基于安全帽佩戴检测的矿山人员违规行为研究

doi:10.16180/j.cnki.issn1007-7820.2019.09.006

电子科技 ›› 2019, Vol. 32 ›› Issue (9): 26-31.doi: 10.16180/j.cnki.issn1007-7820.2019.09.006

基于安全帽佩戴检测的矿山人员违规行为研究

仝泽友^1,^2,³,冯仕民^1,²,侯晓晴^1,^2,³,丁恩杰^1,²

1.中国矿业大学物联网(感知矿山)研究中心,江苏徐州 221000
2.矿山互联网应用技术国家地方联合工程实验室,江苏徐州 221000
3.中国矿业大学信息与控制工程学院,江苏徐州 221000)

收稿日期:2018-09-06 出版日期:2019-09-15 发布日期:2019-09-19
作者简介:仝泽友(1994-),男,硕士研究生。研究方向:人体行为识别,计算机视觉等。|冯仕民(1983-),男,博士,讲师。研究方向:多传感器智能信息处理、融合与人机交互。
基金资助:
2017年国家重点研发计划项目(2017YFC0804401)

Recognition of Underground Miners’ Rule-Violated Behavior Based on Safety Helmet Detection

TONG Zeyou^1,^2,³,FENG Shimin^1,²,HOU Xiaoqing^1,^2,³,DING Enjie^1,²

1.Internet of Things (Perception Mine) Research Center,China University of Mining and Technology,Xuzhou 221000,China
2.The National and Local Joint Engineering Laboratory of Internet Application Technology on Mine,Xuzhou 221000,China
3.Institute of Information and Control Engineering,China University of Mining and Technology,Xuzhou 221000,China

Received:2018-09-06 Online:2019-09-15 Published:2019-09-19
Supported by:
2017 National Key R&D Projects(2017YFC0804401)

摘要/Abstract

摘要：

针对矿山人员安全帽佩戴检测问题,文中提出了一种基于人脸的身份识别及安全帽佩戴检测的违规行为识别方法。首先在视频图像中检测人脸以识别身份,然后运用卷积神经网络方法检测人员是否佩戴安全帽,实验阶段将此方法与传统的图像处理方法进行测试对比。实验结果显示,基于深度学习的安全帽检测方法的鲁棒性强于传统方法,在不同条件下识别率和运行效率均优于传统方法,深度学习方法的平均识别率高达97%,所需平均运行时间少于传统方法的1/7。

关键词: 安全帽检测, 身份识别, 违规行为, 深度学习, 准确率与速度, 平均时间

Abstract:

In order to solve the problem of the miner's helmet-wearing recognition, this paper proposed a method of detecting the safety helmet based rule-violated activity. It was based on the face recognition and helmet detection. Firstly, detect the human faced in the picture and identify them, and then detect whether the person wears a safety helmet or not with the Convolutional Neural Network approach. This paper compared the depth learning method with the traditional method in different conditions in the experiments.The experimental results showed that the robustness of the depth learning based safety helmet detection method was stronger than the traditional method. The recognition rate of the depth learning method was higher than that of the traditional approach under different conditions. The deep learning method achieved an average recognition rate of 97% and the average running time was less than 1/7 of that of the traditional approach.

Key words: safety helmet detection, identity recognition, rule-violated activity, deep learning, accuracy and efficiency, average time

中图分类号:

TN399

仝泽友,冯仕民,侯晓晴,丁恩杰. 基于安全帽佩戴检测的矿山人员违规行为研究[J]. 电子科技, 2019, 32(9): 26-31.

TONG Zeyou,FENG Shimin,HOU Xiaoqing,DING Enjie. Recognition of Underground Miners’ Rule-Violated Behavior Based on Safety Helmet Detection[J]. Electronic Science and Technology, 2019, 32(9): 26-31.

