PDF下载
[1]孙卉梅,刘路,王德刚,等.面向高后果区工程车辆视觉检测的YOLO-MMCE算法[J].油气储运,2024,43(09):1031-1038.[doi:10.6047/j.issn.1000-8241.2024.09.008]
　SUN Huimei,LIU Lu,WANG Degang,et al.YOLO-MMCE algorithm for visual detection of engineering vehicles in high-consequence area[J].Oil & Gas Storage and Transportation,2024,43(09):1031-1038.[doi:10.6047/j.issn.1000-8241.2024.09.008]
点击复制

面向高后果区工程车辆视觉检测的YOLO-MMCE算法

《油气储运》[ISSN:1000-8241/CN:13-1093/TE] 卷: 43 期数: 2024年09期 页码: 1031-1038 栏目: 检测与完整性 出版日期: 2024-09-25
Title:
YOLO-MMCE algorithm for visual detection of engineering vehicles in high-consequence area
文章编号:
1000-8241（2024）09-1031-08
作者:
孙卉梅1; 刘路1; 王德刚2; 王太勇3
1.天津职业技术师范大学机械工程学院；2.国家管网集团北方管道有限责任公司长沙输油气分公司；3.天津大学机械工程学院
Author(s):
SUN Huimei1; 3; LIU Lu1; 3; WANG Degang2; WANG Taiyong3
1.School of Mechanical Engineering, Tianjin University of Technology and Education; 2.Changsha Oil and Gas Transmission Branch, PipeChina North Pipeline Co. Ltd.; 3.School of Mechanical Engineering, Tianjin University
关键词:
YOLOv5; 工程车辆; 数据增强; 坐标注意力机制; 损失函数
Keywords:
YOLOv5;  engineering vehicle;  data augmentation;  coordinate attention (CA) mechanism;  loss function
分类号:
TE832
DOI:
10.6047/j.issn.1000-8241.2024.09.008
文献标志码:
A
摘要:
【目的】高后果区大型施工现场的工程车辆给埋地管道带来严重的安全隐患。针对当前常用方法在工程车辆重叠目标检测和日照变化场景下的目标检测方面存在漏检率高、检测精度低的问题，以挖掘机、装载机、压路机及重型货车4类常见工程车辆作为识别对象，提出了一种基于改进YOLOv5的工程车辆目标检测方法——YOLO-MMCE。【方法】采用Mosaic+Mixup结合的数据增强方式，增强对不同场景的适应能力，提高模型在实际复杂环境和模糊情况下的鲁棒性和泛化性。针对目标重叠和光照变化导致的特征不明显问题，在YOLOv5网络模型中引入坐标注意力（Coordinate Attention,CA）机制，增强网络模型的特征提取能力；为了提升预测边框回归精度，引入了高效率交并比（Efficient Intersection over Union,EIOU）函数，计算预测框与真实框的宽高差异值并取代纵横比，进一步提高算法检测精度。【结果】以兰郑长成品油管道高后果区监控摄像机获取的施工现场照片为数据集，对YOLO-MMCE算法进行验证。结果表明，对YOLOv5算法3个方面的改进均能提高其在实际工况下工程车辆目标检测的精度，总体平均精度均值（Mean Average Precision,mAP）达到84.8％，比原始YOLOv5算法提高了6.9％。对挖掘机、装载机、压路机及重型货车的目标检测mAP分别提高了4.4％、7.5％、9.5％、6.0％。【结论】YOLO-MMCE算法有效解决了重叠目标检测和日照变化场景下的工程车辆目标检测问题，具备实际应用价值。（图3，表5，参28）
Abstract:
[Objective] Engineering vehicles operating on large-scale construction sites in high-consequence areas pose severe safety hazards to buried pipelines. This paper addresses the shortcomings of current common techniques for detecting overlapped targets of engineering vehicles and target detection in scenarios with varying sunlight, highlighting issues such as high miss rates and low detection accuracies. The paper introduces a target detection method for engineering vehicles named YOLO-MMCE, which is based on an improved version of YOLOv5. This method focuses on recognizing four main types of engineering vehicles: excavators, loaders, rollers, and heavy trucks. [Methods] The Mosaic + Mixup combined data augmentation approach was adopted to improve the model’s adaptability to diverse scenarios and strengthen its robustness and generalization in intricate real-world settings and ambiguous conditions. In response to challenges pertaining to overlapping targets and inconspicuous features due to illumination variations, a coordinate attention (CA) mechanism was integrated into the YOLOv5 network model to amplify its feature extraction capacity. Additionally, to improve the regression accuracy of prediction borders, an Efficient Intersection over Union (EIOU) function was incorporated to calculate the width-height difference between prediction and real borders to replace the aspect ratio, thus further elevating the detection accuracy of the algorithm. [Results] The YOLO-MMCE algorithm was validated using the datasets comprised of construction site photos captured by surveillance cameras in the high-consequence areas along the Lanzhou-Zhengzhou-Changsha product oil pipeline. The results revealed precision enhancements in engineering vehicle target detection under real-world conditions, resulting from the YOLOv5 algorithm improvements in three aspects. These advancements led to an overall Mean Average Precision (mAP) of 84.8%, a 6.9% increase over the original YOLOv5 algorithm. The target detection mAPs for excavators, loaders, rollers, and heavy trucks were raised by 4.4%, 7.5%, 9.5%, and 6.0% respectively. [Conclusion] The YOLO-MMCE algorithm provides an efficient solution for detecting overlapped targets and engineering vehicle targets in environments with varying sunlight conditions, illustrating its values in practical applications. (3 Figures, 5 Tables, 28 References)

