不安全的环境和行为是导致安全事故的主要根源。创造一个良好的工作环境已成为人们的迫切需求，而安全标志则扮演着传达安全标准和规则约束的重要信息媒介的角色。安全标志目标检测技术在保障煤矿安全、提高工作效率以及推动智慧煤矿安全建设方面发挥着巨大的作用。通过将先进的目标检测技术与安全标志结合，我们能够实现更高效的自动化安全监管，为人们提供更加安全可靠的工作环境，进而推动整体社会的安全意识和生产效率。本文主要研究内容如下：

（1）提出基于VGG16改进的安全标志分类识别网络模型，首先，改进SoftMax分类器由1000类改为10类以适应本研究的数据集，利用评估指标对模型性能进行分析，微调参数进行优化，探究不同超参数如epoch、batchsize对模型性能的影响；其次，将改进的模型在数据增强前后的数据集上分别进行试验，探究数据增强对模型性能的影响；最后将改进后的网络与未改进的网络以及其余网络模型进行对比分析。试验结果表明：当迭代次数为50、批大小为64时，top-5 accuracy最高，达到99.83%，损失函数能够快速并稳定的收敛，且该模型在经过数据增强的数据集上的precision比未经过数据增强的数据集高27.09%。

（2）为了提高小目标的检测精度，本文提出一种改进YOLOv4-tiny模型的安全标志检测算法。首先，将注意力机制引入模型中，加强网络对图像中重要信息的学习；其次，使用 Soft-NMS 算法替代YOLOv4-tiny中原有的NMS算法，避免由于非极大值抑制而导致的的漏检问题；最后，通过引入Focal Loss来优化网络的损失函数，以增强模型对困难样本的预测能力并提高预测框的回归精度。将改进后的模型与未改进的传统YOLOv4-tiny模型以及双阶段算法代表算法Faster RCNN等主流模型进行对比。试验结果表明：改进YOLOv4-tiny算法的mAP达到了97.76%，比未改进的YOLOv4-tiny算法的mAP提高7.55%。

（3）构建安全标志数据集，本文对多种环境中的安全标志进行收集，经过图像预处理与数据增强建立包含必须戴防护手套、必须佩戴安全帽、必须佩戴防护口罩、当心触电、当心中毒、紧急出口、禁止攀登、禁止吸烟、禁止烟火、应急避难场所的4014幅图片的安全标志数据集。根据（训练集+验证集）：测试集= 9：1和训练集：验证集= 9：1的比值，将整个数据集分为训练集、验证集和测试集。同时为使本数据集适用于目标检测模型，使用labelimg工具进行人工标注，生成对应的xml文件。

Unsafe environments and behaviors are identified as the primary root causes of incidents. Creating a conducive work environment has become an urgent necessity, with safety signs playing a pivotal role as a key information medium for conveying safety standards and regulatory constraints. Safety sign target detection technology plays an important role in ensuring coal mine safety, improving work efficiency and promoting intelligent coal mine safety construction. By integrating advanced object detection techniques with safety signs, we can achieve more efficient automated safety supervision, providing people with safer and more reliable work environment, thereby promoting overall societal safety awareness and productivity. The main research contents of this paper are as follows:
(1)    A modified safety sign classification recognition network model based on VGG16 is proposed. Firstly, the SoftMax classifier is improved from 1000 classes to 10 classes to adapt to the dataset used in this study. The performance of the model is analyzed using evaluation metrics, and parameters are fine-tuned for optimization. The effects of different hyperparameters such as epoch and batch size on model performance are investigated. Secondly, the improved model is tested on datasets before and after data augmentation to explore the impact of data augmentation on model performance. Finally, the improved network is compared with the unmodified network and other network models. Experimental results indicate that when the number of epoch is 50 and the batch size is 64, the top-5 accuracy is the highest, reaching 99.83%. The loss function converges quickly and steadily. Moreover, the precision of this model on the dataset after data augmentation is 27.09% higher than that on the dataset without data augmentation.
(2)    To enhance the detection accuracy of small targets, a real-time safety sign detection algorithm based on improved YOLOv4-tiny is proposed. Firstly, an attention mechanism is introduced into the model to strengthen the network's learning of important information in images. Secondly, the Soft-NMS algorithm is used to replace the original NMS algorithm in YOLOv4-tiny to avoid missed detections caused by non-maximum suppression. Finally, Focal Loss is introduced to optimize the network's loss function, enhancing the model's prediction capability for difficult samples and improving the regression accuracy of predicted boxes. The improved model is compared with the unmodified traditional YOLOv4-tiny model and mainstream models such as Faster RCNN. Experimental results show that the m AP of the improved YOLOv4-tiny algorithm reaches 97.76%, which is 7.55% higher than that of the unmodified YOLOv4-tiny algorithm YOLOv4-tiny model and mainstream models such as the two-stage algorithm representative, Faster R-CNN. 
(3)    To construct a safety signs dataset, this study collected various safety signs from multiple environments. After undergoing image preprocessing and data augmentation, a safety sign dataset containing 4014 images was established, including signs such as " Wearing protective gloves," " Wearing safety helmet," " Wearing dustproof mask," " Warning electric shock," " Warning poisoning," "Emergency Exit," "No Climbing," "No Smoking," "No Fireworks," and "Emergency Shelter." Following the ratio of (training set + validation set): test set = 9:1 and training set: validation set = 9:1, the entire dataset was divided into training, validation, and test sets. Moreover, to make this dataset suitable for object detection models, manual annotation was conducted using the labelimg tool, generating corresponding XML files.
 

