發展應用深度學習之傳統儀表影像判讀技術

Development of an Image Interpretation Technique for Traditional Gauges Using Deep Learning

林怡德(Yi-Der Lin)

指導教授 ： 陳冠宇

中原大學/工學院/機械工程研究所/碩士(2020年)

https://doi.org/10.6840/cycu202000887

摘要

本研究目的在於建立傳統機台之自動化系統。傳統工廠中不同機台間沒有訊號溝通，需要透過人工進行操作及紙本資料的收集，在沒有足夠資金添購自動化設備的情況下，不僅耗費人力進行數據收集且如無後續建檔程序，數據亦無法使用電子系統進行整合分析及統計，本研究希望透過架設攝影機，利用機器視覺判讀儀表數值，將不同機台之數值收集彙整於電腦、網路或是手機軟體裡，除可以將數值電子化，並且也可使公司進一步發展機台之風險控制及建置警示系統。
本文利用霍夫偵測、機器視覺及深度學習，對工廠機台之儀表讀數進行辨識，此監控系統主要將儀表分成兩部份進行判讀：指針式及七段顯示器。指針式主要利用霍夫偵測找出指針，並計算角度求出數值；七段顯示器利用深度學習辨識，藉由事前訓練數字0~9產生模型，並以此模型進行判斷，再將兩者數值顯示於圖形使用者介面，使使用者可遠端知悉機台數據。

關鍵字

深度學習 ； 機器學習 ； 機器視覺 ； 霍夫偵測 ； python ； opencv

並列摘要

The purpose of this research is to build an automation system for traditional machines. In the traditional factory, there is no signal communication between different machines, therefore it is need to be operated and collected on paper manually. Without sufficient funds to purchase automation equipment, not only is it labor-intensive to collect data, but without subsequent documentation procedures, the data cannot be integrated for analysis and statistics using an electronic system. This research hopes by setting up a camera, using machine vision to recognize the value of the gauges on the instrument, and collecting the values of different machines in a computer, network or mobile phone software. In addition to digitizing the values, it can also enable the company to further develop risk control and warning system.
In this paper, Hough detection, machine vision and deep learning are used to identify the gauge readings of the factory machine. This system mainly divides the gauge into two parts, pointer type and seven-segment display. The pointer type mainly uses Hough detection to find the pointer and calculate the angle to find the value. The seven-segment display uses deep learning recognition to generate a model by training numbers 0-9 beforehand, and judges based on this model. Then, the two values are displayed on the graphical user interface so that the user can know the gauge data remotely.

並列關鍵字

deep learning ； machine learning ； machine vision ； Hough transformation ； python ； opencv

參考文獻

[1]P. V. Hough, “General Purpose Visual Input for a Computer,” 1962

[2]R. O. Duda, P. E. Hart, “Use of the Hough transformation to detect lines and curves in pictures.” Communications of the ACM, 1972, 15.1:11-15.

[3]D.H.Ballard, “Generalizing the Hough transform to detect arbitrary shapes.” Pattern Recognition Vol.13, pp.111-122, 1981

[4]C. R. Dyer, “Gauge Inspection Using Hough Transforms”, IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. PAMI-5 , Issue: 6 , 1983

[5]Y. LeCun, L. Bottou, Y. Bengio, and P. Haffner, "Gradient-based learning applied to document recognition," Proceedings of the IEEE, vol. 86, pp. 2278-2324, 1998