图/表 11

表1

图1

图2

图3

图4

表2

图5

表3

图6

表4

图7

参考文献 17

[1]	Liu P, Han S, Meng Z, et al. Facial expression recognition via a boosted deep belif network [C].Washington D C:Proceedlings of the IEEE Conference on Computer Vision and Pattern Recognition,IEEE, 2014.
[2]	熊海朋, 陈洋洋, 陈春玮 . 基于卷积神经网络的场景图像文本定位研究[J]. 电子科技, 2018,31(1):50-53.
	Xiong Haipeng, Chen Yangyang, Chen Chunwei . Text location in image based on convolution neual network[J]. Electronic Science and Technology, 2018,31(1):50-53.
[3]	Rafael C. Gonzalez digital image processing[M]. Beijing: Publishing House of Electronics Industry, 2017.
[4]	完颜勇 . 基于对比度增强和背景估计的文档图像二值化[J]. 电子科技, 2018,31(4):20-25.
	Wanyan Yong . Doucument image binarization based on contrast enhancement and background estimation[J]. Electronic Science and Technology, 2018,31(4):20-25.
[5]	Zhang Z, Luo P, Loy C C, et al. Facial landmark detection bydeep multi-task learning [C].Berlin:Proceedings of the European Conference on Computer Vision,Springer, 2014.
[6]	朱煜, 赵江坤, 王逸宁 , 等. 基于深度学习的人体行为识别算法综述[J]. 自动化学报, 2016,42(6):848-875. doi: 10.16383/j.aas.2016.c150710
	Zhu Yu, Zhao Jiangkun, Wang Yining , et al. A review of human action recognition based on deep learning[J]. Acta Automatica Sinica, 2016,42(6):848-875. doi: 10.16383/j.aas.2016.c150710
[7]	Fischer A, Igel C . Training restricted boltzmann machines:an introduction[J]. Pattern Recognition, 2014,47(1):25-39.
[8]	Christian Szegedy. Vincent vanhoucke rethinking the inception architecture for computer vision [C].Boston:IEEE Conference on Computer Vision and Pattern Recognition, 2016.
[9]	Ding Changxing, Tao Dacheng . Trunk-branch ensemble convolutional neural networks for video-based face recognition[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2018,40(4):89-95.
[10]	Schmidhuber J . Deep learning in neural networks: an overview[J]. Neural Networks, 2015,61(7):85-117.
[11]	龚爱平 . 基于嵌入式机器视觉的信息采集与处理技术研究[D]. 杭州:浙江大学, 2013.
	Gong Aiping . Study of the information acquisition and processing technology based on embaedded computer vision[D]. Hangzhou:Zhejiang University, 2013.
[12]	Joe Minichino. Learning OpenCV3 computer vision with python[M]. Paris: Packet Publishing, 2015.
[13]	毕林, 谢伟, 崔君 . 基于卷积神经网络的矿工安全帽佩戴识别研究[J]. 黄金科学技术, 2017,25(4):73-80.
	Bi Lin, Xie Wei, Cui Jun . Identification research on the miner’s safety helmet wear based on convolutional neural network[J]. Gold Science and Technology, 2017,25(4):73-80.
[14]	黄文坚, 唐源 . Tensorflow实战[M]. 北京: 电子工业出版社, 2017.
	Huang Wenjian, Tang Yuan. Practice of tensorflow[M]. Beijing: Publishing House of Electronics Industry, 2017.
[15]	Jose M Chaquet, Enrique J Carmona . A survey of video datasets for human actions and activity recognition[J]. Computer Vision and Image Understanding, 2013(9):633-659.
[16]	Alejandro Newell, Yang Kaiyu, Deng Jia. Stacked hourglass networks for human pose estimation [C].Guangzhou:ECCV, 2016.
[17]	Chen X, Yuille A. Articulated pose estimation by agraphical model with image dependent pairwise relations [C].Shanghai:NIPS, 2014.

