OS15-4 Action recognition based on binocular vision

Title:	OS15-4 Action recognition based on binocular vision
Publication:	ICAROB2017
Volume:	22
Pages:	160-164
ISSN:	2188-7829
DOI:	10.5954/ICAROB.2017.OS15-4
Author(s):	Yiwei Ru,, Hongyue Du, Shuxiao Li, Hongxing Chang
Publication Date:	January 19, 2017
Keywords:	action recognition, binocular version, convolutional neural networks, motion history image
Abstract:	Aimed at the problem that the recognition accuracy of the monocular camera is low, we propose a binocular vision recognition algorithm for action recognition based on HART-Net(Human action recognition networks).Firstly, the left and right views obtained by the binocular camera are matched to obtain the depth map of the human body .Then, the depth information is projected onto the three planes, the projection images of three directions are used to construct MHI (motion history image), and are combined into a new image. Finally, we use HART-Net to train a classifier for action recognition. Experimental results show that the binocular recognition algorithm is 18% more accurate than the monocular recognition algorithm.
PDF File:	https://alife-robotics.co.jp/members2017/icarob/data/html/data/OS_pdf/OS15/OS15-4.pdf
Copyright:	© The authors. This article is distributed under the terms of the Creative Commons Attribution License 4.0, which permits non-commercial use, distribution and reproduction in any medium, provided the original work is properly cited. See for details: https://creativecommons.org/licenses/by-nc/4.0/