参考文献/References:

[1] GIRSHICK R, DONAHUE J, DARRELL T, MALIK J. Rich feature hierarchies for accurate object detection and semantic segmentation[C]. Columbus: 2014 IEEE Conference on Computer Vision and Pattern Recognition, 2014: 580-587.
[2] WEI L，DRAGOMIR A，DUMITRU E，CHRISTIAN S， SCOTT R，CHENG-YANG F，et al. SSD: Single Shot MultiBox Detector[J].Springer, Cham, 2016: 21-37. DOI:10.1007/978-3-319-46448-0_2.
[3] GIRSHICK R. Fast R-CNN[C]. Santiago: 2015 IEEE International Conference on Computer Vision (ICCV), 2015:1440-1448.
[4] REN S Q, HE K M, GIRSHICK R, SUN J. 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. DOI: 10.1109/TPAMI.2016.2577031.
[5] 刘德祥，梁苗，王钰棋.改进Faster-RCNN的工程车辆识别方法[J]. 城市勘测，2023（1）：110-113. 10.3969/j.issn.1672-8262. 2023.01.029. LIU D X, LIANG M, WANG Y Q. Improved Faster-RCNN method for engineering vehicle detection[J]. Urban Geotechnical Investigation & Surveying, 2023(1): 110-113.
[6] 周斌，苏鹏，高鹏.基于改进YOLOv3算法的挖掘机检测方法[J]. 数字制造科学，2022，20（2）：141-145. 10.3963/j.issn.1672-3236.2022.02.012. ZHOU B, SU P, GAO P. Excavator detection method based on improved algorithm of YOLOv3[J]. Digital Manufacture Science, 2022, 20(2): 141-145.
[7] 张旭天.基于深度学习的工程车辆检测与追踪[D].石家庄：石家庄铁道大学，2021. ZHANG X T. Construction vehicle detection and tracking based on deep learning[D]. Shijiazhuang: Shijiazhuang Tiedao University, 2021.
[8] REDMON J, DIVVALA S, GIRSHICK R, FARHADI A. You only look once: unified, real-time object detection[C]. Las Vegas: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2016: 779-788.
[9] 刘先禄.基于深度卷积神经网络的小目标检测[D].淮南：安徽理工大学，2022. LIU X L. Small object detection based on deep convolutional neural networks[D]. Huainan: Anhui University of Science and Technology, 2022.
[10] 刘传洋，吴一全.基于深度学习的输电线路视觉检测方法研究进展[J].中国电机工程学报，2023，43（19）：7423-7445. 10.13334/j.0258-8013.pcsee.221139. LIU C Y, WU Y Q. Research progress of vision detection methods based on deep learning for transmission lines[J]. Proceedings of the CSEE, 2023, 43(19): 7423-7445.
[11] 陈业红，姜国龙，褚云飞，张慧仪，张璐，吴朝军.基于锚框的深度学习物体目标检测算法概览[J].微电子学与计算机， 2021，39（7）：12-23. 10.19304/J.ISSN1000-7180.2021. 1340. CHEN Y H, JIANG G L, CHU Y F, ZHANG H Y, ZHANG L, WU C J. An overview of object detection algorithms of deep learning rely on anchor box[J]. Microelectronics & Computers, 2021, 39(7): 12-23.
