基于计算机视觉的石油库施工人员未穿工服行为识别系统研究

Behavior recognition system for construction workers in oil depots failing to wear protective clothing based on computer vision

提出一种基于计算机视觉技术的石油库施工人员未穿工服行为识别系统。该系统采用YOLOv5s和ResNet-50级联检测器,通过构建丰富的石油库施工人员数据集并进行实验验证,实现了对未穿工服行为的精准检测。实验结果表明,ResNet-50模型在测试集上展现出高准确度,为石油库安全生产管理提供了有力的技术支撑。相较于SqueezeNet,ResNet-50在识别准确率上表现更优,但后者在显存占用和推理速度上更具优势。本文研究不仅有助于提升石油库安全管理的智能化水平,降低人为因素导致的安全风险,还为类似场景下的安全监控提供了有益的参考和借鉴。随着计算机视觉技术的不断发展,该系统有望在石油库安全生产管理中发挥更加重要的作用。

This paper proposes a behavior recognition system for oil depot construction personnel not wearing uniforms based on computer vision technology.The system adopts a cascaded detector combining YOLOv5s and ResNet-50,and achieves precise detection of non-uniform wearing behavior through the construction of a rich dataset of oil depot construction personnel and experimental verification.Experimental results show that the ResNet-50 model exhibits high accuracy on the test set,providing strong technical support for the safe production management of oil depots.Compared to SqueezeNet,ResNet-50 performs better in recognition accuracy,while the latter excels in memory usage and inference speed.The research in this paper not only helps to improve the intelligent level of safety management in oil depots and reduce safety risks caused by human factors,but also provides useful references and lessons for safety monitoring in similar scenarios.With the continuous development of computer vision technology,this system is expected to play a more important role in the safe production management of oil depots.