智慧影像辨識技術之產業應用導入研發

Research and Development of Industrial Applications Using Intelligent Pattern Recognition Techniques

林子怡(ZI-YI LIM)

指導教授 ： 曾顯文 ； 廖珗洲

朝陽科技大學/資訊學院/資訊管理系/博士(2020年)

摘要

當今影像處理相關技術已經發展的非常成熟，影像傳輸速度、品質足以廣泛應用於各種產業需求中，例如工業製造之自動光學檢測與量測、保全系統之自動監視系統、醫學影像之檢測與分析、無人駕駛汽車等各種領域之應用。本研究利用影像辨識技術與人工智慧實現於兩套系統並且應用於產業界，第一部分為高速追蹤之藥丸自動光學檢測系統，第二部分為智慧影像自動代理操作之系統。
第一部分，高速追蹤之藥丸檢測系統透過自動光學檢測技術實現，自動光學檢測(AOI: Automated Optical Inspection)是生產製造流程中確保產品品質的重要技術，對於具有高速與高準確度的生產流程來說，AOI檢測技術更是重要，因為在這種情況下，使用人眼進行視覺檢測已經幾乎是不可能的。因此，本研究開發一個高速追蹤之藥丸自動光學檢測系統，與台灣一間國際上包裝與標記設備的領導廠商配合，其藥丸「數粒填充機」的設備，具有最高每分鐘6,000顆藥丸的填充速度，在本研究中已開發一套高速AOI系統來剔除有瑕疵的藥丸。這裡運用圖形辨識與物體追蹤的方法來達成高速檢測的需求。開發之AOI可以避免瑕疵藥丸被填充到藥瓶中，以確保藥瓶的填充品質，所實作的技術可以達成高速製造、提供高精密度的檢測以及降低人工成本。
第二部分，智慧影像自動代理操作之系統(簡稱代操系統)，對於重複性高且單純的操作，本研究提出了一個自動化代理操作系統，其為一套軟硬體結合的解決方案，能透過事先設定好的自動化腳本與影像辨識模組搭配，精準控制鍵盤滑鼠執行代操，達到真正的全自動化操作，而出現非代操系統預期的狀況時，也能藉由多人多機機制通知人員介入處理。此外，本研究也融入模組化的軟體設計，讓腳本設定具備高度的可擴充性，可針對不同的需求與應用動態地設定客製化腳本，使得產製過程的管理更加容易並且更有效率。本研究所實作的代理操作系統已導入代工製程產業，基於舊式並需要大量人工操作的設備，藉由代操系統可以取代大多數的作業人員，完成許多繁瑣的動作，達成產業自動化升級。
本研究所開發之系統均已完成測試並且導入生產線上使用，確實解決產業問題並符合產業需求，系統也隨著生產線的調整進行系統調校以及優化。

關鍵字

自動光學檢測 ； 圖形識別 ； 移動物體追蹤 ； 樣板匹配 ； 自動代理操作 ； 產業應用

並列摘要

Nowadays, image processing is widely used in various fields. The research will propose two part of systems, it will be realized through pattern recognition methods and artificial intelligence technology, in order to solve the practical problems that need to be applied in the industrial field.
The first part is the “High Speed Tablet Automated Optical Inspection(AOI) System”. Automated optical inspection (AOI) is an important technique in the manufacturing process to ensure the quality of products. AOI is also especially important when the manufacturing process is in high speed and high accuracy operations. Visual inspection by the operators is difficult to handle under such conditions, a high-speed tracking tablet AOI system is expected to be developed. This project is cooperated with a company that is one of the leading packaging and labeling equipment manufacturers in the world. An AOI system will be designed for the tablet counting and filling system. The maximum filling speed is about 6,000 tablets per minute. Three kinds of defects, called NG (not good) tablets, will be detected by the AOI system, including broken tablet, size variation, and color variation, in order to increase the filling quality. Several pattern recognition and object tracking methods are utilized to meet the high-speed inspection requirement.
The second part is the “Smart Automatic Image Processing Agent Operating System”. The highly repeatable and simple operation can be replaced by this agent operating system. The pre-configured automation script is design with the image recognition module to control the keyboard and mouse accurately, to achieve a real automated operation. When the unexpected situation occurs, our system able to notify the operator to handle through the multi-person multi-machine mechanism. The script setting of operating system is also highly scalable and extensible, which can also customize the script according to difference types of requirement. It makes the management of production process become easier and more efficient.
Two systems have already deployed and integrated into the production line. They solve the industry issues and satisfy the industry demands. They are refined and sophisticated with the change of the production line.

並列關鍵字

Automated optical inspection ； Pattern recognition ； Object tracking ； Pattern match ； Automatic agent operation ； Industrial application

參考文獻

[1] N. Bhardwaj, S. Agarwal, and V. Bhardwaj (2015), “An imaging approach for the automatic thresholding of photo defects,” Pattern Recognition Letters, Vol. 61, pp. 32-40.

[2] M. H. Hung and C. H. Hsieh (2015), “A novel algorithm for defect inspection of touch panels,” Image and Vision Computing, Vol. 41, pp. 11-25.2

[3] C. F. Kuo, K. C. Peng, H. C. Wu, and C. C. Wang (2015), “Automated inspection of micro-defect recognition system for color filter,” Optics and Lasers in Engineering, Vol. 70, pp. 6-17.

[4] S. W. Yang, C. S. Lin, S. K. Lin, and H. T. Chiang (2014), “ Automatic defect recognition of TFT array process using gray level co-occurrence matrix,” Optik, Vol. 125, pp. 2671-2676.

[5] X. W. Zhang, Y. Q. Ding, Y. Y. Lv, A. Y. Shi, and R. Y. Liang (2011), “A vision inspection system for the surface defects of strongly reflected metal based on multi-class SVM,” Expert Systems with Applications, Vol. 38, pp. 5930-5939.