應用於機器人導航與控制之室內三維地圖視覺重建

Vision-Based 3-D Indoor Map Building for Navigation and Control of Mobile Robots

李廷軒(Ting-Shiuan Li 李廷軒)

指導教授 ： 張文中

國立臺北科技大學/電資學院/電機工程系研究所/碩士(2007年)

摘要

在機器人導航與控制中，精確的環境地圖對於導航與控制是相當重要的資訊，而在傳統已知的方法中精確的地圖重建必須仰賴昂貴的感測器，如聲納、雷射測距儀等。因此，本論文提出以視覺為基礎的兩種架構，完成未知室內環境之三維地圖重建，並利用建立的地圖做為機器人導航與控制的依據。第一種架構使用兩支攝影機作為感測器，利用所提出的影像處理方法找出現實空間中的特徵點所投影在兩影像中的位置，進而藉由兩攝影機之間的已知關係重建出所有特徵點在現實空間的座標，同時判斷特徵點之間的幾何關係，找出機器人可通行的道路。第二種架構為使用單支攝影機搭配水平雷射結構光為感測器，使用影像處理擷取出水平雷射結構光在影像中的線段，藉由水平雷射結構光與攝影機之間已知的幾何關係，進而重建出線段在現實空間的位置以完成三維地圖的重建，同時找出可供機器人通行的路徑。在本論文中將所提出之兩架構已藉自製輪式機器人系統驗證於現實環境，並藉由道路與機器人之間的關係設計導航控制法則以完成定點控制，讓機器人可自主的在未知環境中移動。

關鍵字

雙眼視覺 ； 地圖重建 ； 移動型機器人 ； 機器人導航 ； vSLAM

並列摘要

An accurate environment map provides essential information for navigation and control of mobile robots. In conventional control techniques, building an accurate environment map requires expensive sensing devices, such as sonar and laser range finder. This research proposes two vision-based approaches to build three-dimensional maps to facilitate navigation and control of mobile robots in unknown environment. In the first approach, binocular vision is employed. Based on an image processing method described in the research, keypoint positions in the environment projected on image planes are detected. Then, the reconstructed keypoints are used to identify the roadway for mobile robots. In the second approach, a camera and a laser projector are adopted. The line made by laser projector on the observed image plane is first retrieved and then rebuilt in three-dimension space according to the geometry relationship between the camera and the laser projector. The roadway is determined accordingly. In this thesis, two approaches are proposed and implemented through actual applications on a mobile robot in a real laboratory environment while control rules are designed based on the relationship between the roadway and the mobile robot to help navigate the robot in such an unknown environment.

並列關鍵字

Stereo Vision ； Map Building ； Mobile Robot ； Mobile Robot Navigation ； vSLAM

參考文獻

[1] Wen-Chung Chang. Hybrid force and vision-based contour following of planar robots.

Journal of Intelligent and Robotic Systems, 47(3):215–237, November 2006.

[2] Wen-Chung Chang. Binocular vision-based trajectory following for autonomous

[3] Wen-Chung Chang. Precise positioning of binocular eye-to-hand robotic manipulators.

Journal of Intelligent and Robotic Systems, 49(3):219–236, July 2007.