基于轻量化深度学习模型的安全帽检测方法  

Helmet Detection Method Based on Lightweight Deep Learning Model

作　　者：秦子豪 雷鸣[1] 宋文广[2] 张维[1] QIN Zi-hao;LEI Ming;SONG Wen-guang;ZHANG Wei(School of Urban Construction, Yangtze University, Jingzhou 434023, China;School of Computer Science, Yangtze University, Jingzhou 434023, China)

机构地区：[1]长江大学城市建设学院,荆州434023 [2]长江大学计算机科学学院,荆州434023

出　　处：《科学技术与工程》2022年第14期5659-5665,共7页Science Technology and Engineering

基　　金：中石油创新基金(2017D-5007-0604);湖北省教育厅省级教研项目(2018284)。

摘　　要：基于对施工现场管理中安全帽检测重要性的认识,同时考虑工程项目中硬件设施的成本控制等现实问题,提出了一种基于深度学习网络Tiny-YOLO v3的轻量化改进版本LT-YOLO的安全帽检测技术方法。LT-YOLO增加了网络的输出层,并包含一种创新的R-DSC特征提取模块,该模块能够在不改变网络输入与输出大小的前提下,极大地降低模型的复杂度。实验结果表明,LT-YOLO在轻量化效果与检测性能之间取得了优良的平衡,在3.5 M参数量的基础上达到了59.3 mAP(mean average precision)和59.4%Recall。因此LT-YOLO拥有极低的参数量和计算量,对高算力硬件的依赖性低,适用于实际工程管理应用的施工现场安全管理,能够极大地降低企业成本,提升施工安全管理的水平。Based on the importance of helmet detection in construction site management and the cost control of hardware facilities in engineering projects,a helmet detection approach based on Lighter and Tiny-YOLO(LT-YOLO),a lightweight and improved version of the deep learning network Tiny-YOLO v3 was proposed.The number of prediction layer was increased and an innovative R-DSC feature extraction module was introduced in LT-YOLO.The complexity of the model could be greatly reduced by R-DSC module without changing the size of the network inputs and outputs.The experimental results show that LT-YOLO achieve an excellent balance between light weight and detection performance,reaching 59.3 mAP(mean average precision)and 59.4%Recall with only 3.5 M parameters.Because of very few parameters and very low computation,LT-YOLO has low dependence on high computing hardware,and is suitable for actual construction site safety management.LT-YOLO can greatly reduce the cost of enterprises and improve the level of construction safety management.

关 键 词：施工现场管理 安全帽检测 深度学习 轻量化 工程管理 

分 类 号：TN277[电子电信—物理电子学]