[1] 中华人民共和国国务院：《建设工程安全生产管理条例》，2004 年 2 月 1 日，第 393 号

[2] 2002 年《中华人民共和国安全生产法》第一章第十五条（2014 修正）

[3] 马永杰, 李雪燕, 宋晓凤. 基于改进深度卷积神经网络的交通标志识别 [J]. 激光与光电子学进展, 2018, 55(12): 8.

[4] LIU Y, HUANG A, LUO Y, et al. Federated learning-powered visual object detection for safety monitoring [J]. AI magazine: Artificial intelligence, 2021, (2): 42.

[5] 顾润龙.大数据下的机器学习算法探讨[J].通讯世界,2019, 26(05): 279-280.

[6] ZHOU L, PAN S, WANG J, et al. Machine learning on big data: Opportunities and challenges [J]. Neurocomputing, 2017, 237: 350-361.

[7] WANG X, HAN T X, YAN S. An HOG-LBP human detector with partial occlusion handling; proceedings of the 2009 IEEE 12th international conference on computer vision, F, 2009 [C]. IEEE.

[8] ERFANI, SARAH M, LECKIE, et al. High-dimensional and large-scale anomaly detection using a linear one-class SVM with deep learning [J]. Pattern Recognition: The Journal of the Pattern Recognition Society, 2016.

[9] KIM Y. Convolutional Neural Networks for Sentence Classification [J]. Eprint Arxiv, 2014.

[10] HAN S, MAO H, DALLY W J. Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding [J]. Fiber, 2015, 56(4): 3-7.

[11] PERDANA A B, PRAHARA A. Face Recognition Using Light-Convolutional Neural Networks Based On Modified Vgg16 Model; proceedings of the 2019 International Conference of Computer Science and Information Technology (ICoSNIKOM), F, 2019 [C].

[12] SZEGEDY C, LIU W, JIA Y, et al. Going Deeper with Convolutions [J]. IEEE Computer Society, 2014.

[13] HE K, ZHANG X, REN S, et al. Deep residual learning for image recognition; proceedings of the Proceedings of the IEEE conference on computer vision and pattern recognition, F, 2016 [C].

[14] GIRSHICK R. Fast R-CNN [J]. Computer Science, 2015.

[15] REN S, HE K, GIRSHICK R, et al. Faster r-cnn: Towards real-time object detection with region proposal networks [J]. Advances in neural information processing systems, 2015, 28.

[16] REDMON J, FARHADI A. YOLOv3: An Incremental Improvement [J]. arXiv e-prints, 2018.

[17] BOJARSKI M, DEL TESTA D, DWORAKOWSKI D, et al. End to End Learning for Self-Driving Cars [J]. 2016.

[18] ZHAN S, BENENSON R, OMRAN M, et al. Towards Reaching Human Performance in Pedestrian Detection [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018.

[19] GEERT, LITJENS, THIJS, et al. A survey on deep learning in medical image analysis [J]. Medical Image Analysis, 2017.

[20] SHEN D, WU G, SUK H I. Deep Learning in Medical Image Analysis [J]. Annual Review of Biomedical Engineering, 2017, 19(1): 221-248.

