基于卷积神经网络的安全标识分类算法研究

doi:10.16265/j.cnki.issn1003-3033.2023.S1.0070

摘要/Abstract

摘要：

为解决安全标识数据集、安全标识特殊图形及复杂背景缺乏等问题,采用卷积神经网络(CNN)提取安全标识的特征,在VGG-16网络结构和CNN的基础上构建能够识别17种安全标识的VGG16-17模型。原始数据有816张,通过数据增强扩展数据集,得到4 708张图片,按照4∶1的比例将数据集划分为训练集和验证集。通过调节模型中部分参数,分析迭代次数和批量大小对模型识别分类效果的影响。结果表明:当迭代次数为20次、批量大小为32时,模型结果最理想,识别准确率为97.92%,相较于基于未经过数据增强数据集的改进模型的准确率提高19.39%,同时,改进模型相较于传统VGG16模型,识别准确率提高4.3%,证明模型改进和数据增强对图像识别能力的提高有一定帮助。

关键词: 卷积神经网络(CNN), 安全标识, 计算机视觉, 图像识别, 数据增强

Abstract:

In order to solve the problems of lack of safety sign data sets, special safety sign graphics, and complex backgrounds, CNN was used to extract the characteristics of safety signs. Based on the VGG-16 network structure and CNN, a VGG16-17 model that could identify 17 kinds of safety signs was constructed. There were 816 original data, and 4 708 images were obtained by expanding the data set through data augmentation. The data set was divided into a training set and a validation set according to the ratio of 4∶1. By adjusting some parameters in the model, the influence of epoch times and batch size on the recognition and classification effect of the model was analyzed. The results show that when the epoch is 20, and the batch size is 32, the model results are the best, and the recognition accuracy is 97.92%. Compared with the improved model based on the data set with no data augmentation, the accuracy is increased by 19.39%. Meanwhile, compared with the traditional VGG16 model, the recognition accuracy of the improved model is improved by 4.3%. This demonstrates that model improvement and data augmentation contribute to the improvement of image recognition.

Key words: convolution neural network (CNN), safety signs, computer vision, image recognition, data augmentation

王瑶涵, 宋泽阳, 张利冬. 基于卷积神经网络的安全标识分类算法研究[J]. 中国安全科学学报, 2023, 33(S1): 263-269.

WANG Yaohan, SONG Zeyang, ZHANG Lidong. Research on safety sign classification based on CNN[J]. China Safety Science Journal, 2023, 33(S1): 263-269.

