Intelligent Dust-Occlusion Restoration for Visual Inspection in Underground Coal Mining

作者：Guoqing Wu, Zhenduo Song, Junjie Li, Qingyou Jiang, Zhenzhong Pang, Zhaoyuan Ma · 发表于：Journal of Physics Conference Series · 年份：2025 · DOI：10.1088/1742-6596/3055/1/012007 · 研究领域：3D Surveying and Cultural Heritage、Geological Modeling and Analysis

Abstract In the process of constructing an augmented reality scene for the coal mining face, the occlusion of the shearer by coal lumps, dust, and other objects that fall during the mining process affects the complete visualization of the shearer in the virtual scene, which brings negative impact on the intelligent detection and control of the shearer based on visual methods. To address this issue, this paper presents an intelligent dust-occlusion recovery method based on advanced algorithmic techniques. Utilizing the YOLOv8 model to segment the occluded shearer from the original image significantly reduces the complexity of image restoration. The FastGAN network has been modified for image inpainting tasks by integrating FFC to enhance contextual awareness and training speed, while incorporating an attention mechanism, significantly improving the model’s perceptual capability in scenarios with large-area image missing, thus enabling the proposed method to achieve rapid and intact visualization of the coal mining machine shearer in augmented reality scenarios.