Performance Improvement of ICP-SLAM by Human Removal Process Using YOLO

DOI
  • Akiba Keigo
    School of Science and Engineering, Chuo University
  • Suzuki Ryuki
    School of Science and Engineering, Chuo University
  • Ji Yonghoon
    School of Materials Science/Intelligent Robotics Area, Japan Advanced Institute of Science and Technology (JAIST)
  • Pathak Sarthak
    Faculty of Science and Engineering, Chuo University
  • Umeda Kazunori
    Faculty of Science and Engineering, Chuo University

Description

In this paper, we propose a novel iterative closest point (ICP)-based simultaneous localization and mapping (SLAM) approach that can build robust map infor-mation even in indoor environments where humans coexist. Several SLAM methods that have been studied so far assume a stationary environment. But there are challenges in operating in a dynamic environment with moving objects such as humans. Specifically, when a mobile robot constructs a map in an environ-ment where humans coexist nearby, humans cause false matching in alignment sensor data. Furthermore, human occlusion also makes it difficult to construct a map with high accuracy. Therefore, we propose a human removal process that utilizes You Look Only Once (YOLO) to detect humans in image data. In this paper, by using this process with ICP-SLAM, we aim to improve the accuracy of map construction in an environment where humans coexist nearby. In our exper-iments, we verified the accuracy of map construction in comparison with conven-tional methods. This experiment is conducted in an indoor corridor where hu-mans coexist nearby. Although we used ICP-SLAM for verification this time, the human removal process can be adapted to other SLAMS.

Journal

Keywords

Details 詳細情報について

  • CRID
    1390295901064116864
  • DOI
    10.14865/ahi.5.1.1
  • ISSN
    24332372
  • Text Lang
    en
  • Data Source
    • JaLC
  • Abstract License Flag
    Disallowed

Report a problem

Back to top