在影像量測辨識領域中，使用機器視覺辨識儀表是一重要的研究領域。其中儀表辨識的目標為類比式電表、水表、瓦斯表及各式工廠的儀表，通常採用人力檢測數值，然而各式儀表龐大的使用範圍與數量，都導致長期監控效率低下，造成人力浪費。故逐漸將傳統的類比式儀表更換成智慧型儀表。但是其更換成本是一大考慮的重點，同時許多工業界儀表也有多種考量，例如體積笨重、更換成本高昂、及智慧型儀表可能受環境電磁場影響致使精度下降等原因，而無法更換成智慧型儀表。
使用機器視覺影像量測方法就成了過度期的解決方法，傳統方法通常使用影像處理技巧擷取目標輪廓特徵，但缺點是這類方法通常具有許多限制條件，無法適用於複雜環境與多重格式的儀表影像量測。故後來研究方向逐漸轉向使用深度學習網路訓練複雜環境的儀表影像量測模型。
因此，本論文建立一套基於機器學習的數字儀表量測系統，使用640×480的RGB網路攝影機讀取影像，利用系統內部的二階段方法偵測儀表區域及辨識數字數值，進行數字儀表的影像量測。不同於傳統影像處理方法只能處理單一格式的儀表量測，本系統結合兩個深度卷積網路訓練模型量測儀表影像數值，可以量測複雜性的儀表數據集，提升量測目標的泛化性。
由於深度學習網路通常需要大量數據資料來輔助特徵的擷取，所以需要人工Label大量的數據，此步驟通常耗費大量時間與心力，在本篇論文中，提出一種合成訓練資料的方法，目前實驗結果在不降低準確度的前提下提升製作訓練資料的效率。

關鍵字：影像處理，儀表辨識，機器學習。

In the field of image measurement and identification, the use of machine vision identification instrument is an important research field. Among them, the object of meter identification is analog meter, water meter, gas meter and various factory instruments. The manpower detection value is usually used. However, the large scope and quantity of various instruments lead to low efficiency of long-term monitoring and waste of manpower. Therefore, the traditional analog meter is gradually replaced with a smart meter. However, the cost of replacement is a major consideration. At the same time, many industrial instruments have many considerations, such as bulky, high replacement costs, and intelligent instruments may be affected by environmental electromagnetic fields, resulting in reduced accuracy, etc., and cannot be replaced with smart meters.
The use of machine vision image measurement methods has become an overdue solution. Traditional methods usually use image processing techniques to capture target contour features, but the disadvantage is that such methods usually have many limitations and cannot be applied to complex environments and multiple formats. Instrument image measurement. Therefore, the research direction gradually turned to the use of deep learning networks to train instrument image measurement models in complex environments.
Based on the above research, this thesis establishes a digital instrument measurement system based on machine learning. The system uses 640x480 RGB network camera to read images, and uses the internal two-stage method to detect the instrument area and identify the digital classification values. Image measurement of digital meters. Different from the traditional image processing method, it can only process meter measurement in a single format. The system combines two deep convolutional network training models to measure the image value of the instrument. It can also measure and improve the measurement of the instrument data set of complexity. The generalization of the goal.
Since deep learning networks usually require a large amount of data to assist in feature extraction, a large amount of data is required for manual Label. This step usually takes a lot of time and effort. In this paper, a method for synthesizing training materials is proposed. Improve the efficiency of production training materials while reducing accuracy.

Key word : Image Process, Meter Recognition, Machine Learning

[1] 法新社,哥斯大黎加國家電力協會計劃安裝15.8萬個智慧電表,”[Online].
Available:https://www.digitimes.com.tw/iot/article.asp?cat=158&id=0000496488_n20513in7kijzj9z7yvbg
[Accessed on 28 Jun. 2019].
[2] 臺北自來水事業處, "漏水現況與防治成果," [Online].
Available:https://www.water.gov.taipei/cp.aspx?n=E69EC16A68D92530
[Accessed on 28 Jun. 2019].
[3] 中華民國自來水協會, "智慧水表之建置與應用發展," [Online].
Available:http://www.ctwwa.org.tw/download.aspx?AttachmentFile=SysUpDo
c/message/newone/9704C4E71E1A4B2.pdf
[Accessed on 28 Jun. 2019].
[4] S. Rasheed et al., "Automated Number Plate Recognition Using Hough Lines and Template Matching," Proceedings of the World Congress on Engineering and Computer Science, vol. 1, pp. 24-26, Oct. 2012.
[5] S. Zhao et al., “Research on Remote Meter Automatic Reading Based on Computer Vision,” in Proc. Transmission and Distribution Conference and Exhibition, Feb. 2005, pp.1-4.
[6] D. Shu et al., “Study of The Automatic Reading of Watt Meter Based on Image Processing Technology,” in Proc. 2nd IEEE Conference on Industrial Electronics and Applications, May 2007, pp. 2214-2217.
[7] B.V. Kakani et al., "Improved OCR Based Automatic Vehicle Number Plate Recognition Using Features Trained Neural Network,"in Proc. 8th International Conference on Computing, Communication and Networking Technologies (ICCCNT), Jul. 2017, pp. 1-6.
[8] I. Gallo et al., “Robust Angle Invariant GAS Meter Reading,” in Proc. International Conference on Digital Image Computing: Techniques and Applications (DICTA), Nov. 2015, pp. 1-7.
[9] D. B. P. Quintanilha et al., “Automatic Consumption Reading on Electromechanical Meters Using HoG and SVM,”in Proc.7th Latin American Conference on Networked and Electronic Media, Nov. 2017, pp. 57-61.
[10] Y. Gao et al., “Automatic Watermeter Digit Recognition on Mobile Devices,” Internet Multimedia Computing and Service, vol. 819, pp. 87–95, Mar. 2018.
[11] R. Laroca et al., “Convolutional Neural Networks for Automatic Meter Reading,” Journal of Electronic Imaging, vol. 28, no. 1, pp. 013023-1-013023-14, Feb. 2019.

