Optics and Precision Engineering, Volume. 32, Issue 6, 857(2024)

RGB-D SLAM method of dynamic scene based on instance segmentation and optical flow

Chenggen WANG1, Jinlong SHI1、*, Haowei ZHU1, Suqin BAI1, Yunhan SUN2, Jiawen LU1, and Shucheng HUANG1
Author Affiliations
  • 1School of Computer Science and Engineering, Jiangsu University of Science and Technology, Zhenjiang22000,China
  • 2State Key Laboratory for Novel Software Technology, Nanjing University, Nanjing10046,China
  • show less
    Figures & Tables(10)
    System framework diagram
    Effect diagram of non rigid object detection
    Diagram of relationship between optical flow and self flow
    Diagram of Absolute trajectory error comparison
    Background reconstruction map
    Comparison of reconstruction results
    Dynamic rigid object fusion
    • Table 1. Comparison of absolute trajectory error of Bonn dataset

      View table
      View in Article

      Table 1. Comparison of absolute trajectory error of Bonn dataset

      序列SF36DS22RF33AF31OURS
      Balloon0.2330.0300.1750.0280.028
      Balloon20.2930.0290.2540.0300.029
      Balloon Tracking0.2210.0490.3020.0450.041
      Balloon Tracking20.3660.0350.3220.0330.059
      Crowd3.5860.0160.2040.0160.027
      Crowd20.2150.0310.1550.0270.028
      Crowd30.1680.0380.1370.0230.038
      Kidnapping0.3360.0290.1480.0300.029
      Kidnapping20.2630.0350.1610.0300.025
      Moving No Box0.1410.2320.0710.0700.025
      Moving No Box20.3640.0390.1790.0290.037
      Moving O Box0.3310.0440.3430.3430.262
      Moving O Box20.3090.2630.5280.4430.134
      Person Tracking0.4840.0610.2890.0700.041
      Person Tracking20.6260.0780.4630.0710.056
      Placing No Box0.1250.5750.1060.0880.021
      Placing No Box20.1770.0210.1410.0200.022
      Placing No Box30.2560.0580.1740.0510.043
      Placing O Box0.3300.2250.5710.3240.177
      Removing No Box0.1360.0160.0410.0200.017
      Removing No Box20.1290.0210.1110.0250.025
      Removing O Box0.3340.2910.2220.3140.197
      Synchronous0.4460.0150.4100.0140.012
      Synchronous20.0370.0090.0220.0100.009
    • Table 2. Comparison of absolute trajectory error of TUM dataset

      View table
      View in Article

      Table 2. Comparison of absolute trajectory error of TUM dataset

      序列DS22RF33Rigid37AF31OURS
      Sitting Static0.0070.0110.0190.0280.006
      Sitting XYZ0.0150.0260.0540.0210.014
      Sitting Halfsphere0.0280.0380.1290.0350.019
      Walking Static0.0070.0140.0180.0110.007
      Walking XYZ0.0170.0740.0900.0250.015
      Walking Halfsphere0.0260.0480.0760.0350.027
    • Table 3. Comparison of Absolute Trajectory Errors for Different Strategies in the Bonn Dataset

      View table
      View in Article

      Table 3. Comparison of Absolute Trajectory Errors for Different Strategies in the Bonn Dataset

      序列全程剔除动态剔除
      Balloon0.0280.028
      Balloon20.0290.029
      Balloon Tracking0.0490.041
      Balloon Tracking20.0890.059
      Moving No Box0.0350.025
      Moving No Box20.0420.037
      Moving O Box0.2620.262
      Moving O Box20.1340.134
      Placing No Box0.0220.021
      Placing No Box20.0250.022
      Placing No Box30.0450.043
      Placing O Box0.1740.177
      Removing No Box0.0190.017
      Removing No Box20.0280.025
      Removing O Box0.1970.191
    Tools

    Get Citation

    Copy Citation Text

    Chenggen WANG, Jinlong SHI, Haowei ZHU, Suqin BAI, Yunhan SUN, Jiawen LU, Shucheng HUANG. RGB-D SLAM method of dynamic scene based on instance segmentation and optical flow[J]. Optics and Precision Engineering, 2024, 32(6): 857

    Download Citation

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

    Category:

    Received: Sep. 7, 2023

    Accepted: --

    Published Online: Apr. 19, 2024

    The Author Email: Jinlong SHI (shi_jinlong@163. com)

    DOI:10.37188/OPE.20243206.0857

    Topics