[21] LUO W, LI Y, URTASUN R, et al. Understanding the Effective Receptive Field in Deep Convolutional Neural Networks [Z]. 2017

[22] XIA G S, BAI X, DING J, et al. DOTA: A Large-scale Dataset for Object Detection in Aerial Images [J]. IEEE, 2018.

[23] LIU X, YAN W Q, KASABOV N. Vehicle-related scene segmentation using CapsNets; proceedings of the 2020 35th International Conference on Image and Vision Computing New Zealand (IVCNZ), F, 2020 [C]. IEEE.

[24] LIU X, NEUYEN M, YAN W Q. Vehicle-related scene understanding using deep learning; proceedings of the Asian Conference on Pattern Recognition, F, 2019 [C]. Springer.

[25] HOUBEN S, STALLKAMP J, SALMEN J, et al. Detection of traffic signs in real-world images: The German Traffic Sign Detection Benchmark; proceedings of the The 2013 international joint conference on neural networks (IJCNN), F, 2013 [C]. Ieee.

[26] GREENHALGH J, MIRMEHDI M. Real-time detection and recognition of road traffic signs [J]. IEEE transactions on intelligent transportation systems, 2012, 13(4): 1498-1506.

[27] KO J G, MUSALOIU-ELEFTERI R, LIM J H, et al. MEDiSN: Medical emergency detection in sensor networks [J]. Acm Transactions on Embedded Computing Systems, 2008.

[28] WANG Y, WANG H. Multilingual convolutional, long short-term memory, deep neural networks for low resource speech recognition [J]. Procedia Computer Science, 2017, 107: 842-847.

[29] KRIZHEVSKY A, SUTSKEVER I, HINTON G E. Imagenet classification with deep convolutional neural networks [J]. Communications of the ACM, 2017, 60(6): 84-90.

[30] 张逞逞. 交通标志识别及其目标检测深度学习算法研究 [D]. 燕山大学.

[31] SIMONYAN K, ZISSERMAN A. Very deep convolutional networks for large-scale image recognition [J]. arXiv preprint arXiv:14091556, 2014.

[32] SZEGEDY C, WEI L, JIA Y, et al. Going deeper with convolutions; proceedings of the 2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), F, 2015 [C].

[33] GAJHEDE N, BECK O, PURWINS H. Convolutional Neural Networks with Batch Normalization for Classifying Hi-hat, Snare, and Bass Percussion Sound Samples [J]. ACM, 2016: 111-511.

[34] DATAL N. Histograms of oriented gradients for human detection; proceedings of the Proc 2005 International Conference on Computer Vision and Pattern Recognition, F, 2005 [C]. IEEE Computer Society.

[35] LOWE D G. Object recognition from local scale-invariant features; proceedings of the Proceedings of the seventh IEEE international conference on computer vision, F, 1999 [C]. Ieee.

[36] PAPAGEORGIOU C P, OREN M, POGGIO T. A general framework for object detection; proceedings of the Sixth International Conference on Computer Vision (IEEE Cat No 98CH36271), F, 1998 [C]. IEEE.

[37] HINTON G E, SALAKHUTDINOV R R. Reducing the dimensionality of data with neural networks [J]. science, 2006, 313(5786): 504-507.

[38] LECUN Y, BOTTOU L, BENGIO Y, et al. Gradient-based learning applied to document recognition [J]. Proceedings of the IEEE, 1998, 86(11): 2278-2324.

[39] HINTON G E, OSINDERO S, TEH Y-W. A fast learning algorithm for deep belief nets [J]. Neural computation, 2006, 18(7): 1527-1554.

[40] DENG J, DONG W, SOCHER R, et al. Imagenet: A large-scale hierarchical image database; proceedings of the 2009 IEEE conference on computer vision and pattern recognition, F, 2009 [C]. Ieee.

[41] GIRSHICK R, DONAHUE J, DARRELL T, et al. Rich feature hierarchies for accurate object detection and semantic segmentation; proceedings of the Proceedings of the IEEE conference on computer vision and pattern recognition, F, 2014 [C].

[42] UIJLINGS J R, VAN DE SANDE K E, GEVERS T, et al. Selective search for object recognition [J]. International journal of computer vision, 2013, 104(2): 154-171.

[43] 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.

[44] GIRSHICK R. Fast r-cnn; proceedings of the Proceedings of the IEEE international conference on computer vision, F, 2015 [C].

