基于改进YOLOv4的小目标行人检测算法

doi:10.19678/j.issn.1000-3428.0063623

计算机工程 ›› 2023, Vol. 49 ›› Issue (2): 296-302,313. doi: 10.19678/j.issn.1000-3428.0063623

基于改进YOLOv4的小目标行人检测算法

王程^1,2, 刘元盛^1,2, 刘圣杰^1,2

1. 北京联合大学北京市信息服务工程重点实验室, 北京 100101;
2. 北京联合大学机器人学院, 北京 100101

收稿日期:2021-12-27 修回日期:2022-01-28 发布日期:2022-05-02
作者简介:王程(1999-),女,硕士研究生,主研方向为无人驾驶技术、计算机视觉、数字图像处理;刘元盛(通信作者),教授、博士生导师;刘圣杰,硕士研究生。
基金资助:
国家自然科学基金“无人车多视视频信息获取与定位关键技术”（61871038）；国家自然科学基金“基于视觉计算的智能驾驶实时城市道路场景理解”（61871039）；北京联合大学研究生科研创新项目（YZ2020K001）；北京联合大学人才强校优选-拔尖计划“无人驾驶车复杂场景中可靠性定位技术研究”（BPHR2020BZ01）。

Small-Target Pedestrian-Detection Algorithm Based on Improved YOLOv4

WANG Cheng^1,2, LIU Yuansheng^1,2, LIU Shengjie^1,2

1. Beijing Key Laboratory of Information Service Engineering, Beijing Union University, Beijing 100101, China;
2. College of Robotics, Beijing Union University, Beijing 100101, China

Received:2021-12-27 Revised:2022-01-28 Published:2022-05-02

摘要/Abstract

摘要： 行人检测在无人驾驶环境感知领域具有重要应用。现有行人检测算法多数只关注普通大小的行人目标，忽略了小目标行人特征信息过少的问题，从而造成检测精度低、应用于嵌入式设备中实时性不高等情况。针对该问题，提出一种小目标行人检测算法YOLOv4-DBF。引用深度可分离卷积代替YOLOv4算法中的传统卷积，以降低模型的参数量和计算量，提升检测速度和算法实时性。在YOLOv4骨干网络中的特征融合部分引入scSE注意力模块，对输入行人特征图的重要通道和空间特征进行增强，促使网络学习更有意义的特征信息。对YOLOv4颈部中特征金字塔网络的特征融合部分进行改进，在增加少量计算量的情况下增强对图像中行人目标的多尺度特征学习，从而提高检测精度。在VOC07+12+COCO数据集上进行训练和验证，结果表明，相比原YOLOv4算法，YOLOv4-DBF算法的AP值提高4.16个百分点，速度提升27%，将该算法加速部署在无人车中的TX2设备上进行实时测试，其检测速度达到23FPS，能够有效提高小目标行人检测的精度及实时性。

关键词: 无人驾驶, 小目标行人, 深度可分离卷积, scSE注意力模块, 特征金字塔网络

Abstract: Pedestrian detection is vital to applications in unmanned environment perception.Most existing pedestrian-detection algorithms focus only on ordinary pedestrian targets and do not consider the low accuracy caused by the insufficient pedestrian feature information of small targets;furthermore, they do not offer favorable real-time performance when applied to embedded devices.Hence, a small-target pedestrian-detection algorithm, YOLOv4-DBF, is proposed herein.The conventional convolution is replaced with deeply separable convolution in the YOLOv4 algorithm, which reduces the number of parameters and the computation time of the model, as well as improves the detection speed and real-time performance of the algorithm.Additionally, the concurrent spatial and channel Squeeze & Excitation(scSE) attention module is introduced into the feature fusion component of the YOLOv4 backbone network to enhance the important channels and spatial features of the input pedestrian feature map as well as to enable the network to learn more meaningful feature information.The feature fusion component of the Feature Pyramid Network(FPN) in the YOLOv4 neck is improved to enhance the multiscale feature learning of the pedestrian target in the image, which improves the detection accuracy but increases the amount of computation.After training and verification based on the VOC07+12+COCO dataset, the results show that compared with the original YOLOv4 algorithm, YOLOv4-DBF increases the Average Precision(AP) by 4.16 percentage points and the speed by 27%.Finally, YOLOv4-DBF is accelerate deployed on the TX2 equipment of an unmanned vehicle for real-time testing, where the maximum speed reaches 23FPS.The algorithm proposed herein can effectively improve the accuracy and real-time performance of small-target pedestrian detection.

Key words: driverless vehicle, small-target pedestrian, deeply separable convolution, scSE attention module, Feature Pyramid Network(FPN)

中图分类号:

TP391

王程, 刘元盛, 刘圣杰. 基于改进YOLOv4的小目标行人检测算法[J]. 计算机工程, 2023, 49(2): 296-302,313.