类型	kernel尺寸/步长	输入尺寸
卷积	3×3/2	416×416×3
池化	3×3/1	207×207×32
卷积	3×3/1	103×103×64
池化	3×3/1	101×101×64
卷积	3×3/2	99×99×64
池化	3×3/1	45×45×64
线性logits	1×1×64
softmax	分类输出	1×1×10

实验次数	传统方法		深度学习方法
实验次数	50	100	50	100
安全帽正面	94%	95%	98%	97%
安全帽侧面	12%	10%	96.6%	98%
光线弱(200 lux)	92%	90%	98%	97%
光线强(3 000 lux)	89%	98%	98%
距离2 m	96%	94%	96%	97%
距离4 m	26%	24%	94%	96%

实验次数	传统方法		深度学习方法
实验次数	50	100	50	100
正常光照条件	92%	90%	86%	85%
光线弱(200 lux)	42%	39%	25%	30%
光线强(3 000 lux)	56%	60%	24%	20%
距离2 m	94%	91%	84%	86%
距离4 m	20%	23%	62%	65%

次数方法	50	100
传统方法	12 338 ms	12 260 ms
深度学习方法	1 703 ms	1 681 ms

基于安全帽佩戴检测的矿山人员违规行为研究

Recognition of Underground Miners’ Rule-Violated Behavior Based on Safety Helmet Detection

RichHTML

PDF (PC)

赞

可视化

摘要/Abstract

引用本文

使用本文

图/表 11

参考文献 17

相关文章 15

Metrics

本文评价

推荐阅读 8

[1]	程晓雅,张雷. 基于深度学习的有遮挡人脸识别方法研究[J]. 电子科技, 2022, 35(1): 35-39.
[2]	赵崇,迟蒙蒙,储聪,张鹏. 导盲犬行走机构运动仿真及其视觉识别算法研究[J]. 电子科技, 2021, 34(9): 66-72.
[3]	马立新,豆晨飞,宋晨灿,杨天笑. 基于特征融合卷积神经网络的绝缘子无损检测[J]. 电子科技, 2021, 34(7): 26-30.
[4]	薛永杰,巨志勇. 基于改进AlexNet的鱼类识别算法[J]. 电子科技, 2021, 34(4): 12-17.
[5]	邵杭,王永雄,秦宇龙. 基于深度学习与显著性的数字图像构图优化[J]. 电子科技, 2021, 34(3): 36-42.
[6]	闫超,孙占全,田恩刚,赵杨洋,范小燕. 基于深度学习的医学图像分割技术研究进展[J]. 电子科技, 2021, 34(2): 7-11.
[7]	程俊华,曾国辉,刘瑾. 基于深度学习的复杂背景图像分类方法研究[J]. 电子科技, 2020, 33(12): 59-66.
[8]	邬倩,吴飞,骆立志. 基于深度学习的人体骨架动作识别[J]. 电子科技, 2020, 33(11): 79-83.
[9]	段雨佳,鞠婷. 基于深度学习的代码审查意见有效性评估[J]. 电子科技, 2020, 33(1): 39-45.
[10]	胡少聪. 基于深度学习的人脸识别方法研究[J]. 电子科技, 2019, 32(6): 82-86.
[11]	姜猛,王子牛,高建瓴. 基于异构数据联合训练的中文分词法[J]. 电子科技, 2019, 32(4): 29-33.
[12]	李荣瑞,施霖,赵薇. 基于卷积神经网络的少数民族头饰识别[J]. 电子科技, 2019, 32(2): 51-55.
[13]	孟昕. 基于深度学习的法律文书识别方法研究[J]. 电子科技, 2019, 32(12): 84-86.
[14]	袁小平,王岗,王晔枫,汪喆远,孙辉. 基于改进卷积神经网络的交通标志识别方法[J]. 电子科技, 2019, 32(11): 28-32.
[15]	史册,裘炅. 基于深度学习机制的火警协议智能分析模型[J]. 电子科技, 2019, 32(1): 80-85.