[12] 何幸，黄永明，朱勇.基于改进YOLOv5的路面坑洼检测方法[J/OL].电子科技：1-8[2024-07-15]. https://doi.org/10.16180/j. cnki.issn1007-7820.2024.07.007. 10.16180/j.cnki.issn1007-7820. 2024.07.007. HE X, HUANG Y M, ZHU Y. Pavement pothole detection method based on improved YOLOv5[J/OL]. Electronic Science and Technology: 1-8[2024-07-15]. https://doi.org/10.16180/j.cnki.issn1007-7820.2024.07.007.
[13] 邵延华，张铎，楚红雨，张晓强，饶云波.基于深度学习的YOLO目标检测综述[J].电子与信息学报，2022，44（10）：3697-3708. 10.11999/JEIT210790. SHAO Y H, ZHANG D, CHU H Y, ZHANG X Q, RAO Y B. A review of YOLO object detection based on deep learning[J]. Journal of Electronics & Information Technology, 2022, 44(10):3697-3708.
[14] ZHANG H Y, CISSE M, DAUPHIN Y N, LOPEZ-PAZ D. Mixup: beyond empirical risk minimization[EB/OL]. (2018-04-27) [2024-07-15]. https://arxiv.org/abs/1710.09412.
[15] 王宇博，马廷淮，陈光明.基于改进YOLOv5算法的农田杂草检测[J]. 中国农机化学报，2023，44（4）：167-173. 10.13733/j.jcam.issn.2095-5553.2023.04.023. WANG Y B, MA T H, CHEN G M. Weeds detection in farmland based on a modified YOLOv5 algorithm[J]. Journal of Chinese Agricultural Mechanization, 2023, 44(4):167-173.
[16] 张立艺，武文红，牛恒茂，石宝，段凯博，苏晨阳.深度学习中的安全帽检测算法应用研究综述[J].计算机工程与应用，2022，58（16）：1-17. 10.3778/j.issn.1002-8331.2203-0580. ZHANG L Y, WU W H, NIU H M, SHI B, DUAN K B, SU C Y. Summary of application research on helmet detection algorithm based on deep learning[J]. Computer Engineering and Applications, 2022, 58(16): 1-17.
[17] HU J, SHEN L, SUN G. Squeeze-and-excitation networks[C]. Salt Lake City: 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2018: 7132-7141.
[18] WANG Q L, WU B G, ZHU P F, LI P H, ZUO W M, HU Q H. ECA-Net: efficient channel attention for deep convolutional neural networks[C]. Seattle: 2020 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020:11531-11539.
[19] WOO S, PARK J, LEE J Y, KWEON I S. CBAM:convolutional block attention module[C]. Munich: Computer Vision–ECCV 2018, 2018: 3-19.
[20] HOU Q B, ZHOU D Q, FENG J S. Coordinate attention for efficient mobile network design[C]. Nashville: 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2021: 13708-13717.
[21] 张宸嘉，朱磊，俞璐.卷积神经网络中的注意力机制综述[J].计算机工程与应用，2021，57（20）：64-72. 10.3778/j.issn.1002-8331.2105-0135. ZHANG C J, ZHU L, YU L. Review of attention mechanism in convolutional neural networks[J]. Computer Engineering and Applications, 2021, 57(20): 64-72.
