目前在大型建筑施工区域，仍然通过人工检查的方式监督施工人员佩戴安全帽，这种监管措施劳动成本大，效率低下，已经完全无法满足建筑行业高速发展的需求。近年来，在我国的建筑行业中，由于未佩戴安全帽造成的事故占总事故的60%以上，随着目标检测技术与智能摄像机的发展，使得实现智能化的安全帽佩戴检测成为可能。本文通过对主流的目标检测算法的分析，权衡检测精度与速度，选择YOLOv4作为基础算法开发安全帽佩戴检测算法。针对YOLOv4对小目标检测精度低，遮挡严重目标识别困难等问题，对YOLOv4进行改进提出一种新的安全帽佩戴检测算法。

首先，在YOLOv4的主干特征提取网络中引入多频谱通道注意力机制，对卷积神经网络提取的特征通道之间的相互依赖关系进行建模，自适应地调整各通道的特征响应值，提高模型的特征提取能力；其次，对YOLOv4的路径聚合网络和特征金字塔网络结构进行改进，将主干特征提取网络中分辨率低，细粒度特征丰富的浅层特征和语义信息丰富的深层特征进行更好地融合，输出分辨率更大的特征图检测小目标；最后对损失函数进行改进，降低了大量简单负样本在训练中所占的权重。

为了获得最佳模型，以每个类别的AP值和mAP值作为评价指标，对模型的改进点进行两两融合实验研究。最后通过综合实验研究和实际的图片测试，验证了本文所改进算法的有效性，通过以上三点改进后，YOLOv4的mAP值由原来的80.90%提升到90.11%，并且通过实际的图片测试后，改进后的模型具有更好地检测效果。

为了将改进后的算法应用到实际的建筑现场，通过华为软件定义摄像机进行二次开发，将训练好的模型通过转换，打包和签名等操作，移植到华为软件定义摄像机上形成了一套完整的安全帽佩戴检测系统。

Construction workers are still required to wear safety helmets in large-scale construction zones, as determined by manual inspection. This type of supervision has high personnel costs and low efficiency, and it has entirely failed to satisfy the demands of the construction industry's rapid growth. Accidents caused by not wearing a helmet account for more than 60% of all accidents in the construction business in my nation in recent years. With the advancement of target detection technologies and smart cameras, intelligent helmet wearing detection is now achievable. This study examines popular target detection algorithms, weighs detection accuracy and speed, and chooses YOLOv4 as the foundation for a helmet-wearing detection method. YOLOv4 is upgraded, and a new helmet wearing detection algorithm is developed, in order to address the difficulties of YOLOv4's low detection accuracy for small targets and difficulty in detecting substantially obscured targets.

To model the interdependence between the feature channels extracted by the convolutional neural network, adaptively adjust the feature response value of each channel, and improve the model, the multi-spectral channel attention mechanism is first introduced into the backbone feature extraction network of YOLOv4. Second, YOLOv4's path aggregation network and feature pyramid network structure have been improved, and the backbone feature extraction network's low-resolution, fine-grained feature-rich shallow features and semantic information-rich deep features have been improved. Finally, the loss function is improved to reduce the weight of a large number of simple negative samples in training. Fusion, the output feature map with higher resolution, detects small targets; finally, the loss function is improved to reduce the weight of a large number of simple negative samples in training.

The AP value and mAP value of each category are utilized as evaluation indicators, and a pairwise fusion experiment research is conducted on the model's improvement points in order to produce the optimal model. Finally, the usefulness of the modified algorithm in this study is confirmed through extensive experimental research and actual picture testing. The mAP value of YOLOv4 has grown from 80.90 percent to 90.11 percent following the above three changes, and the upgraded model has a superior detection impact after passing the actual photo test.

The trained model is transplanted to Huawei software-defined cameras through operations such as conversion, packaging, and signature, forming a complete set of Safety helmet wearing detection system, in order to apply the improved algorithm to the actual construction site, through secondary development of Huawei software-defined cameras.

[1] 邱浩然.基于改进YOLOv3的安全帽检测算法研究与实现[D].西南交通大学,2020.002334.

[2] 闫金鹏.2009年至2017年房屋市政工程建设领域生产安全事故分析与安全控制重[J].建筑安全,2019,34(02):61-63.