WANG Cheng, LIU Yuansheng, LIU Shengjie. Small-Target Pedestrian-Detection Algorithm Based on Improved YOLOv4[J]. Computer Engineering, 2023, 49(2): 296-302,313.

http://www.ecice06.com/CN/Y2023/V49/I2/296

图/表 13

20230216183008

20230216183022

20230216183025

20230216183029

20230216183033

20230216183037

20230216183042

20230216183046

20230216183050

20230216183053

20230216183057

20230216183100

20230216183103

参考文献

[1] MUKHERJEE A, ADARSH S, RAMACHANDRAN K I.ROS-based pedestrian detection and distance estimation algorithm using stereo vision, leddar and CNN[EB/OL].[2021-11-05].https://link.springer.com/chapter/10.1007/978-981-15-5400-1_13.
[2] HAO Y T.Research on multi-feature and machine learning hierarchical pedestrian detection method based on deep learning[J].Journal of Physics:Conference Series, 2021, 1748(2):022001.
[3] ANSARI M F, LODI K A.A survey of recent trends in two-stage object detection methods[J].Renewable Power for Sustainable Growth, 2021, 723:669-677.
[4] ZHANG Y F, LI X, WANG F Y, et al.A comprehensive review of one-stage networks for object detection[C]//Proceedings of IEEE International Conference on Signal Processing, Communications and Computing.Washington D.C., USA:IEEE Press, 2021:1-6.
[5] REN S Q, HE K M, 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.
[6] HE K M, GKIOXARI G, DOLLÁR P, et al.Mask R-CNN[C]//Proceedings of IEEE International Conference on Computer Vision.Washington D.C., USA:IEEE Press, 2017:2980-2988.
[7] SHAO X Q, WEI J Y, GUO D F, et al.Pedestrian detection algorithm based on improved Faster RCNN[C]//Proceedings of IEEE Advanced Information Technology, Electronic and Automation Control Conference.Washington D.C., USA:IEEE Press, 2021:1368-1372.
[8] LAI K C, ZHAO J, LIU D J, et al.Research on pedestrian detection using optimized Mask R-CNN algorithm in low-light road environment[J].Journal of Physics:Conference Series, 2021, 1777(1):012057.
[9] 音松, 陈雪云, 贝学宇.改进Mask RCNN算法及其在行人实例分割中的应用[J].计算机工程, 2021, 47(6):271-276, 283. YIN S, CHEN X Y, BEI X Y.Improved Mask RCNN algorithm and its application in pedestrian instance segmentation[J].Computer Engineering, 2021, 47(6):271-276, 283.(in Chinese)
[10] LIU W, ANGUELOV D, ERHAN D, et al.SSD:single shot multibox detector[C]//Proceedings of European Conference on Computer Vision.Berlin, Germany:Springer, 2016:21-37.
[11] BOCHKOVSKIY A, WANG C Y, LIAO H Y M.YOLOv4:optimal speed and accuracy of object detection[EB/OL].[2021-11-05].https://arxiv.org/pdf/2004.10934.pdf.
[12] DONG C, LUO X S.Research on a pedestrian detection algorithm based on improved SSD network[J].Journal of Physics:Conference Series, 2021, 1802(3):032073.
[13] WEN B Y, WU M Q.Study on pedestrian detection based on an improved YOLOv4 algorithm[C]//Proceedings of IEEE International Conference on Computer and Communications.Washington D.C., USA:IEEE Press, 2020:1198-1202.
[14] CAO Z, YANG H, ZHAO J, et al.Attention fusion for one-stage multispectral pedestrian detection[J].Sensors(Basel, Switzerland), 2021, 21(12):4184.
[15] 黄凤琪, 陈明, 冯国富.基于可变形卷积的改进YOLO目标检测算法[J].计算机工程, 2021, 47(10):269-275, 282. HUANG F Q, CHEN M, FENG G F.Improved YOLO object detection algorithm based on deformable convolution[J].Computer Engineering, 2021, 47(10):269-275, 282.(in Chinese)
[16] EVERINGHAM M, GOOL L, WILLIAMS C K I, et al.The pascal Visual Object Classes(VOC) challenge[J].International Journal of Computer Vision, 2010, 88(2):303-338.
[17] LIN T Y, MAIRE M, BELONGIE S, et al.Microsoft coco:common objects in context[C]//Proceedings of European Conference on Computer Vision.Berlin, Germany:Springer, 2014:740-755.
[18] REDMON J, DIVVALA S, GIRSHICK R, et al.You only look once:unified, real-time object detection[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2016:779-788.
[19] REDMON J, FARHADI A.YOLO9000:better, faster, stronger[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2017:6517-6525.
[20] REDMON J, FARHADI A.YOLOv3:an incremental improvement[EB/OL].[2021-11-05].https://arxiv.org/abs/1804.02767.
[21] TSOTSOS J K.A computational perspective on visual attention[M].Cambridge, USA:MIT Press, 2011.
[22] HU J, SHEN L, SUN G.Squeeze-and-excitation networks[C]//Proceedings of IEEE/CVF Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2018:7132-7141.
[23] ROY A G, NAVAB N, WACHINGER C.Concurrent spatial and channel squeeze & excitation in fully convolutional networks[EB/OL].[2021-11-05].https://arxiv.org/pdf/1803.02579v1.pdf.
[24] HOWARD A G, ZHU M L, CHEN B, et al.MobileNets:efficient convolutional neural networks for mobile vision applications[EB/OL].[2021-11-05].https://arxiv.org/abs/1704.04861.
[25] LIN T Y, DOLLÁR P, GIRSHICK R, et al.Feature pyramid networks for object detection[C]//Proceedings of IEEE Conference on Computer Vision and Pattern Recognition.Washington D.C., USA:IEEE Press, 2017:936-944.

