

Nathan J. Gaul, Air Force Institute of Technology
Robert C. Leishman, Air Force Institute of Technology

Document Type

Conference Proceeding

Publication Date

8-2021

Abstract

Aircraft visual inspection is both essential to the maintenance of an aircraft, and expensive and time-consuming to perform. Augmenting trained maintenance professionals with automated UAVs to collect and analyze images for aircraft inspection is an active research topic and a potential application of convolutional neural networks (CNNs). Training datasets for niche research topics such as aircraft visual inspection are small and challenging to produce, and the manual labeling process of these datasets produces subjective annotations. Self-driving car researchers have experimented with generating artificial datasets with modern computer graphics that can train for real-world driving scenarios. Our research borrows this idea and proposes a work-in-progress artificial data generation pipeline to create 3D rendered automatically annotated images for training CNNs for automated visual aircraft inspection.

Comments

This conference paper is available via subscription or purchase through the DOI link below.

DOI

10.1109/NAECON49338.2021.9696375

Source Publication

Proceedings of the IEEE National Aerospace Electronics Conference, NAECON

Recommended Citation

N. J. Gaul and R. C. Leishman, "Artificial Dataset Generation for Automated Aircraft Visual Inspection," NAECON 2021 - IEEE National Aerospace and Electronics Conference, Dayton, OH, USA, 2021, pp. 302-306, doi: 10.1109/NAECON49338.2021.9696375.