Paper Abstract and Keywords |
Presentation |
2019-09-04 16:25
Inter-frame interpolation by height constraint for sequential human silhouette images
-- Application to forensic gait analysis under low frame-rate condition -- Daisuke Imoto, Kenji Kurosawa, Masakatsu Honma, Ryo Yokota, Manato Hirabayashi, Yoshinori Hawai (NRIPS) PRMU2019-19 MI2019-38 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
Boundary information is important for defining the shape of an object or a person as a first approximation, and widely used for quantitative analysis in fields such as fluid, cells, medical images, and human video analysis. For example, gait-based human identification technology, which describes the shape of a person as a silhouette (boundary information), has been beginning to be used in forensic investigation. However, under low frame-rate condition, accuracy has dropped due to lack of the in-between information of a pair of successive silhouettes. In order to compensate for such an information loss, a reliable boundary tracking technique for interpolating in-between silhouettes (i.e. boundary tracking) is necessary, but a method suitable for human video analysis has not been proposed yet. In order to solve this problem, in this paper, we proposed a novel simply-extending boundary tracking method based on the assumption that the positions of the body points do not merely change along the height or vertical direction. As a result, by using the proposed method, it is elucidated that we can perform frame interpolation with much higher accuracy than the previous methods. Furthermore, the proposed method has improved the accuracy of gait-based human identification under low frame-rate conditions. It is expected that the proposed method may be effective in other applications or fields such as action recognition, physical motion science, and sports video analysis, etc. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Boundary tracking / inter-frame interpolation / human video analysis / gait-based human identification / / / / |
Reference Info. |
IEICE Tech. Rep., vol. 119, no. 192, PRMU2019-19, pp. 43-48, Sept. 2019. |
Paper # |
PRMU2019-19 |
Date of Issue |
2019-08-28 (PRMU, MI) |
ISSN |
Online edition: ISSN 2432-6380 |
Copyright and reproduction |
All rights are reserved and no part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording, or any information storage and retrieval system, without permission in writing from the publisher. Notwithstanding, instructors are permitted to photocopy isolated articles for noncommercial classroom use without fee. (License No.: 10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) |
Download PDF |
PRMU2019-19 MI2019-38 |
|