• 北大核心期刊(《中文核心期刊要目总览》2017版)
  • 中国科技核心期刊(中国科技论文统计源期刊)
  • JST 日本科学技术振兴机构数据库(日)收录期刊

留言板

尊敬的读者、作者、审稿人, 关于本刊的投稿、审稿、编辑和出版的任何问题, 您可以本页添加留言。我们将尽快给您答复。谢谢您的支持!

姓名
邮箱
手机号码
标题
留言内容
验证码

递归特征融合与并行缩放的航拍工程车辆检测

马学森 储昭坤 马吉

马学森, 储昭坤, 马吉. 递归特征融合与并行缩放的航拍工程车辆检测[J]. 微电子学与计算机, 2022, 39(8): 39-46. doi: 10.19304/J.ISSN1000-7180.2022.0109
引用本文: 马学森, 储昭坤, 马吉. 递归特征融合与并行缩放的航拍工程车辆检测[J]. 微电子学与计算机, 2022, 39(8): 39-46. doi: 10.19304/J.ISSN1000-7180.2022.0109
MA Xuesen, CHU Zhaokun, MA Ji. Engineering vehicle detection in aerial images with recursive feature fusion and parallel scaling[J]. Microelectronics & Computer, 2022, 39(8): 39-46. doi: 10.19304/J.ISSN1000-7180.2022.0109
Citation: MA Xuesen, CHU Zhaokun, MA Ji. Engineering vehicle detection in aerial images with recursive feature fusion and parallel scaling[J]. Microelectronics & Computer, 2022, 39(8): 39-46. doi: 10.19304/J.ISSN1000-7180.2022.0109

递归特征融合与并行缩放的航拍工程车辆检测

doi: 10.19304/J.ISSN1000-7180.2022.0109
基金项目: 

国家重点研发计划资助项目 2020YFC1512601

2020安徽省自然科学基金联合基金 2008085UD08

中央高校基本科研业务费专项资金资助 PA2021GDGP0061

详细信息
    作者简介:

    马学森  男,(1976-),博士,副教授,硕士生导师.研究方向为深度学习、目标检测. E-mail: mxs@hfut.edu.cn

    储昭坤  男,(1996-),硕士研究生.研究方向为深度学习、目标检测

    马吉  男,(1998-),硕士研究生.研究方向为深度学习、图像处理

  • 中图分类号: TP391.4

Engineering vehicle detection in aerial images with recursive feature fusion and parallel scaling

  • 摘要:

    针对无人机航拍输电走廊图像中背景复杂多变、目标偏小且尺度变化大导致检测精度差的问题,本文提出基于RetinaNet递归特征融合与并行缩放的工程车辆检测方法.该方法更适合检测复杂背景中的工程车辆:首先,增添C2层为基础层,与原始骨干网输出层共同用于生成特征金字塔,避免小目标特征被高度压缩;其次,调整原始特征金字塔层次结构,将具有反馈连接的递归结构用于特征提取增强表征能力,设计新颖轻巧的特征融合策略重构特征金字塔,充分利用上下文信息,提高对复杂背景中目标的检测能力;最后,在骨干网C5层的基础上使用多个反卷积块和平均池化层构造并行输出的特征缩放分支,进一步增加特征图的分辨率,提高对小目标的检测精度.在本文构造的工程车辆APEV数据集和公开的PASCAL VOC数据集上分别进行对比实验,结果表明,所提方法的检测速度在满足工程应用需求的前提下,检测精度比原始RetinaNet网络分别提升4.9%和2.7%,与Faster R-CNN、SSD、YOLOv3、YOLOv5、LSN、S-RetinaNet等方法相比精度更高.

     

  • 图 1  无人机电力巡检示意图

    Figure 1.  Schematic diagram of UAV electricity inspection

    图 2  RFS-RetinaNet网络结构图

    Figure 2.  The network structure of RFS-RetinaNet

    图 3  残差单元

    Figure 3.  Residual unit

    图 4  多尺度目标检测方法

    Figure 4.  Some methods of multi-scale objects detection

    图 5  特征融合策略

    Figure 5.  The strategy of feature fusion

    (a)Feature Pyramid Networks; (b)Recursive fusion feature pyramid with feedback connection

    图 6  RFS-RetinaNet特征缩放结构

    Figure 6.  The feature scaling structure of RFS-RetinaNet

    图 7  APEV数据集图片部分示例

    Figure 7.  Some images of APEV Dataset

    图 8  不同方法在APEV数据集上的实验结果对比

    Figure 8.  Experimental results of different methods on the APEV dataset

    ((a) RetinaNet; (b) RFS-RetinaNet)

    表  1  特征融合结果对比

    Table  1.   Results comparison of feature fusion

    Number Construct Dataset mAP(%)
    (a) add APEV 86.0
    (b) Recursion+add APEV 88.1
    (c) Recursion+concat APEV 89.4
    下载: 导出CSV

