Title: | Research on vision-based hand gesture recognition for astronaut virtual training |
Authors: | Qingchao Xie, Author |
Material Type: | ISU internship report |
Publisher: | Illkirch-Graffenstaden (France) : International Space University, 2020 |
Size: | 1 electronic resource (v, 37 p.) / col. ill. |
Bibliography note: | Includes bibliographical references |
Languages: | English |
Class number: | TL850 |
Subjects: | Astronauts--Training ; Virtual reality |
Description: | To solve the problem of self-occlusion gesture recognition in astronaut virtual training, and provide a more natural virtual interactive training method for astronauts, this report discussed methods of hand pose estimation in astronaut virtual training. Based on the characteristics of astronauts' virtual interactive training, the advantages and disadvantages of the existing methods of hand pose estimation vision-based were analyzed, and the corresponding improvement methods also were put forward. To solve self-occlusion, a method of rapid occlusion prediction based on Autoencoder (AE) was proposed. Combined with mature hand pose estimation from a single image, a general multi-camera fusion framework based on hand joints occlusion prediction was constructed, which can improve the accuracy of hand pose estimation and solve the problem of severe self-occlusion in astronaut virtual training. |
ISU program : | Master of Space Studies |
Permalink: | https://isulibrary.isunet.edu/index.php?lvl=notice_display&id=11094 |
Read online (1)
![]() Xie, Qingchao_INT report (1.58MB) Adobe Acrobat PDF |