基于ILF-YOLOv3的人员在岗状态检测算法研究On-the-job Status Detection Algorithm Based on ILF-YOLOv3
谢斌红,栗宁君,陈立潮,张英俊
摘要(Abstract)：
为了解决员工在岗状态的实时监测和管理问题,提出了一种改进YOLOv3的目标检测算法,ILF-YOLOv3(Improve Loss and Feature-YOLOv3).首先,使用二分交叉熵损失函数和添加制衡权重参数的方式对YOLOv3算法的损失函数进行改进;然后,增加了模型多尺度特征检测模块的特征融合密度;最后,针对采样数据集单一性的问题,采用生成式对抗网络对其进行定向增强。实验结果表明,改进后的算法在自制的StaffSData-Strong数据集上mAP值提高了7.9%,召回率提高了14%.
关键词(KeyWords)： 目标检测;YOLOv3网络;交叉熵损失函数;多尺度特征融合;ILF-YOLOv3网络;在岗状态检测
基金项目(Foundation): 山西省重点计划研发项目(201803D121048);山西省重点计划研发项目(201803D121055);; 山西省重点研发计划重点项目(201703D111027);; 山西省科技重大专项项目(20141101001)
作者(Author): 谢斌红,栗宁君,陈立潮,张英俊
参考文献(References)：
- [1] KRIZHEVSKY A,SUTSKEVER I,HINTON G E.ImageNet classification with deep convolutional neural networks[C]//Pro-c of the 25th International Conference on Neural Inform-ation Processing Systems,2012:1097-1105.
- [2] SERMANET P,EIGEN D,ZHANG X,et al.Overfeat:Integrate-d recongnition,localization and detection using convolutional networks[J/OL].(2014-02-14).https://arxiv.org/pdf/1312.6229.pdf.
- [3] SZEQEDY C,LIU W,JIA Y,et al.Going deeper withConvolution[C/OL].https://arxiv.org/pdf/1409.4842pdf.
- [4] SIMONYAN K,ZISSERMAN A.Very Deep Convoluti-onal Networks for Large-Scale Image Recongnition[J/OL].https://arxiv.org/pdf/1409.1556.pdf.
- [5] HE KAIMING,ZHANG XIANGYU,REN SHAOQING,et al.Deepresidual learning for Image recongnition[J/OL].(2015-10-10).https://arxiv.org/pdf/1512.03385.pdf.
- [6] GIRSHICK R,DONAHUE J,DAREELL T,et al.Rich feature h-ierarchies for accurate object detection and Semantic seg-mentation[C]//Proc of IEEE Conference on Computer Vi-sion and Pattern Recognition,2014:580-587.
- [7] GIRSHICK R.Fast R-CNN [C]//Proc of IEEE InternationalConference on ComputerVision,2015:1440-1448.
- [8] OUYANG WANLI,WANG XIAOGANG,ZENG XINGYU,et al.D-eepID-Net:deformable deep convolution neural networks for object detection [C]//Proc of IEEE Conference on Computer Vision and Pattern Recongnition,2015:2403-2412.
- [9] REN SHAOQING,HE KAIMING,GIRSHICK R,et al.Faster R-CNN:towards real-time object detetion with region pro-posal netw-orks[C]//Proc of Advances in Neural Informa-tion Processing Systems,2015:91-99.
- [10] REDMON J,DIVVALA S,GIRSHICK R,et al.You Only Look Once:Unified,Real-Time Object Detection[C]//Proc of the IEEE Conference on Computer Vision and Pattern Recognition.Washington,USA:IEEE,2016:779-788.
- [11] REDMON J,FARHADI A.YOLO9000:Better,Faster,Strong-er[J/OL].(2016-10-25).https://arxiv.org/pdf/1612.08242.pdf.
- [12] REDMON J,FARHADI A.YOLOv3:An incremental improvement.(2018-04-08).https://arxiv.org/pdf/1804.02767.pdf.
- [13] 施辉，陈先桥，杨英.改进YOLO v3的安全帽佩戴检测方法[J/OL].计算机工程与应用.http://kns.cnki.net/kcms/detail/11.2127.TP.20190311.1538.012.html.
- [14] 郑志强，刘妍妍，潘长城，等.改进 YOLO V3 遥感图像飞机识别应用[J].电光与控制，2019,26(4):28-32.
- [15] 鞠默然，罗海波，王仲博，等.一种改进的YOLO V3及其在小目标检测中的应用[J].光学学报，2019,39(7):245-252.
- [16] LIN T Y,GOYAL P,GIRSHICK P,et al.Focal loss for de-nse object detection[J/OL].(2017-08-07).https://arxiv.org/pdf/1708.02002.pdf.
- [17] RADFORD A,METZ L,CHINTALA S.Unsupervised Representation Learning with Deep Convolutional Generative Adversarial Networks[J/OL].(2016-01-07).https://arxiv.org/pdf/1511.06434v1.pdf.