以二維人像分析及三維電腦視覺技術作跟人自動車之研究

Abstract

本研究主要是提出一套基於電腦視覺技術，讓自動車航行在室內環境中具有偵測與跟隨人物的能力，並且具備量測與使用者間的距離及和使用者產生互動的功能。我們利用一台具有機器手臂裝置的自動車做為實驗平台，並且利用無線操控的方式讓自動車航行在室內的環境中。我們提出了利用參考資料的方式，以便對於不同身高的人來進行人物和自動車的距離量測。對於環境中光線的變化，我們建立改善的膚色模型來偵測跟隨的目標人物。並且基於形狀限制的概念，在近距離時利用區域偵測來跟隨人物。針對真實環境中不均勻光線的情況，我們提出了漸進式的人臉偵測方法，將畫面分割為320x240到40x30的區塊，直到偵測出人臉。此外，我們提出了自動車轉彎模式，利用十字路口位置和機械手臂的角度資訊，來適應跟隨人物迅速轉彎並且行走在狹窄的路徑上的情況。我們同時提供人機互動的策略，藉由偵測人物的朝向以及揮手的行為來使自動車提供更智慧型的服務。最後我們以成功的偵測和跟隨實驗證明本系統的完整性與可行性。

An intelligent vision-based vehicle system for person following by distance measurement, human detection and following, as well as interaction with humans in indoor environments is proposed. A vehicle with wireless control and a web camera is used as a test bed. A robot arm is equipped on the vehicle to hold the camera at a higher position and is used to change the orientation of the camera. First, a method for computing the distance between a person and the vehicle for different people heights is proposed by the use of learned data of a reference person. Next, a human face detection technique is proposed, which uses a skin-colored model adapting to luminance changes. Besides, a technique of area tracking for human following using images only including part of clothes is proposed, which is based on the concept of shape circumscribing. Furthermore, because the intensity is non-uniform in the real environment, also proposed is a progressive face detection method which uses blocks with sizes varying from 320□240 to 40□30 until the system detects the human face. To follow a target person who turns fast at a corner in a narrow path, a technique of using crossroad points and the image information taken by the robot arm to search the disappearing person is proposed. For avoiding hitting the wall, the vehicle is also enabled to adjust its orientation for monitoring the surrounding when it follows the person. In addition, techniques for intelligent interaction with humans to provide more intelligent services are also proposed, which can be employed to detect the facing direction and the hand movement of a person. Good experimental results show the flexibility and feasibility of the proposed methods for the application of indoor person following.