    表  2  APEV数据集上结果对比

    Table  2.   Comparison of experimental results on APEV Dataset

    Method Backbone mAP(%) FPS
    Faster R-CNN[14] ResNet-50 84.8 5.4
    SSD[15] VGG-16 85.4 20.1
    YOLOv3[21] DarkNet-53 87.3 18.2
    YOLOv5[22] DarkNet-53 88.6 21.3
    LSN[23] ResNet-101 87.4 12.2
    RetinaNet[13] ResNet-50 86.0 17.9
    RFFP ResNet-50 89.4 17.4
    DetectoRS[18] ResNet-50 89.5 3.6
    RFS-RetinaNet(本文) ResNet-50 90.9 16.4
    下载: 导出CSV

    表  3  VOC数据集上结果对比

    Table  3.   Comparison of experimental results on VOC

    Method Backbone mAP(%) FPS
    Faster R-CNN[14] VGG-16 72.4 6.3
    SSD[15] VGG-16 75.1 24
    RetinaNet[13] ResNet-50 77.2 19.3
    S-RetinaNet[24] ResNet-50 79.5 12
    RFFP ResNet-50 79.0 18.7
    DetectoRS[18] ResNet-50 80.8 4.2
    RFS-RetinaNet(本文) ResNet-50 79.9 17.4
    下载: 导出CSV

    表  4  消融实验结果对比

    Table  4.   Results of ablation experiments

    Method Backbone RFFP FS mAP(%)
    RetinaNet ResNet-50 - - 86.0
    RetinaNet ResNet-50 - 88.9
    RetinaNet ResNet-50 - 89.4
    RetinaNet ResNet-50 90.9
    下载: 导出CSV
  • [1] 温新叶, 杨忠伟, 陈昌. 输电线路无人机智能巡检应用研究[J]. 中国设备工程, 2021(23): 31-32. DOI: 10.3969/j.issn.1671-0711.2021.23.018.

    WEN X Y, YANG Z W, CHEN C. Application research on UAV intelligent inspection of transmission lines[J]. China Plant Engineering, 2021(23): 31-32. DOI: 10.3969/j.issn.1671-0711.2021.23.018.
    [2] LI C Z, XU C Y, CUI Z, et al. Learning object-wise semantic representation for detection in remote sensing imagery[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Long Beach, CA, USA: IEEE, 2019: 20-27.
    [3] VNEL F Ö, ÖZKALAYCI B O, CILA C. The power of tiling for small object detection[C]//Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops. Long Beach, CA: IEEE, 2019: 582-591. DOI: 10.1109/CVPRW.2019.00084.
    [4] GUO Y P, XU Y, LI S L. Dense construction vehicle detection based on orientation-aware feature fusion convolutional neural network[J]. Automation in Construction, 2020, 112: 103124. DOI: 10.1016/j.autcon.2020.103124.
    [5] LIN T Y, DOLLÁR P, GIRSHICK R, et al. Feature pyramid networks for object detection[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu, HI, USA: IEEE, 2017: 936-944. DOI: 10.1109/CVPR.2017.106.
    [6] 江波, 屈若锟, 李彦冬, 等. 基于深度学习的无人机航拍目标检测研究综述[J]. 航空学报, 2021, 42(4): 524519. DOI: 10.7527/S1000-6893.2020.24519.

    JIANG B, QU R K, LI Y D, et al. Review of research on drone aerial target detection based on deep learning[J]. Acta Aeronautica et Astronautica Sinica, 2021, 42(4): 524519. DOI: 10.7527/S1000-6893.2020.24519.
    [7] WANG J W, DING J, GUO H W, et al. Mask OBB: A semantic attention-based mask oriented bounding box representation for multi-category object detection in aerial images[J]. Remote Sensing, 2019, 11(24): 2930. DOI: 10.3390/rs11242930.
    [8] 张智, 王进, 王杰, 等. 多尺度和纹理特征增强的小尺寸人脸检测[J]. 计算机应用研究, 2021, 38(3): 914-918. DOI: 10.19734/j.issn.1001-3695.2019.12.0696.

    ZHANG Z, WANG J, WANG J, et al. Multi-scale and texture feature enhancement for small face detection[J]. Application Research of Computers, 2021, 38(3): 914-918. DOI: 10.19734/j.issn.1001-3695.2019.12.0696.
    [9] 万子伦, 张彦波, 王多峰, 等. 复杂环境下多任务识别的人脸口罩检测算法[J]. 微电子学与计算机, 2021, 38(10): 21-27. DOI: 10.19304/J.ISSN1000-7180.2021.0056.

