本研究主要目的為開發一套結合3D視覺系統及六軸串聯式機械手臂，可進行修復汽車鈑金凹陷之全自動系統。此系統可藉由掃描板金凹陷區域和比對找出板金變形分布，並引導機械手臂對該位置進行鈑金修復工作。為找出汽車鈑金的凹陷分佈資訊，3D曲面匹配是整個系統的核心，本研究使用的技術是以迭代最近點(Iterative Closest Point, ICP)演算法作為基礎，並修改部分演算法以達到匹配兩個部分相似曲面的目的。
汽車板金三維曲面配對系統部分，先是透過RGB-D相機擷取結合RGB資訊及深度資訊並轉換成點雲，再使用深度濾波去除雜訊及環境點雲，保留目標物的點雲資訊進行匹配，可藉此節省配對演算法的時間，並有效提升ICP演算法的配對精度。另外，本研究使用基於ICP演算法之改良曲面匹配法執行匹配兩相似之對稱自由曲面。利用ICP演算法中去除不可靠點之特性，改變及分類點集之對應關係，達到只配對相似自由曲面之效果。匹配完成後便可根據板金凹陷分佈資訊，引導機械手臂夾取適當修復工具進行全自主修復。

The purpose of this study is to a 3D integrated vision system that can command the robotic arm to fully autonomous car sheet metal repairing, combines the 6-axis robot for car sheet metal repairing. This system can find out the deformation distribution of the sheet metal by scanning the recessed area of the sheet metal and comparing it, and guide the mechanical arm to perform the sheet metal repair work on the certain position. In order to find out the depression distribution information of sheet metal, 3D surface matching is the core of the entire system. The technology used in this study is based on the Iterative Closest Point (ICP) algorithm, and modified some parts of the algorithm to matching two similar surface.
In terms of the 3D surface matching system, this study captures the RGB information and depth information through an RGB-D camera and converts it into a point cloud, then uses depth filtering to remove noise and point clouds of the environment, and retains the information of the target point cloud for matching. In this way, it will save the time of the matching algorithm and effectively improves the matching accuracy of the ICP algorithm. In addition, this study uses an improved surface matching method based on the ICP algorithm to match two similar symmetric free-form surfaces. Using the feature of removing unreliable points in the ICP algorithm, changing and classifying the corresponding point set, achieve the effect of only matching the similar part of free-form surfaces. After the matching, the system will guide the robot to pick up the repair tools for fully autonomous repair the dent of the sheet metal.

1. 劉妃芳，台灣汽車售後維修市場發展之研究，碩士論文，逢甲大學經營管理碩士在職專班，2015
2. Gruen, A., & Akca, D. (2005). Least squares 3D surface and curve matching. ISPRS Journal of Photogrammetry and Remote Sensing, 59(3), 151-174.
3. Besl, P. J., & McKay, N. D. (1992, April). Method for registration of 3-D shapes. In Sensor fusion IV: control paradigms and data structures (Vol. 1611, pp. 586-606). International Society for Optics and Photonics.
4. Heikkila, J., & Silven, O. (1997, June). A four-step camera calibration procedure with implicit image correction. In Proceedings of ieee computer society conference on computer vision and pattern recognition (pp. 1106-1112). IEEE.
5. Hartley, R., & Zisserman, A. (2003). Multiple view geometry in computer vision. Cambridge university press.
6. 蔡高岳，機械系統設計，高立圖書，2004。
7. 飛時測距及結構光介紹，取自< https://reurl.cc/e89laL>
8. K-d樹基礎介紹，取自< https://en.wikipedia.org/wiki/K-d_tree>
9. ICP演算法基礎介紹，取自< https://reurl.cc/14gAWp>
10. 江宗錡，六軸關節型機械手臂之手眼校正研究，成功大學電機工程學系學位論文, 2014: p. 1-66.