[45] REDMON J, DIVVALA S, GIRSHICK R, et al. You only look once: Unified, real-time object detection; proceedings of the Proceedings of the IEEE conference on computer vision and pattern recognition, F, 2016 [C].

[46] REDMON J, FARHADI A. YOLO9000: better, faster, stronger; proceedings of the Proceedings of the IEEE conference on computer vision and pattern recognition, F, 2017 [C].

[47] LIU W, ANGUELOV D, ERHAN D, et al. Ssd: Single shot multibox detector; proceedings of the European conference on computer vision, F, 2016 [C]. Springer.

[48] LIN T-Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection; proceedings of the Proceedings of the IEEE international conference on computer vision, F, 2017 [C].

[49] LAW H, DENG J. Cornernet: Detecting objects as paired keypoints; proceedings of the Proceedings of the European conference on computer vision (ECCV), F, 2018 [C].

[50] REZATOFIGHI H, TSOI N, GWAK J, et al. Generalized intersection over union: A metric and a loss for bounding box regression; proceedings of the Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, F, 2019 [C].

[51] LIU W, LIAO S, REN W, et al. High-level semantic feature detection: A new perspective for pedestrian detection; proceedings of the Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, F, 2019 [C].

[52] 曹捷, 郭志彬, 潘立志, 等. 高空作业场景下的安全带穿戴检测 [J]. 湖南科技大学学报(自然科学版), 2022, 37(01): 92-99.

[53] 房霆宸, 左俊卿, 龚剑. 计算机视觉在建筑工程施工领域的研究与应用 [J]. 建筑施工, 2021, 43(11): 2376-9+82.

[54] TEIZER J, CALDAS C H, HAAS C T. Real-time three-dimensional occupancy grid modeling for the detection and tracking of construction resources [J]. Journal of Construction Engineering and Management, 2007, 133(11): 880-8.

[55] CHENG T, TEIZER J. Real-time resource location data collection and visualization technology for construction safety and activity monitoring applications [J]. Automation in construction, 2013, 34: 3-15.

[56] 曾浩. 工程施工智能安全帽系统研究 [D]; 哈尔滨工业大学, 2017.

[57] 夏明华, 刘主光, 钱碧甫, 等. 基于人工智能的智能安全帽的研制 [J]. 电工技术, 2018, (15): 80-81.

[58] 杨登杰,钟伦珑,兰二斌,等.基于北斗的智能安全帽系统的设计[J].工业控制计算机,2019,32(07):15-17.

[59] 裴兴旺, 周晓, 庞喆, 等. 钢混排架厂房再生利用施工安全监测传感器优化布置 %J 建筑经济 [J]. 2023, 44(S2): 510-514.

[60] DONG S, HE Q, LI H, et al. Automated PPE misuse identification and assessment for safety performance enhancement [M]. ICCREM 2015. 2015: 204-214.

[61] KELM A, LAUßAT L, MEINS-BECKER A, et al. Mobile passive Radio Frequency Identification (RFID) portal for automated and rapid control of Personal Protective Equipment (PPE) on construction sites [J]. Automation in construction, 2013, 36: 38-52.

[62] CHEHRI H, CHEHRI A, SAADANE R. Traffic signs detection and recognition system in snowy environment using deep learning; proceedings of the The Proceedings of the Third International Conference on Smart City Applications, F, 2020 [C]. Springer.

[63] JIN M, CHEN X, LAI G, et al. Glove detection system based on VGG-16 network; proceedings of the 2020 13th International Symposium on Computational Intelligence and Design (ISCID), F, 2020 [C]. IEEE.

[64] KOLAR Z, CHEN H, LUO X. Transfer learning and deep convolutional neural networks for safety guardrail detection in 2D images [J]. Automation in construction, 2018, 89(MAY): 58-70.

[65] FANG Q, LI H, LUO X, et al. Detecting non-hardhat-use by a deep learning method from far-field surveillance videos [J]. Automation in Construction, 2018, 85: 1-9.

[66] FANG W, DING L, ZHONG B, et al. Automated detection of workers and heavy equipment on construction sites: A convolutional neural network approach [J]. Advanced Engineering Informatics, 2018, 37: 139-49.

[67] WANG H, HU Z, GUO Y, et al. A real-time safety helmet wearing detection approach based on CSYOLOv3 [J]. Applied Sciences, 2020, 10(19): 6732.

