Visual-Based Detection Method for Oil Leakage in Antarctic Power-Generation Cabin

作者：Xiao Ge, Tao Wang, Kanjian Zhang · 年份：2024 · DOI：10.1145/3651671.3651761 · 研究领域：Image Enhancement Techniques、Fire Detection and Safety Systems、Advanced Neural Network Applications

In the unattended periods of Antarctic Power-Generation Cabin, it is imperative to promptly identify oil leakage within the cabin to assess potential malfunctions in the oil storage module or generator. We proposed a visual-based method for oil leakage detection in the Antarctic Power-Generation Cabin, employing an orbital-type inspection robot platform. The detection method integrates an improved MBLLEN image enhancement algorithm, utilizing color features in the HSV color space for image segmentation and combines prior knowledge with perspective transformation for size estimation. The proposed image enhancement algorithm outperforms the mainstream algorithms in terms of brightness enhancement and color restoration. We also established an image set depicting oil leakage scenarios within the cabin where the method achieved an accuracy of about 98% and a precision of about 10 <?TeX $c{m}^2$?> . Furthermore, it exhibited resilience to robot positioning errors and insensitivity to varying lighting conditions, thereby demonstrating robust applicability.