选择文件类型/文献管理软件名称

选择包含的内容

基于改进YOLOv4的小目标行人检测算法

Small-Target Pedestrian-Detection Algorithm Based on Improved YOLOv4

RichHTML

PDF

可视化

被引次数

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献

相关文章 15

编辑推荐

Metrics

本文评价

[1]	李松江, 耿兰兰, 王鹏. 基于改进Yolov4的车辆目标检测[J]. 计算机工程, 2023, 49(4): 272-280.
[2]	郭克友, 王苏东, 李雪, 张沫. 基于Dim env-YOLO算法的昏暗场景车辆多目标检测[J]. 计算机工程, 2023, 49(3): 312-320.
[3]	史宝岱, 张秦, 李瑶, 李宇环. 面向图像目标识别的轻量化卷积神经网络[J]. 计算机工程, 2022, 48(6): 257-262.
[4]	杜田, 李欣, 赖成喆, 郑东. 面向无人驾驶地图更新的安全信任管理方案[J]. 计算机工程, 2022, 48(6): 154-166.
[5]	柳聪, 屈丹, 司念文, 魏紫薇. 基于深度可分离卷积的轻量级图像超分辨率重建[J]. 计算机工程, 2022, 48(6): 228-234.
[6]	史钰祜, 张起贵. 基于局部注意的快速视频目标检测方法[J]. 计算机工程, 2022, 48(5): 314-320.
[7]	邹慧海, 侯进. 改进SSD算法的道路小目标检测研究[J]. 计算机工程, 2022, 48(5): 281-288.
[8]	徐增敏, 陈凯, 郭威伟, 赵汝文, 蒋占四. 面向轻量级卷积网络的激活函数与压缩模型[J]. 计算机工程, 2022, 48(5): 242-250.
[9]	候瑞环, 杨喜旺, 王智超, 高佳鑫. 一种基于YOLOv4-TIA的林业害虫实时检测方法[J]. 计算机工程, 2022, 48(4): 255-261.
[10]	宁小娟, 巩亮, 张金磊. 基于激光点云的道路可通行区域检测方法[J]. 计算机工程, 2022, 48(4): 22-29.
[11]	汪常建, 丁勇, 卢盼成. 融合改进FPN与关联网络的Faster R-CNN目标检测[J]. 计算机工程, 2022, 48(2): 173-179.
[12]	王军, 冯孙铖, 程勇. 深度学习的轻量化神经网络结构研究综述[J]. 计算机工程, 2021, 47(8): 1-13.
[13]	林杰, 陈春梅, 刘桂华, 祝礼佳. 室内服务机器人的实时场景分割算法[J]. 计算机工程, 2021, 47(7): 21-29.
[14]	音松, 陈雪云, 贝学宇. 改进Mask RCNN算法及其在行人实例分割中的应用[J]. 计算机工程, 2021, 47(6): 271-276,283.
[15]	胡江颢, 王红雨, 乔文超, 马靖煊. 基于轻量级网络的实时人体关键点检测算法[J]. 计算机工程, 2021, 47(4): 218-225.

模态框（Modal）标题

选择文件类型/文献管理软件名称

选择包含的内容

基于改进YOLOv4的小目标行人检测算法

Small-Target Pedestrian-Detection Algorithm Based on Improved YOLOv4

RichHTML

PDF

可视化

被引次数

摘要/Abstract

引用本文

使用本文

图/表 13

参考文献

相关文章 15

编辑推荐

Metrics

本文评价