[3] 常欣,刘鑫萌.建筑施工人员不合理佩戴安全帽事故树分析[J].吉林建筑大学学报,2018,35(06):65-69.

[4] Girshick R,Donahue J,Darrell T,et al.Rich feature hierarchies for accurate object detection and semantic segmentation[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2014:580-587

[5] Dalal N,Triggs B.Histograms of oriented gradients for human detection[C]. // Computer Vision and Pattern Recognition, 2005. CVPR 2005. IEEE Computer Society Conference on vol.1.2005:886-893.

[6] Felzenszwalb P, McAllester D, Ramanan D. A discriminatively trained, multiscale, deformable part model[C]//2008 IEEE conference on computer vision and pattern recogniion.Ieee, 2008:1-8.

[7] Alex Krizhevsky et al.“ImageNet classification with deep convolutional neural networks” Communications of The ACM, 2017, 60(6): 84-90.

[8] He K, Zhang X, Ren S, et al. Spatial pyramid pooling in deep convolutional networks for visual recognition[J].IEEE transactions on pattern analysis and machine intelligence, 2015,37(9):1904-1916.

[9] Girshick R.Fast r-cnn[C]//Proceedings of the IEEE international conference on computer vision.2015:1440-1448.

[10] Ren S,He K,Girshick R,et al.Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2017,39(6):1137-1149.

[11] Lin T Y, Dollár P, Girshick R, et al. Feature pyramid networks for object detection[C]// Proceedings of the IEEE conference on computer vision and pattern recognition.2017: 2117-2125.

[12] Redmon J, Divvala S, Girshick R, et al.You only look once: Unified,real-time object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2016:779-788.

[13] Liu W, Anguelov D, Erhan D, et al. Ssd:Single shot multibox detector[C]//European conference on computer vision.Springer, Cham,2016:21-37.

[14] Redmon J, Farhadi A. YOLO9000 : better, faster, stronger[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2017:7263-7271.

[15] Redmon J, Farhadi A.Yolov3: An incremental improvement[J].arXiv preprint arXiv:1804. 02767,2018.

[16] Bochkovskiy A, Wang C Y,Liao H Y M.Yolov4:Optimal speed and accuracy of object detection[J].arXiv preprint arXiv:2004.10934, 2020.

[17] Agnes Kelm et al. “Mobile passive Radio Frequency Identification (RFID) portal for automated and rapid control of Personal Protective Equipment (PPE) on construction sites”Automation in Construction.2013,36(Dec.):38-52.

[18] Dong S, He Q, Li H,et al. Automated PPE misuse identification and assessment for safety performance enhancement[M]//ICCREM 2015.2015:204-214.

[19] Zhang H,Yan X, Li H,et al. Real-time alarming, monitoring, and locating for non-hard-hat use in construction[J]. Journal of construction engineering and management, 2019, 145(3): 1943-7862.

[20] Wu H, Zhao J. An intelligent vision-based approach for helmet identification for work safety[J]. Computers in Industry, 2018, 100:267-277.

[21] Mneymneh B E, Abbas M, Khoury H.Vision-based framework for intelligent monitoring of hardhat wearing on construction sites[J].Journal of Computing in Civil Engineering, 2019,33(2):04018066.

[22] 杨莉琼,蔡利强,古松.基于机器学习方法的安全帽佩戴行为检测[J].中国安全生产科学技术,2019,15(10):152-157.

[23] 徐守坤,倪楚涵,吉晨晨等.一种基于安全帽佩戴检测的图像描述方法研究[J].小型微型计算机系统,2020,41(04):812-819.

[24] 王雨生,顾玉宛,庄丽华等.复杂姿态下的安全帽佩戴检测方法研究[J].计算机工程与应用,2022,58(01):190-196.

[25] 肖体刚,蔡乐才,高祥等.改进YOLOv3的安全帽佩戴检测方法[J].计算机工程与应用,2021,57(12):216-223.

[26] 葛青青,张智杰,袁珑,李秀梅,孙军梅.融合环境特征与改进YOLOv4的安全帽佩戴检测[J].中国图象图形学报,2021,26(12):2904-2917.

