利用增強式骨架識別技術的智慧型避障自走車之人體跟隨控制

本文的主要目的是開發出人體跟隨及閃避障礙物的智慧型自走車。在本文提出的方法可以解決勞力短缺或幫助人們生活更便利。該系統可分為兩個部分：自走車設計及骨架識別。結合這兩部份，達成避開障礙物及人體跟隨的目標。在骨架識別中，此系統使用微軟Kinect和Kinect SDK來獲得深度及目標的骨架的資訊來做控制設計。根據四個節點的位置，來判斷目標位置並且再去穩定地跟隨目標物。當目標遺失時，自走車可以利用骨架識別來判斷，且不使用色彩資訊，亦可有效地找到目標。我們的系統使用PID控制和模糊控制用來改善及提高系統的穩定性。而且，當系統避開障礙物的同時，自走車亦可跟隨目標的能力。此系統使用紅外感測器，來計算與障礙物的距離，並且避免碰撞障礙物。因此，自走車能夠在有干擾及複雜環境達到目的。

關鍵字

人體跟隨；智慧型自走車；避障；骨架辨識

並列摘要

In this thesis, the main purpose is to develop an intelligent vehicle with obstacle avoidance for person following. The proposed intelligent system has the abilities to solve the various fields of labor shortage for useful assistance. This system can be roughly divided into two major parts: mobile vehicle design and skeleton image recognition. Combining these two parts to reach the person following with obstacle avoidance is our desired control objective. In the portion of skeleton image processing, the system employs Microsoft Kinect and Kinect SDK to obtain the depth and skeleton information of the moving target. Then, this intelligent system utilizes the enhanced skeleton information and gesture of person to achieve the following control design. According to the position of four joints, the system receives the target position and stably follows the path of the target. When the target is lost in the complex environments, the intelligent vehicle can apply skeleton image recognition without the use of color information to find the target effectively. For the following design of the system, both PID control and fuzzy control are employed to improve and enhance the system stability of motors. In addition, the system provides the ability of obstacle avoidance while following the moving target. The system determines the distance of obstacle with infrared sensors to avoid collision. As a result, the intelligent vehicle system can successfully achieve the purpose of person following with obstacle avoidance under existence of disturbance and complex environments.