Title: The object localisation based on multiple image information for a humanoid robot

Authors: Qinjun Du; Yang Wang; Shuai Xu

Addresses: School of Electrical and Electronic Engineering, Shandong University of Technology, 12 Zhangzhou road, Shandong Zibo, 255091, China ' School of Electrical and Electronic Engineering, Shandong University of Technology, 12 Zhangzhou road, Shandong Zibo, 255091, China ' School of Electrical and Electronic Engineering, Shandong University of Technology, 12 Zhangzhou road, Shandong Zibo, 255091, China

Abstract: The stable and fast segmentation of the object image is a key technology for the humanoid robot in the complex background, which attributes to identification and localisation the target object. Only using single image information cannot guarantee to accurately segment the target object image. We selected image feature information using multi-sensor information fusion method to segment the target object. The depth, colour, shape and size are the ideal image information for the robots visual perception system. Using the humanoid robot stereo vision system and the depth, colour, shape and size of four types of image information, a fast and gradually approaching target area image segmentation method has been designed. With the identification and localisation target object image method, the humanoid robot can find and locate the target object. The experiments show that based on multi-image information features, the humanoid robot can locate the target object, and the localisation accuracy has been improved.

Keywords: humanoid robots; visual localisation; information fusion; binocular vision; object localisation; multiple images; robot vision; image segmentation; multiple sensors; sensor fusion; depth; colour; shape; size; visual perception; stereo vision.

DOI: 10.1504/IJCSM.2016.078710

International Journal of Computing Science and Mathematics, 2016 Vol.7 No.4, pp.371 - 380

Accepted: 04 Dec 2015
Published online: 01 Sep 2016 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article