{"id":260,"date":"2014-10-30T13:07:20","date_gmt":"2014-10-30T11:07:20","guid":{"rendered":"http:\/\/www.mpateraki.org\/?p=260"},"modified":"2015-03-03T09:30:31","modified_gmt":"2015-03-03T07:30:31","slug":"robust-articulated-upper-body-pose-tracking-under-severe-occlusions","status":"publish","type":"post","link":"https:\/\/mpateraki.org\/?p=260","title":{"rendered":"Robust articulated upper body pose tracking under severe occlusions"},"content":{"rendered":"

Sigalas M., Pateraki M.<\/strong> and Trahanias P., 2014. Robust articulated upper body pose tracking under severe occlusionsIn Proc. of the IEEE\/RSJ International Conference on Intelligent Robots and Systems (IROS), 14-18 September, Chicago, USA. [doi<\/a>] [pdf<\/a>] [bib<\/a>]<\/p>\n

Abstract:<\/p>\n

Articulated human body tracking is one of the\u00a0most thoroughly examined, yet still challenging, tasks in Human\u00a0Robot Interaction. The emergence of low-cost real-time depth\u00a0cameras has greatly pushed forward the state of the art in the\u00a0field. Nevertheless, the overall performance in complex, real\u00a0life scenarios is an open-ended problem, mainly due to the\u00a0high-dimensionality of the problem, the common presence of\u00a0severe occlusions in the observed scene data, and errors in the\u00a0segmentation and pose initialization processes.\u00a0In this paper we propose a novel model-based approach\u00a0for markerless pose detection and tracking of the articulated\u00a0upper body of multiple users in RGB-D sequences. The main\u00a0contribution of our work lies in the introduction and further\u00a0development of a virtual User Top View, a hypothesized view\u00a0aligned to the main torso axis of each user, to robustly estimate\u00a0the 3D torso pose even under severe intra- and inter-personal\u00a0occlusions, exempting at the same time the requirement of\u00a0arbitrary initialization. The extracted 3D torso pose, along with\u00a0a human arm kinematic model, gives rise to the generation of\u00a0arms hypotheses, tracked via Particle Filters, and for which\u00a0ordered rendering is used to detect possible occlusions and\u00a0collisions.\u00a0Experimental results in realistic scenarios, as well as comparative\u00a0tests against the NiTE (TM) user generator middleware using\u00a0ground truth data, validate the effectiveness of the proposed\u00a0method.<\/p>\n\n\n\n
\"\"<\/a><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n
    \n
  • Supplementary material (video<\/strong>):
    \n