[68] BOCHKOVSKIY A, WANG C-Y, LIAO H-Y M. Yolov4: Optimal speed and accuracy of object detection [J]. arXiv preprint arXiv:200410934, 2020.

[69] YUNYUN L, JIANG W. Detection of wearing safety helmet for workers based on YOLOv4; proceedings of the 2021 International Conference on Computer Engineering and Artificial Intelligence (ICCEAI), F, 2021 [C]. IEEE.

[70] 吴冬梅, 闫宗亮, 宋婉莹, 等. 改进YOLOv4的安全帽佩戴检测及身份识别方法 [J]. 计算机仿真, 2022, 39(12): 290-3+377.

[71] 杨永波, 李栋. 改进 YOLOv5 的轻量级安全帽佩戴检测算法 [J]. 计算机工程与应用, 2022, 58(9): 201-7.

[72] 徐先峰, 赵万福, 邹浩泉, et al. 基于 MobileNet-SSD 的安全帽佩戴检测算法 [J]. 计算机工程, 2021, 47(10): 298-305.

[73] 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: 102894.

[74] 杨天云. 基于卷积神经网络的多尺度目标检测研究 [D]. 华中科技大学, 2017.

[75] 梅莹,尹艺璐,石称华,等.基于改进VGG卷积神经网络的叶菜霜霉病智能识别算法研究[J].上海蔬菜,2021,(06):76-84.

[76] 孙志琳. 基于深度学习的行人再识别研究 [D]. 山西大学, 2019.

[77] MARTINS A F T, ASTUDILLO R F. From Softmax to Sparsemax: A Sparse Model of Attention and Multi-Label Classification [J]. JMLRorg, 2016.

[78] HAIBIN L, YUAN S, WENMING Z, et al. The detection method for coal dust caused by chute discharge based on YOLOv4-tiny [J]. Opto-Electronic Engineering, 2021, 48(6): 210049-1-14.

[79] LIN T-Y, MAIRE M, BELONGIE S, et al. Microsoft coco: Common objects in context; proceedings of the Computer Vision–ECCV 2014: 13th European Conference, Zurich, Switzerland, September 6-12, 2014, Proceedings, Part V 13, F, 2014 [C]. Springer.

[80] PASZKE A, GROSS S, MASSA F, et al. PyTorch: An Imperative Style, High-Performance Deep Learning Library [J]. 2019.

[81] 于佃海, 吴甜. 深度学习技术和平台发展综述 [J]. 人工智能, 2020, (3): 12.

[82] QIN Z, YAN W Q. Traffic-sign recognition using deep learning; proceedings of the International Symposium on Geometry and Vision, F, 2021 [C]. Springer.

[83] WANG W, WU B, YANG S, et al. Road damage detection and classification with faster R-CNN; proceedings of the 2018 IEEE international conference on big data (Big data), F, 2018 [C]. IEEE.

[84] LOPEZ-ANTEQUERA M, GOMEZ-OJEDA R, PETKOV N, et al. Appearance-invariant place recognition by discriminatively training a convolutional neural network [J]. Pattern Recognition Letters, 2017, 92: 89-95.

[85] 周云成,许童羽,郑伟,等.基于深度卷积神经网络的番茄主要器官分类识别方法[J].农业工程学报,2017,33(15):219-226.

[86] FENG L, PO L-M, LI Y, et al. Integration of image quality and motion cues for face anti-spoofing: A neural network approach [J]. Journal of Visual Communication and Image Representation, 2016, 38: 451-460.

[87] ULLAH M B. CPU Based YOLO: A Real Time Object Detection Algorithm; proceedings of the 2020 IEEE Region 10 Symposium (TENSYMP), F, 2020 [C].

[88] GUO M H, XU T X, LIU J J, et al. Attention mechanisms in computer vision:A survey [J]. Computational visual media, 2022, 8(3): 38.

[89] HU J, SHEN L, SUN G. Squeeze-and-excitation networks; proceedings of the Proceedings of the IEEE conference on computer vision and pattern recognition, F, 2018 [C].

[90] WANG Q, WU B, ZHU P, et al. ECA-Net: Efficient channel attention for deep convolutional neural networks; proceedings of the Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, F, 2020 [C].

[91] 王素珍, 许浩, 邵明伟, 等. 基于改进YOLOv4-Tiny算法的绝缘子缺陷检测 [J] 国外电子测量技术 [J]. 2022, 41(09): 155-162.