Paper Abstract and Keywords |
Presentation |
2018-10-11 14:30
A study on ship type identification by use of deep neural network Ryouichi Nishimura, Katsuhiro Temma (NICT), Kiyohiko Hattori (Saitama Inst. of Tech.), Kenji Kaneko (TEAMS), Akinori Ito (Tohoku Univ.), Toyonobu Fujii (TEAMS), Akihiro Kijima (Tohoku Univ.) EA2018-54 |
Abstract |
(in Japanese) |
(See Japanese page) |
(in English) |
Poaching has recently become a serious problem due to the globalization of food culture and the accompanied rising prices. Considering bad effects on the global ecosystems by the overfishing, potential deficit would be enormous in the future. Workers on aquaculture are now forced to keep surveying a poaching boat throughout the night because such a boat should be caught in flagrante delicto. If a machine can do the job instead of humans, it would be a promising solution. Poaching boats usually move around in the dark. Therefore, it is desirable to detect them using characteristics of the sound, namely audio fingerprint, produced by each boat. We tried to develop such a system using Deep Neural Network (DNN). Sound and video of ships were continuously recorded and databases were constructed using the data for 26 days. Classification into 14 classes combined with a post process of moving average and binarization showed a performance of approximately 0.9 at most in F-measure depending on ship types. |
Keyword |
(in Japanese) |
(See Japanese page) |
(in English) |
Aquaculture / Poaching surveillance / Cepstrum / Real-time system / DNN / / / |
Reference Info. |
IEICE Tech. Rep., vol. 118, no. 234, EA2018-54, pp. 1-6, Oct. 2018. |
Paper # |
EA2018-54 |
Date of Issue |
2018-10-04 (EA) |
ISSN |
Print edition: ISSN 0913-5685 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 |
EA2018-54 |
|