图/表 8

图1

表1

图2

图3

图4

图5

图6

图7

参考文献 25

[1]	LIU Xiaoxu, YAN Weiqi, KASABOV N. Vehicle-related scene segmentation using CapsNets[C]. 35^th International Conference on Image and Vision Computing New Zealand (IVCNZ), 2020: 1-6.
[2]	LIU Xiaoxu, NEUYEN M, YAN Weiqi. Vehicle-related scene understanding using deep learning[C]. Asian Conference on Pattern Recognition, 2019: 61-73.
[3]	RUBAIYAT A H, TOMA T T, KALANTARI-KHANDANI M, et al. Automatic detection of helmet uses for construction safety[C]. IEEE/WIC/ACM International Conference on Web Intelligence Workshops (WIW), 2016: 135-142.
[4]	WU Hao, ZHAO Jinsong. An intelligent vision-based approach for helmet identification for work safety[J]. Computers in Industry, 2018, 100: 267-277. doi: 10.1016/j.compind.2018.03.037
[5]	SILVA R R V, AIRES K R T, VERAS R D M S. Helmet detection on motorcyclists using image descriptors and classifiers[C]. 27^th SIBGRAPI Conference on Graphics, Patterns and Images, 2014: 141-148.
[6]	张建明, 王伟, 陆朝铨, 等. 基于压缩卷积神经网络的交通标志分类算法[J]. 华中科技大学学报:自然科学版, 2019, 47(1): 103-108.
	ZHANG Jianming, WANG Wei, LU Chaoquan, et al. Traffic sign classification algorithm based on compressed convolutional neural network[J]. Journal of Huazhong University of Science and Technology: Natural Science Edition, 2019, 47(1): 103-108.
[7]	WU Linxiu, LI Houjie, HE Jianjun, et al. Traffic sign detection method based on Faster R-CNN[J]. Journal of Physics: Conference Series, 2019, 1713(3): DOI: 10.1088/1742-6596/1176/3/032045.
[8]	CHEHRI H, CHEHRI A, SAADANE R. Traffic signs detection and recognition system in snowy environment using deep learning[C]. The Proceedings of the Third International Conference on Smart City Applications, 2020: 503-513.
[9]	WANG Canyong. Research and application of traffic sign detection and recognition based on deep learning[C]. International Conference on Robots & Intelligent System (ICRIS), 2018: 150-152.
[10]	JIN Miao, CHEN Xiwen, LAI Guoshu, et al. Glove detection system based on VGG-16 network[C]. 13^th International Symposium on Computational Intelligence and Design (ISCID), 2020: 172-175.
[11]	汪洋, 王俊刚. 基于深度学习算法的铁路列车运行安全检测[J]. 中国安全科学学报, 2018, 28(增2): 41-45.
	WANG Yang, WANG Jungang. Study on safety inspection of railway train operation based on deep learning algorithm[J]. China Safety Science Journal, 2018, 28(S2): 41-45. doi: 10.16265/j.cnki.issn1003-3033.2018.S2.008
[12]	ZHOU Yanqing, XUE Heru, JIANG Xinhua, et al. Low-resolution safety helmet image recognition combining local binary pattern with statistical features[J]. Comput Syst Appl, 2015, 24(7): 211-215.
[13]	MNEYMNEH B E, ABBAS M, KHOURY H. Automated hardhat detection for construction safety applications[J]. Procedia engineering, 2017, 196: 895-902. doi: 10.1016/j.proeng.2017.08.022
[14]	PARK C, LEE D, KHAN N. An analysis on safety risk judgment patterns towards computer vision based construction safety management[C]. Creative Construction e-Conference, 2020: 31-38.
[15]	HOUBEN S, STALLKAMP J, SALMEN J, et al. Detection of traffic signs in real-world images: the German traffic sign detection benchmark[C]. International Joint Conference on Neural Networks, 2013:1-8.
[16]	GREENHALGH J, MIRMEHDI M. Real-time detection and recognition of road traffic signs[J]. IEEE Transactions on Intelligent Transportation Systems, 2012, 13(4): 1 498-1 506. doi: 10.1109/TITS.2012.2208909
[17]	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, 10(1): 11-29.
[18]	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(6): 76-84.
	MEI Ying, YIN Yilu, SHI Chenghua, et al. Research on intelligent recognition algorithm of leafy vegetable downy mildew based on improved VGG convolutional neural network[J]. Shanghai Vegetables, 2021(6): 76-84.
[19]	QIN Zhongbing, YAN Weiqi. Traffic-sign recognition using deep learning[C]. International Symposium on Geometry and Vision, 2021: 13-25.
[20]	WANG Wenzhe, WU Bin, YANG Sixiong, et al. Road damage detection and classification with faster R-CNN[C]. IEEE International Conference on Big data (Big data), 2018:5 220-5 223.
[21]	孙志琳. 基于深度学习的行人再识别研究[D]. 太原: 山西大学, 2019.
	SUN Zhilin. Research on person re-identification based on deep learning[D]. Taiyuan: Shanxi University, 2019.
[22]	张建华, 孔繁涛, 吴建寨, 等. 基于改进 VGG 卷积神经网络的棉花病害识别模型[J]. 中国农业大学学报, 2018, 23(11): 161-171.
	ZHANG Jianhua, KONG Fantao, WU Jianzhai, et al. Cotton disease identification model based on improved VGG convolution neural network[J]. Journal of China Agricultural University, 2018, 23(11): 161-171.
[23]	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. doi: 10.1016/j.patrec.2017.04.017
[24]	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.
	ZHOU Yuncheng, XU Tongyu, ZHENG Wei, et al. Classification and recognition approaches of tomato main organs based on DCNN[J]. Transactions of the Chinese Society of Agricultural Engineering, 2017, 33(15): 219-226.
[25]	FENG Litong, PO L-M, LI Yuming, 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. doi: 10.1016/j.jvcir.2016.03.019

大类	类别	备注	数量	增强后数量
禁止类	禁止吸烟	A	46	260
禁止类	禁止攀登	B	54	303
禁止类	禁止烟火	C	49	268
禁止类	禁止入内	D	55	308
警告类	当心触电	E	37	314
警告类	注意安全	F	45	298
警告类	当心机械伤人	G	45	254
警告类	当心中毒	H	46	269
警告类	当心坠落	I	52	274
警告类	当心滑倒	J	39	271
指令类	必须佩戴安全帽	K	69	255
指令类	必须戴防护手套	L	38	259
指令类	必须穿防护服	M	40	278
指令类	必须佩戴防毒面具	N	50	287
指令类	必须佩戴防尘口罩	O	35	256
提示类	应急避难场所	P	45	258
提示类	紧急出口	Q	71	296
总计			816	4 708