基于LEDNet的露天矿监控低光照图像增强模型

A Low-Light Image Enhancement Model for Open-Pit Mine Monitoring Based on LEDNet

摘要: 露天矿监控视频常受到图像严重退化造成的影响，图像退化起因包括防爆要求对监控设备性能的限制、矿区内大型设备和建筑物的遮挡，以及生产作业多在夜间进行导致的光照不足等。这些因素共同导致监控图像出现噪点多、分辨率低、照度不足，影响图像的清晰度和可见度，从而对生产调度和安全监控产生负面影响。为解决这一问题，本研究引入了一个针对露天矿监控视频特有退化问题的高效图像恢复技术。开发了一个新颖的模拟露天矿监控视频退化流程，专门模拟露天矿环境中的光照不足和遮挡造成的图像退化情况，包括针对夜间低光照和防爆限制下的图像质量改善。在此基础上，构建了一个大规模数据集，提出了LEDNet深度学习网络，该网络专为联合提升低光照图像的亮度和清晰度而设计，考虑到了低光照增强和去模糊任务之间的相互作用。通过这一创新方法，不仅显著提高了露天矿监控图像的质量，而且为在恶劣环境下提高监控系统性能提供了有效途径，这对于提高生产效率和确保作业安全具有重要意义。大量实验证明，本文方法在处理真实露天矿监控视频中的图像时表现出色，有效地提升了图像的可见度和清晰度。

Abstract: Open-pit mine monitoring videos are often severely affected by image degradation, mainly due to the limitations on monitoring equipment performance imposed by explosion-proof requirements, the obstruction by large equipment and structures within the mining area, and inadequate lighting caused by nighttime operations. These factors collectively lead to noisy images, low resolution, and insufficient illumination, which affect the clarity and visibility of the images, thereby negatively impacting production scheduling and worker safety monitoring. To address this issue, this study introduced an efficient image restoration technique specifically tailored to the degradation problems unique to open-pit mine monitoring videos. A novel simulation of the degradation process in open-pit mine monitoring videos was developed, designed to replicate image degradation scenarios caused by low light and obstructions typical in the mining environment, including image quality improvements under low-light conditions and explosion-proof constraints. Based on this, a large-scale dataset was constructed, and a deep learning network named LEDNet was proposed. The network was specifically designed to jointly enhance the brightness and clarity of low-light images, taking into account the interaction between low-light enhancement and deblurring tasks. This innovative approach not only significantly improves the quality of open-pit mine monitoring images but also provides an effective means to enhance the performance of monitoring systems in harsh environments, which is of great significance for improving production efficiency and ensuring operational safety. Extensive experiments demonstrate that the method performs excellently in processing images in real open-pit mine monitoring videos, effectively enhancing image visibility and clarity.