    WAN Z L, ZHANG Y B, WANG D F, et al. Face mask detection algorithm for multi task recognition in complex environment[J]. Microelectronics & Computers, 2021, 38(10): 21-27. DOI: 10.19304/J.ISSN1000-7180.2021.0056.
    [10] YANG X, YANG J R, YAN J C, et al. SCRDet: Towards more robust detection for small, cluttered and rotated objects[C]//Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision. Seoul, Korea (South): IEEE, 2019: 8231-8240. DOI: 10.1109/ICCV.2019.00832.
    [11] 李红艳, 李春庚, 安居白, 等. 注意力机制改进卷积神经网络的遥感图像目标检测[J]. 中国图象图形学报, 2019, 24(8): 1400-1408. DOI: 10.11834/jig.180649.

    LI H Y, LI C G, AN J B, et al. Attention mechanism improves CNN remote sensing image object detection[J]. Journal of Image and Graphics, 2019, 24(8): 1400-1408. DOI: 10.11834/jig.180649.
    [12] YU X H, GONG Y Q, JIANG N, et al. Scale match for tiny person detection[C]//Proceedings of the 2020 IEEE Winter Conference on Applications of Computer Vision. Snowmass, CO, USA: IEEE, 2020: 1246-1254. DOI: 10.1109/WACV45572.2020.9093394.
    [13] LIN T Y, GOYAL P, GIRSHICK R, et al. Focal loss for dense object detection[C]//Proceedings of the 2017 IEEE International Conference on Computer Vision. Venice, Italy: IEEE, 2017: 2999-3007. DOI: 10.1109/ICCV.2017.324.
    [14] 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. DOI: 10.1109/TPAMI.2016.2577031.
    [15] LIU W, ANGUELOV D, ERHAN D, et al. SSD: Single shot MultiBox detector[C]//14th European Conference on Computer Vision. Amsterdam, The Netherlands: Springer, 2016: 21-37. DOI: 10.1007/978-3-319-46448-0_2.
    [16] PINHEIRO P O, LIN Y T, COLLOBERT R, et al. Learning to refine object segments[C]//14th European Conference on Computer Vision. Amsterdam, The Netherlands: Springer, 2016: 75-91. DOI: 10.1007/978-3-319-46448-0_5.
    [17] LI Z X, ZHOU F Q. FSSD: feature fusion single shot multibox detector[Z]. arXiv preprint arXiv: 1712.00960, 2018.
    [18] QIAO S Y, CHEN L C, YUILLE A. DetectoRS: Detecting objects with recursive feature pyramid and switchable atrous convolution[C]//Proceedings of the 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Nashville, TN, USA: IEEE, 2021: 10208-10219. DOI: 10.1109/CVPR46437.2021.01008.
    [19] LIANG X, ZHANG J, ZHUO L, et al. Small object detection in unmanned aerial vehicle images using feature fusion and scaling-based single shot detector with spatial context analysis[J]. IEEE Transactions on Circuits and Systems for Video Technology, 2020, 30(6): 1758-1770. DOI: 10.1109/TCSVT.2019.2905881.
    [20] HUANG J, RATHOD V, SUN C, et al. Speed/accuracy trade-offs for modern convolutional object detectors[C]//Proceedings of the 2017 IEEE Conference on Computer Vision and Pattern Recognition. Honolulu, HI, USA: IEEE, 2017: 3296-3297. DOI: 10.1109/CVPR.2017.351.
    [21] REDMON J, FARHADI A. YOLOv3: An incremental improvement[Z]. arXiv preprint arXiv: 1804.02767, 2018.
    [22] JOCHER G. Yolov5[EB/OL]. (2020-08-09). https://github.com/ultralytics/yolov5. Accessed 2021. -02-12.
    [23] WANG T C, ANWER R M, CHOLAKKAL H, et al. Learning rich features at high-speed for single-shot object detection[C]//Proceedings of the 2019 IEEE/CVF International Conference on Computer Vision. Seoul, Korea (South): IEEE, 2019: 1971-1980. DOI: 10.1109/ICCV.2019.00206.
    [24] 李成豪, 张静, 胡莉, 等. 基于多尺度感受野融合的小目标检测算法[J/OL]. 计算机工程与应用: 1-7[2021-12-23]. http://kns.cnki.net/kcms/detail/11.2127.TP.20210420.1358.074.html.

    LI C H, ZHANG J, HU L, et al. Small object detection algorithm based on multiscale receptive field fusion[J/OL]. Computer Engineering and Applications: 1-7[2021-12-23]. http://kns.cnki.net/kcms/detail/11.2127.TP.20210420.1358.074.html.
    [25] CHEN K, PANG J M, WANG J Q, et al. Hybrid task cascade for instance segmentation[C]//Proceedings of the 2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach, CA, USA: IEEE, 2019: 4969-4978. DOI: 10.1109/CVPR.2019.00511.
  • 加载中
图(8) / 表(4)
计量
  • 文章访问数:  39
  • HTML全文浏览量:  24
  • PDF下载量:  7
  • 被引次数: 0
出版历程
  • 收稿日期:  2022-02-19
  • 修回日期:  2022-03-17
  • 网络出版日期:  2022-08-15

目录

    /

    返回文章
    返回