Seepage Detection in the Sealed Cabin of a Certain Type of Aircraft Based on an Enhanced Channel Attention Mechanism

Abstract

With the advancement of the aviation industry, aircraft safety is a priority as closed cavities are susceptible to water infiltration during flight, which can lead to short circuits and operational problems. The current standard manual inspection is inefficient. Our paper describes an industrial vision inspection method that combines an enhanced channel attention mechanism with YOLOV9. 3,000 images generated by aircraft water penetration are replicated in a simulated environment, expanded to 6,000 by data enhancement techniques such as rotation and flipping, and then accurately annotated with targets for model training and validation. Feature extraction was improved on the standard YOLOV9 framework by adapting its weighting strategy to improve the SELayer channel attention mechanism to emphasize specific features of water infiltration. Experiments show a 4.7% improvement in accuracy on our dataset, with a single-frame inference time of 23 ms, which meets real-time requirements. This method provides a reliable automated solution for aircraft maintenance.