Overcoming occlusions in semi-autonomous telepresence systems
Overcoming occlusions in semi-autonomous telepresence systems
复制标题
克服半自主远程呈现系统中的遮挡
DOI：
--
复制
发表时间：
2013
期刊：
影响因子：
--
通讯作者：
E. Croft
中科院分区：
文献类型：
--
作者：
Sina Radmard;E. Croft
在本文中，我们提出了一种遮挡处理系统，应用于铰接式临场感平台，同时减少了对操作员的控制负担，并确保视觉交互的连续性。铰接式远程呈现系统通过允许操作者即使在步行对话期间也保持面部接触来改善操作者的存在感。然而，对铰接式系统的附加控制以及控制导航和进行对话沿着会使操作者分心并降低双方的远程呈现体验。因此，我们提出了一个半自主的系统，沿着与视觉跟踪，可以帮助减轻管理铰接平台的负担，通过清除闭塞，可能会出现之间的telepresense平台和第二方（对话者）。特别是，我们演示了一个丢失的目标恢复算法（LTRA）在远程呈现系统中的应用。由于远程呈现系统的性质需要通过平滑运动从遮挡中快速恢复，我们通过采用数值逆Kinetics求解器并引入笛卡尔空间采样方法来生成最佳清除运动来提高应用的LTRA性能。最后，为了验证不同的遮挡清除方法，我们在一个铰接式临场感平台上实现了它们。各种情况下的实验结果支持我们的方法的可行性，以快速恢复在远程呈现系统通信的视觉接触。
In this paper we present an occlusion handling system, applied to articulated telepresence platforms, that simultaneously reduces the control burden on the operator and ensures the continuity of the visual interaction. Articulated telepresence systems improve the sense of presence of the operator by allowing the operator to maintain facial contact even during a walking conversation. However, the additional control of the articulated system, along with controlling navigation and carrying on a conversation can distract the operator and degrade the telepresence experience for both parties. Thus, we propose a semi-autonomous system that, along with visual tracking, can help relieve the burden of managing the articulated platform by clearing occlusions that might arise between the telepresense platform and the second party (the interlocuter). In particular, we demonstrate the application of a lost target recovery algorithm (LTRA) in a telepresence system. As the nature of a telepresence system requires fast recovery from occlusions through smooth motions, we improve the applied LTRA performance by adopting a numerical Inverse Kinematics solver and introducing a Cartesian space sampling method to generate optimal clearing motions. Finally, to validate different occlusion clearing methods, we implement them on an articulated telepresense platform. Experimental results for various scenarios support the feasibility of our approach to quickly recover visual contact in telepresence systems communications.