[27] Cao W, Zhang J, Cai C, et al. CNN-based intelligent safety surveillance in green IoT applications[J]. China Communications, 2021, 18(1) : 108-119.

[28] Han K, Zeng X. Deep Learning-Based Workers Safety Helmet Wearing Detection on Construction Sites Using Multi-Scale Features[J].IEEE Access,2021,10:718-729.

[29] Gu Y,Wang Y, Shi L,et al. Automatic detection of safety helmet wearing based on head region location[J]. IET Image Processing, 2021, 15(11) : 2441-2453.

[30] Li N, Lyu X,Xu S, et al. Incorporate online hard example mining and multi-part combination into automatic safety helmet wearing detection[J].IEEE Access,2020, 9:139536-139543.

[31] Shen J,Xiong X,Li Y,et al.Detecting safety helmet wearing on construction sites with bounding‐box regression and deep transfer learning[J].Computer‐Aided Civil and Infrastructure Engineering,2021,36(2):180-196.

[32] Wu J,Cai N,Chen W,et al.Automatic detection of hardhats worn by construction personnel:A deep learning approach and benchmark dataset[J].Automation in Construction,2019,106(Oct.):102894.1-102894.7.

[33] Wang C Y,Liao H Y M, Wu Y H,et al.CSPNet:A new backbone that can enhance learning capability of CNN[C]//Proceedings of the IEEE/CVF conference on computer vision and pattern recognition workshops.2020:390-391.

[34]Liu S,Qi L,Qin H,et al.Path aggregation network for instance segmentation [C]// Proceedings of the IEEE conference on computer vision and pattern recognition. 2018:8759-8768.

[35] Chen X, Kundu K,Zhu Y,et al.3d object proposals for accurate object class detection[J]. Advances in neural information processing systems,2015,28.

[36] Li J,Liang X, Wei Y,et al.Perceptual generative adversarial networks for small object detection[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2017:1222-1230.

[37] Bell S,Zitnick C L,Bala K,et al.Inside-outside net: Detecting objects in context with skip pooling and recurrent neural networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2016:2874-2883.

[38] Cao G,Xie X,Yang W,et al.Feature-fused SSD:Fast detection for small objects[C]//Ninth International Conference on Graphic and Image Processing (ICGIP 2017).International Society for Optics and Photonics,2018,10615:106151E.

[39] Menikdiwela M,Nguyen C,Li H,et al.CNN-based small object detection and visualization with feature activation mapping[C]//2017 International Conference on Image and Vision Computing New Zealand (IVCNZ).IEEE,2017:1-5.

[40] Kisantal M,Wojna Z,Murawski J,et al.Augmentation for small object detection[J].arXiv preprint arXiv:1902.07296,2019.

[41]Cheng P,Liu W,Zhang Y,et al.LOCO:local context based faster R-CNN for small traffic sign detection[C]//International conference on multimedia modeling.Springer,Cham, 2018:329-341.

[42]Ren Y,Zhu C,Xiao S.Small object detection in optical remote sensing images via modified faster R-CNN[J].Applied Sciences,2018,8(5):813.

[43] Szegedy C,Liu W,Jia Y,et al.Going deeper with convolutions[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2015:1-9.

[44] He K,Zhang X,Ren S,et al.Deep residual learning for image recognition[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2016:770-778.

[45] Hu J, Shen L, Sun G. Squeeze-and-excitation networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2018:7132-7141.

[46] Qin Z,Zhang P, Wu F, et al.Fcanet:Frequency channel attention networks[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision.2021:783-792.

[47] Woo S,Park J,Lee J Y,et al.Cbam:Convolutional block attention module[C]//Proceedings of the European conference on computer vision (ECCV).2018:3-19.

[48] Oksuz K,Cam B C,Kalkan S,et al.Imbalance problems in object detection:A review[J]. IEEE transactions on pattern analysis and machine intelligence,2020,43(10):3388-3415.

[49] Lin T Y,Goyal P,Girshick R,et al.Focal loss for dense object detection[C]//Proceedings of the IEEE international conference on computer vision.2017:2980-2988.

[50] Zheng Z,Wang P,Liu W,et al.Distance-IoU loss:Faster and better learning for bounding box regression[C]//Proceedings of the AAAI Conference on Artificial Intelligence.2020, 34(07):12993-13000.