Open AccessRobust and Accurate Human Tracking Algorithm for Handling Occlusion and Out of Plane Rotation
Author(s) -
Anshul Pareek*,
Nidhi Arora
Publication year - 2019
Publication title -
international journal of innovative technology and exploring engineering
Language(s) - English
Resource type - Journals
ISSN - 2278-3075
DOI - 10.35940/ijitee.l2682.1081219
Subject(s) - superimposition , affine transformation , artificial intelligence , computer vision , computer science , rotation (mathematics) , template , tracking (education) , transformation (genetics) , classifier (uml) , algorithm , mathematics , psychology , pedagogy , biochemistry , chemistry , gene , programming language , pure mathematics
Robust and accurate human tracking using computer vision is acquiring more and more attention, seeking to meet the demands of the increasing number of applications. This paper presents a novel approach to handle occlusion and out of plane rotation. Both these issues are crucial and continue to be a challenge in all state of art algorithms of computer vision-based human tracking. In this paper a SURF oriented scheme based human tracker is proposed, which searches for the target human in expanded rectangular region surrounding the previous target location. There is an online update of object model by selecting fresh templates every time. Here, superimposition of keypoints obtained from previous templates is done on fresh template using Affine transformation. Whether it’s a pose change or not, is affirmed by affine transformation, by calculating aspect ratio of target enclosed region. An Autotuned classifier discriminates the case of occlusion and pose change and confirms tracking failure. The success rate and computational time proves the accomplishment of the proposed algorithm