Optics and Precision Engineering, Volume. 30, Issue 20, 2538(2022)

Overview of visual pose estimation methods for space missions

Rui ZHOU... Yanfang LIU*, Naiming QI and Jiayu SHE |Show fewer author(s)
Author Affiliations
  • School of Astronautics, Harbin Institute of Technology, Harbin150090, China
  • show less
    Figures & Tables(9)
    Vision-based space mission development
    Classification of visual pose estimation methods
    Flowchart of traditional pose estimation method
    Classification of visual pose estimation methods based on deep learning
    Comparison of target recognition network structure
    • Table 1. Applications based on visual technology in space missions

      View table
      View in Article

      Table 1. Applications based on visual technology in space missions

      项目名称国家时间测量方式视觉系统观测目标
      SRMS39加拿大1981遥操作监视单目相机空间站等
      ROTEX德国1993遥操作监视单目相机-
      TORU40俄罗斯1994遥操作监视双目相机飞船、空间站
      RRM41美国2011遥操作监视双目相机卫星、加注口
      玉兔号42中国2013遥操作监视双目相机月面
      EST-VII43日本1997合作目标测量CCD相机目标星
      OE44美国1999合作目标测量VGS目标星
      太空船44美国2005合作目标测量AVGS目标星
      天宫一号45中国2011合作目标测量雷达、CCD相机对接靶标
      天宫二号45中国2016合作目标测量单目相机对接靶标
      FREND/SUMO46美国终止非合作目标立体视觉对接环
    • Table 1. Applications based on visual technology in space missions

      View table
      View in Article

      Table 1. Applications based on visual technology in space missions

      项目名称国家时间测量方式视觉系统观测目标
      PHOENIX47美国2011年非合作目标立体视觉星箭对接环
      ROGER48欧空局终止非合作目标变焦相机空间碎片、故障航天器
      TECSAS49德国终止非合作目标光学相机、激光雷达对接环
      DEOS50德国2006非合作目标EPOS系统对接环
      CX-OLEV50德国2007非合作目标多相机组合目标星
      SDMR51日本2009非合作目标双目视觉空间碎片
      SMART-OLEV欧洲2009非合作目标双目视觉发动机喷嘴
      天问一号52中国2020非合作目标中高分辨率相机火星地表
      RSGS美国2021非合作目标双目视觉目标星
      祝融号火星车52中国2021非合作目标多光谱、导航相机火星地表
      e.Deorbit53加拿大2021非合作目标-空间碎片
    • Table 2. Comparison of target recognition networks

      View table
      View in Article

      Table 2. Comparison of target recognition networks

      分类算法作者时间检测速度mAP(VOC2007)mAP(COCO2017)

      一阶

      网络

      YOLO69Redmon J, Divvala S201545 FPS63.40%-
      SSD70Liu W201559 FPS73.90%31.20%
      YOLOv273Redmon J, Farhadi A201767 FPS76.80%21.60%
      DSSD71Fu C Y, Liu W, et al20179.5 FPS78.60%-
      RSSD72Jeong J, Park H, Kwak N201735 FPS78.50%-
      YOLOv374Redmon J, Farhadi A201872 FPS-33.00%
      YOLOv475Bochkovskiy A202065 FPS-43.50%
      YOLOv5sJocher G20202 ms-36.70%
      YOLOv5m20202.7 ms-44.50%
      YOLOv5l20203.8 ms-48.20%
      YOLOv5s20206.1 ms-50.40%

      二阶

      网络

      R-CNN76Malik201447 s/张66.00%-
      Fast-RCNN77Girshick R20152 s/张66.90%-
      Faster-RCNN78Ren S Q, He K M20165 FPS66.90%36.80%
      Mask-RCNN79He K M, Gkioxari G20185 FPS-37.70%
      Mask Scoring-RCNN80Huang Z J,Huang L C20195 FPS-41.70%
    • Table 3. Comparison of pose estimation networks

      View table
      View in Article

      Table 3. Comparison of pose estimation networks

      分类名称作者时间对象特点
      整体回归法PoseCNN81Yu X, Schmidt T2017RGB图像高维语义嵌入低维特征
      BB882Rad M, Lepetit V2017RGB图像采用8个顶点的2D投影预测3D姿态
      YOLO-6D83Tekin B, Sinha S N2018RGB图像直接预测投影顶点的2D图像位置
      SSD-6D84Kehl W, Manhardt F2017RGB图像采用离散化视图将回归问题转化为分类问题
      Deep6DPose85Do T, Cai M2018RGB图像端对端体系,附加区域建议网络
      分类投票法MCN86Li C, Bai J2018RGB图像多视图融合,减少单视图的模糊性
      CullNet87Gupta K,Petersson L2019RGB图像利用置信度评分对位姿估计进行筛选
      DenseFusion88Wang C, Xu D F2019多源数据采用稠密融合网络提取像素级特征,迭代细化
      PVNet89Peng S, Liu Y2019RGB图像引入关键点方向,并添加不确定性驱动
      PVNE3D90He Y S,Sun W2020RGB-D图像拓展PVNet到三维关键点,利用几何约束
    Tools

    Get Citation

    Copy Citation Text

    Rui ZHOU, Yanfang LIU, Naiming QI, Jiayu SHE. Overview of visual pose estimation methods for space missions[J]. Optics and Precision Engineering, 2022, 30(20): 2538

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: Information Sciences

    Received: Jan. 17, 2022

    Accepted: --

    Published Online: Oct. 27, 2022

    The Author Email: LIU Yanfang (yanfangliu@hit.edu.cn)

    DOI:10.37188/OPE.20223020.2538

    Topics