[12] M. Rodriguez et al., “HD MR: A New Algorithm for Number Recognition in Electrical Meters,” Turkish Journal of Electrical Engineering and Computer Sciences, vol. 22, no. 1, pp. 87-96, Oct. 2012.
[13] L. A. Elrefaei et al., “Automatic Electricity Meter Reading Based on Image Processing,” in Proc, IEEE Jordan Conference on Applied Electrical Engineering and Computing Technologies (AEECT), Nov. 2015, pp. 1-5.
[14] Y. Zhang et al., “Automatic Reading of Domestic Electric Meter: An Intelligent Device Based on Image Processing and ZigBee/Ethernet Communication,” Journal of Real-Time Image Processing, pp. 133-143, Jun. 2016.
[15] A. Anis et al., “Digital Electric Meter Reading Recognition Based on Horizontal and Vertical Binary Pattern,” in Proc, 3rd International Conference on Electrical Information and Communication Technology (EICT), Dec. 2017, pp.1-6.
[16] V. C. P. Edward et al., “Support Vector Machine Based Automatic Electric Meter Reading System,” in Proc. IEEE International Conference on Computational Intelligence and Computing Research, Dec. 2013, pp.1-5.
[17] B. Shi et al., “An End-to-End Trainable Neural Network for Image-Based Sequence Recognition and Its Application to Scene Text Recognition,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 39, no. 11, pp. 2298-2304, Dec. 2016.
[18] C. Szegedy et al., "Going Deeper with Convolutions," in Proc. Computer Vision and Pattern Recognition(CVPR), Jun. 2014, pp. 1-9.
[19] K. He et al., "Deep Residual Learning for Image Recognition," in Proc. Computer Vision and Pattern Recognition(CVPR), Jun. 2016, pp. 770-778.
[20] T. Y. Lin et al., "Focal Loss for Dense Object Detection," in Proc. Computer Vision and Pattern Recognition(CVPR), Jun. 2017, pp. 2980-2988.
[21] S. Ren et al. "Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 39, no. 6, pp. 1137-1149, Jun. 2017.
[ 22] J. Redmon et al., "You Only Look Once: Unified, Real-Time Object Detection," in Proc. Computer Vision and Pattern Recognition(CVPR), Jun. 2016, pp. 779-788.
[23] W. Liu et al., "Ssd: Single Shot Multibox Detector," in Proc. European Conference on Computer Vision, 2016, pp. 21-37.
[24] J. Ma et al., "Arbitrary-Oriented Scene Text Detection via Rotation Proposals," IEEE Transactions on Multimedia, vol. 20, no. 11, pp. 3111-3122, Nov. 2018.
[25] R. Laroca et al., “Real-Time Automatic License Plate Recognition Through Deep Multi-Task Networks,” in Proc. 31st SIBGRAPI Conference on Graphics, Patterns and Images (SIBGRAPI), Oct. 2018, pp. 110–117.
[26] N. Otsu, "A Threshold Selection Method from Gray-Level Histograms," IEEE Transactions on Systems, Man, and Cybernetics, vol. 9, no. 1, pp. 62-66, 1979.
[27] J. Redmon and A. Farhadi, "YOLO9000: Better, Faster, Stronger," in Proc. Computer Vision and Pattern Recognition(CVPR), Jun. 2017, pp. 7262-7271.
[28] J. Redmon and A. Farhadi, "Yolov3: An Iincremental Improvement," [Online].
Available:https://arxiv.org/abs/1804.02767
[Accessed on 28 Jun. 2019].
[29] C. Szegedy et al., "Rethinking the Inception Architecture for Computer Vision," in Proc. Computer Vision and Pattern Recognition(CVPR), Jun. 2016, pp. 2818-2826.
[30] J. Deng et al., “ImageNet: A Large-Scale Hierarchical Image Database,”in Proc. Computer Vision and Pattern Recognition(CVPR), Jun. 2009, pp. 248–255.
[31] A. Gupta et al., "Synthetic Data for Text Localisation in Natural Images," in Proc. Computer Vision and Pattern Recognition(CVPR), Jun. 2016, pp. 2315-2324.
[32] P. Migdal, “Learning Deep Learning with Keras,”2017. [Online].
Avalible:https://p.migdal.pl/2017/04/30/teaching-deep-learning.html
[Accessed on 28 Jun. 2019].