基於影像特徵辨識之機械手臂模糊控制系統之研究

自動化、智能化科學研究為近年工業發展趨勢，結合機械手臂與計算機視覺於精密製程系統之應用。透過電腦於物件特徵與位置相關資訊進行分析處裡，達至相應的定位、搬運及篩選等任務，已成為工業應用主軸。本文主軸以特徵辨識與模糊控制系統進行研究探討，實驗過程中以直角座標型機器手臂進行操作，特徵辨識系統應用Emgu CV計算機視覺程式庫，模糊控制系統應用Aforge.Net人工智慧程式庫於Visual Studio開發環境平台中進行程式設計，最後將各系統進行軟硬體統一整合。運用特徵辨識定位及模糊控制之技術，完成機械手臂作動控制與路徑規劃，而達成機械手臂的自主夾取。本研究以多邊形物件測試特徵辨識系統之影像特徵辨識與定位，並以SIFT(Scale-Invariant Feature Transform)與SURF(Speeded-Up Robust Features)演算法互相比較。根據檢測特徵與匹配，查看特徵量、運算時間及穩定性，相互比較與驗證，在同條件下，SURF演算法特徵量平均為SIFT的70%、運算時間平均為SIFT的51%、穩定性平均為SIFT的85%，雖然其穩定性較低，但還在容許範圍內，且其即時的特性較符合本旨需求，最終將SURF方法應用於本研究系統之中，並運用模糊控制有效地使控制系統之操作更為精準細微。期許研究成果能有效改善製程中的定點定位作動此項步驟，於未來大批量生產及檢測的製程中，提升生產線之效率與縮短生產週期，實現改善製程之目標。關鍵字：機械手臂、特徵辨識、模糊控制、Emgu CV、Aforge.Net

關鍵字

機械手臂；特徵辨識；模糊控制； Emgu CV ； Aforge.Net

並列摘要

Intelligent automation has become the current trend in scientific research and industrial development, combining robotic arms and computer vision in precision process systems. Using computers to analyze and process object features and locations and achieve tasks such as positioning, transporting, and selecting objects has become the focus of industrial applications. The objective of this study was to develop a feature recognition and fuzzy control system. During the experiment, we used a right-angle coordinate robot. The feature recognition system employed the Emgu CV library, and program design of the fuzzy control system was performed on Visual Studio using the Aforge.Net artificial intelligence library. The software and hardware of the systems were then integrated. Actuation control and route planning of the robotic arm were completed using the feature recognition and fuzzy control technology, thereby enabling the robotic arm to retrieve objects autonomously. We used polygonal objects to test the recognition and positioning functions of the feature recognition system and compared the SIFT (Scale-Invariant Feature Transform) and SURF (Speeded-Up Robust Features) feature recognition methods. According to the detection characteristics and matching, the feature quantity, operation time and stability are checked and compared with each other. Under the same conditions, the SURF algorithm features an average of 70% of SIFT, an arithmetic time average of 51% of SIFT, and an average stability of 85% of SIFT. Although its stability is low, it is still within the allowable range, and its immediate characteristics are more in line with the purpose. Finally, the SURF method is applied to the research system, and used fuzzy control to enhance the precision of control system operations. It is hoped that the achievements of this study can effectively improve the fixed-point positioning operation in the process and can be applied in mass production and testing in the future so increase production line efficiency, shorten production cycles, and improve processes. Keywords : robotic arm, feature recognition, fuzzy control, Emgu CV, Aforge.Net