轻量化高精度卷积神经网络的安全帽识别方法

doi:10.3778/j.issn.1002-8331.2006-0354

摘要/Abstract

摘要：

由于施工环境的复杂性，基于机器视觉的安全帽识别方法常常出现误检与漏检的情况。为提高复杂环境下安全帽识别的准确率，同时满足实时性要求，提出一种基于视觉感受野特性的轻量化高精度卷积神经网络。该卷积神经网络以RFBnet网络为基础，增加特征金字塔网络模块，使神经网络同时兼顾浅层语义信息和深层语义信息的表示能力，以实现复杂施工环境下不同形态与大小安全帽的识别。采用SE-Ghost模块在保持网络特征提取能力不变的情况下，对主干网络结构进行轻量化。为验证方法的性能，将基于感受野特性的轻量化卷积神经网络和当前主要卷积神经网络进行实验对比，结果表明，所提网络模型的检测准确率较YOLO-v3、RFBnet-300和RFBnet-512网络分别提高了1.60个百分点、3.62个百分点和0.98个百分点，检测速度达到20?frame/s。

关键词: 安全帽检测, 卷积神经网络, 感受野结构, 特征金字塔

Abstract:

Due to the complexity of construction environment, machine vision based helmet recognition methods often have false detections and missed detections. In order to improve the accuracy of helmet recognition in complex environments and satisfy the real-time requirement, this paper proposes a lightweight and high-precision convolutional neural network based on the characteristics of the receptive field. The convolutional neural network is based on the RFBnet network, and a feature pyramid network module is embedded to the neural network to represent the shallow semantic information and the deep semantic information simultaneously, so as to improve the accuracy of helmet recognition with different shapes and sizes in complex construction environments. SE-Ghost module is used to lighten the network backbone and reduce the complexity of the network without the network feature extraction capability changed. In the experiment, the proposed convolutional neural network based on receptive field is compared with current main convolutional neural networks, the results show that the proposed network has higher accuracy than YOLO-v3, RFBnet-300 and RFBnet-512 networks, and increases 1.60 percentage points, 3.62 percentage points and 0.98 percentage points respectively, while the detection speed reaches 20 frame/s.

Key words: helmet detection, convolutional neural network, receptive field structure, feature pyramid

陈柳，陈明举，薛智爽，罗仕胜. 轻量化高精度卷积神经网络的安全帽识别方法[J]. 计算机工程与应用, 2021, 57(22): 177-181.

CHEN Liu, CHEN Mingju, XUE Zhishuang, LUO Shisheng. Lightweight and High-Precision Convolutional Neural Network for Helmet Recognition Method[J]. Computer Engineering and Applications, 2021, 57(22): 177-181.

参考文献

[1] PARK M W，ELSAFTY N，ZHU Z.Hardhat-wearing detection for enhancing on-site safety of construction workers[J].Journal of Construction Engineering & Management，2015，141（9）：04015024.
[2] 李琪瑞.基于人体识别的安全帽视频检测系统研究与实现[D].成都：电子科技大学，2017.
LI Q R.A research and implementation of safety-helmet video detection system based on human body recognition[D].Chengdu：University of Electronic Science and Technology of China，2017.
[3] 刘晓慧，叶西宁.肤色检测和Hu矩在安全帽识别中的应用[J].华东理工大学学报（自然科学版），2014，40（3）：365-370.
LIU X H，YE X N.Skin color detection and Hu moments in helmet recognition research[J].Journal of East China University of Science and Technology（Natural Science Edition），2014，40（3）：365-370.
[4] 王生生，王萌，王光耀.基于深度神经网络剪枝的两阶段遥感图像目标检测[J].东北大学学报（自然科学版），2019，40（2）：174-179.
WANG S S，WANG M，WANG G Y.Deep neural network pruning based two-stage remote sensing image object detection[J].Journal of Northeastern University（Natural Science），2019，40（2）：174-179.
[5] 刘俊明，孟卫华.基于深度学习的单阶段目标检测算法研究综述[J].航空兵器，2020，27（3）：44-53.
LIU J M，MENG W H.Research on single-stage object detection algorithm based on deep learning[J].Aero Weaponry，2020，27（3）：44-53.
[6] 黄泽桑.基于深度学习的目标检测技术研究[D].北京：北京邮电大学，2019.
HUANG Z S.Object detection based on deep learning[D].Beijing：Beijing University of Posts and Teleconmunications，2019.
[7] LIU S，HUANG D，WANG Y.Receptive field block net for accurate and fast object detection[C]//15th European Conference on Computer Vision，2018：404-419.
[8] REDMON J，FARHADI A.YOLOv3：an incremental improvement[C]//2018 IEEE Conference on Computer Vision and Pattern Recognition，2018：89-95.
[9] SIMONYAN K，ZISSERMAN A.Very deep convolutional networks for large-scale image recognition[C]//International Conference on Learning Representations，May 2015.
[10] CHEN L C，PAPANDREOU G，KOKKINOS I，et al.Semantic image segmentation with deep convolutional nets and fully connected CRFs[J].Computer Science，2014（4）：357-361.
[11] LIN T Y，DOLLAR P，GIRSHICK R，et al.Feature pyramid networks for object detection[C]//2017 IEEE Conference on Computer Vision and Pattern Recognition，Honolulu，2017：2117-2125.
[12] HU J，SHEN L，AIBANIE S，et al.Squeeze-and-excitation networks[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition，Salt Lake City，2018：7132-7141.
[13] HOWARD A，SANDLER M，CHU G，et al.Searching for MobileNetv3[C]//2019 IEEE/CVF International Conference on Computer Vision，Seoul，2019：1314-1324.
[14] ZHANG X，ZHOU X，LIN M，et al.ShuffleNet：an extremely efficient convolutional neural network for mobile devices[C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition，Salt Lake City，2018：6848-6856.
[15] HAN K，WANG Y，TIAN Q，et al.GhostNet：more features from cheap operations[C]//2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition，Seattle，2020：1577-1586.