臺灣博碩士論文加值系統

本研究以Delta型機械手結合影像辨識及機器學習應用於飛抓系統。現今各類型產業裡常需使用影像辨識進行物品識別，再搭配合適的機械手臂來進行物料分類、產品搬移、物件組合等功能，特別近年來機器學習亦廣泛發展應用，在各類產業逐步擴展提供更好的效能。
此次研究目的為建立一套利用影像辨識進行飛抓系統應用，透過市售webcam在輸送帶辨識物體中心後進行倒數，物體隨著輸送帶往前輸送，當倒數完成即觸發機械手臂飛抓，利用機器學習將時間預測值回傳飛抓系統，達成回授控制自動調整觸發時間作出最佳的飛抓時機，完成機械手臂飛抓系統之流程提升。

In this study, Delta-type manipulator combined with image recognition and machine learning was applied to the flying grasping system. Nowadays, it is necessary to use image recognition for object recognition in various types of industries, and then use appropriate robotic arms to perform functions such as goods classification, product movement, and object assembly. Especially in recent years, machine learning has been widely developed and applied, and it has been gradually expanded in various industries for better performance.
The purpose of this research is to establish a set of application of a flying grab system using image recognition. After identifying the center of object on the conveyor belt through a commercially available webcam, when the countdown is completed. The object is transported forward with the conveyor belt. Using machine learning to return the time prediction value to the flying grasping system to achieve feedback control and automatically adjust the trigger time to make the best flying grasping timing, and complete the process improvement of the robotic arm flying grasping system.

書頁名 I
摘 要 II
Abstract III
誌謝 IV
目錄 V
表目錄 VII
圖目錄 VIII
符號說明 X
第一章、緒論 11
1.1研究動機與背景 11
1.1.1飛抓系統介紹 12
1.1.2飛抓系統流程分析 14
1.2 文獻回顧 15
1.2.1 國外文獻 15
1.2.2 國內文獻 20
1.3 論文目標 22
第二章 系統架構與軟硬體介紹 23
2.1系統架構 23
2.2 硬體及模擬軟體介紹 24
2.2.1 影像處理端硬體 24
2.2.2 輸送帶硬體 25
2.2.3 FANUC M-1ia/0.5S 25
2.2.4 FANUC R-30B控制器 26
2.2.5 FANUC Roboguide模擬軟體 27
2.2.6主程式語言：C# 27
第三章 AForge.NET影像辨識之應用 29
3.1 概述 29
3.2 Sobel Edge Detector 29
3.2.1 Sobel Edge Detector圖像處理 30
3.3 Blob Counter 31
3.3.1 Blob Counter圖像處理 33
3.4 Euclidean Color Filtering 34
3.4.1 Euclidean Color Filtering圖像處理 34
第四章 機器學習與其在飛抓演算法之應用 35
4.1概述 35
4.1.1 ML. NET 機器學習可使用的類型 35
4.2 ML. NET 程式碼工作流程 36
4.2.1 ML. NET 模型建立器 37
4.3 ML. NET演算法 38
4.3.1 ML. NET 評估指標 39
4.3.2 ML. NET 提高模型準確性 40
4.4 ML. NET在飛抓系統之應用 41
第五章 實驗結果 43
5.1 飛抓系統運作環境概述 43
5.2 飛抓系統實驗結果 44
5.3 飛抓系統進行機器學習實驗結果 46
第六章、結論 52
6.1 成果貢獻 52
6.2 未來展望 53
參考文獻 54

[1] 上銀機械手臂Delta飛抓 https://www.youtube.com/watch?v=HLfKOtgOrWU
[2] Qiang Bai, Shaobo Li, Jing Yang ，Qisong Song , Zhiang Li1 , and Xingxing Zhang(2020, October)，”Object Detection Recognition and Robot Grasping Based on Machine Learning: A Survey”, Digital Object Identifier 10.1109/ACCESS.2020.3028740, IEEE.
[3] Jacques Janse Van Vuuren , Liqiong Tang, Ibrahim Al-Bahadly , and Khalid Mahmood Arif(2020, April), “A 3-Stage Machine Learning-Based Novel Object Grasping Methodology”, Digital Object Identifier 10.1109/ACCESS.2020.2987341, IEEE.
[4] Shoujun Wang, Xingmao Shao, Liusong Yang , and Nan Liu(2020, July),“Deep Learning Aided Dynamic Parameter Identification of 6-DOF Robot Manipulators”, Digital Object Identifier 10.1109/ACCESS.2020.3012196,IEEE.
[5] Guanglong Du , Zhiyao Wang , and Zhelin Li(2020, January)，” A New Cloud Robots Training Method Using Cooperative Learning” , Digital Object Identifier 10.1109/ACCESS.2020.2966226,IEEE.
[6] 吳宜儒(2019),基於深度學習與立體影像之機械手臂智慧夾取,國立清華大學動力機械工程學系學位論文，pp.1-57。
[7] 蘇立珩(2018),基於機器學習與影像之機械手臂抓取,國立清華大學動力機械工程學系學位論文，pp.1-58。
[8] 林祐生(2019)，深度學習在機器人視覺辨識中的應用,國立虎尾科技大學機械工程系學位論文，pp.1-60。
[9] AForge.NET http://www.aforgenet.com/
[10] Sobel https://zh.wikipedia.org/wiki/%E7%B4%A2%E8%B2%9D%E7%88%BE%E7%AE%97%E5%AD%90
[11] Sobel Edge Detector http://www.aforgenet.com/framework/features/edge_detectors_filters.html
[12] Blob Counter
http://www.aforgenet.com/framework/features/blobs_processing.html
[13] Hui Kong, Hatice Cinar Akakin and Sanjay E. Sarma, “A Generalized Laplacian of Gaussian Filter for Blob Detection and Its Applications,” IEEE Transactions on Cybernetics, vol. 43, no. 6, Dec. 2013.
[14] Euclidean Color Filtering
http://www.aforgenet.com/framework/features/color_filtering.html
[15] ML.NET https://docs.microsoft.com/zh-tw/dotnet/machine-learning/how-does-mldotnet-work
[16] Jarred Capellman,(2020), Hands-on Machine Learning with ML.NET:Getting started with Microsoft ML.NET to implement popular machine learning algorithms in C#，,Packt Publishing Ltd. ,UK, pp.44-69.
[17] 波式散佈圖https://zh.wikipedia.org/wiki/%E5%8D%9C%E7%93%A6%E6%9D%BE%E5%88%86%E5%B8%83
[18] 波式迴歸模組參數https://docs.microsoft.com/zh-tw/azure/machine-learning/studio-module-reference/poisson-regression