3D Object Recognition and Pose Estimation
           

- 指導教授 黃漢邦 博士 研究生 唐婉儒

- Advisor :Dr.Han-Pang Huang Student : Shu-Wei Chang

Lab. of Robotics., Department of Mechanical Engineering, National Taiwan University, Taiwan
Abstract:

Robotics has been much development over the last few decades. In the near future, there may be multiple robots able to perform household chores, dangerous work or exhausting jobs in families, work places or hospitals. However, most studies do not consider these multi-robot system and human-robot interaction together. The thesis attempts to integrate gesture recognition techniques and multi-robot formation control to promote the functionality of multi-robot coordination under human commanding.

We firstly discuss the method used for gesture recognition system including hand segmentation and localization, classification, post-processing in real-time applications. Then, combining the gesture system, natural multi-robot human interaction is divided into two cases: gesture target identification for a multi-robot system and adaptive humans-robots group behavior formation. The first case deals with developing a strategy for determining the robot that a gesture command is intended for to coordinate a reasonable outcome.

The second case regards robots and humans as companions and as group members and the adaptation of formations according to humans when request. Finally, we demonstrate that naturally multi-robot human interaction based on the gesture recognition to close the gap between humans and multi-robot.



中文摘要:


在過去幾十年中，機器人領域蓬勃發展，因此在不久的將來，很有可能有很多的機器人一起在家幫忙做家務事、在工廠做危險的任務以及在醫院分擔繁重的工作。

然而，儘管機器人領域發展興盛，但大多研究並無考慮多機器人系統和與人互動之結合，因此，本論文旨在提升當有人用手勢指示時多機器人的協調與整合之能力。論文首先討論手勢辨識的方法包括手的分割與定位、類別分類、後處理以便在即時系統中應用。接著，結合手勢辨識系統，多機器人與人之和諧互動可分為兩者: 多機器人手勢目標之確定與適性人與機器人之群體行為。

第一個主要為處理與發展一策略讓多機器人系統能清楚且確切的了解給予指示之人的目標與對象；第二個為將機器人與人們視為一個群體，當一群體有一目標運動隊形時，機器人能因應人們的位置而有相對的移動與動作。最後，我們展示出基於手勢辨識系統多機器人與人類之和諧的互動以縮短人類與機器人之間的距離。