[22] HUANG C, YANG S Y, LUO Y G, WANG Y S, LIU Z. Visual detection and image processing of parking space based on deep learning[J]. Sensors, 2022, 22(17): 6672. DOI: 10.3390/s22176672.
[23] MENG L L, CUI X, LIU R, ZHENG Z, SHAO H L, LIU J X, et al. Research on metallurgical saw blade surface defect detection algorithm based on SC-YOLOv5[J]. Processes, 2023, 11(9): 2564. DOI: 10.3390/pr11092564.
[24] REZATOFIGHI H, TSOI N, GWAK J Y, SADEGHIAN A, REID I, SAVARESE S. Generalized intersection over union: a metric and a loss for bounding box regression[C]. Long Beach:2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2019: 658-666.
[25] ZHENG Z H, WANG P, LIU W, LI J Z, YE R G, REN D W. Distance-IoU loss: faster and better learning for bounding box regression[J]. Proceedings of the AAAI Conference on Artificial Intelligence, 2020, 34(7): 12993-13000. DOI: 10.1609/aaai.v34i07.6999.
[26] 周秦汉，王振.基于多尺度特征增强卷积神经网络遥感目标检测算法[J].电光与控制，2022，29（11）：74-81. 10.3969/j.issn.1671-637X.2022.11.013. ZHOU Q H, WANG Z. A remote sensing target detection algorithm based on multi-scale feature enhancement CNNs[J]. Electronics Optics & Control, 2022, 29(11): 74-81.
[27] ZHANG Y F, REN W Q, ZHANG Z, JIA Z, WANG L, TAN T N. Focal and efficient IOU loss for accurate bounding box regression[J]. Neurocomputing, 2022, 506: 146-157. DOI:10.1016/j.neucom.2022.07.042.
[28] 王素珍，吕基岳，邓成禹，葛润东，李浩儒. 基于改进YOLOv5的钢铁表面缺陷检测算法[J].国外电子测量技术，2023， 42（10）：43-50. 10.19652/j.cnki.femt.2305102. WANG S Z, LV J Y, DENG C Y, GE R D, LI H R. Improved YOLOv5-based surface defect detection algorithm for steel[J].Electronic Measurement Technology, 2023, 42(10):43-50.

相似文献/References:

[1]孙卉梅　刘路　王德刚　王太勇.面向高后果区工程车辆视觉检测的YOLO-MMCE算法[J].油气储运,2024,43(09):1.
　SUN Huimei,LIU Lu,WANG Degang,et al.An Improved YOLOv5 Algorithm for Visual Detection of Engineering Vehicles in High Consequence Areas[J].Oil & Gas Storage and Transportation,2024,43(09):1.

备注/Memo

孙卉梅，女，2000年生，在读硕士生，2022年毕业于山东工商学院电子信息工程专业，现主要从事机器视觉方向的研究工作。地址：天津市河西区大沽南路1310号，300222。电话：17616239629。Email：17616239629@163.com
通信作者：刘路，男，1982年生，高级工程师，2011年博士毕业于天津大学精密仪器及机械专业，现主要从事机器视觉与数字信号处理方向的研究工作。地址：天津市河西区大沽南路1310号，300222。电话：13722655164。Email：lordman1982@163.com
基金项目：国家自然科学基金资助项目“基于泛信息融合的智能制造系统状态感知与虚拟维护”，51975402。
· Received: 2023-11-02 · Revised: 2023-11-28 · Online: 2024-06-26

更新日期/Last Update: 2024-09-25