研究生: |
蘇心瑀 SIN-YU SU |
---|---|
論文名稱: |
基於機器視覺的多個指針式儀表自動識別方法 Auto-recognition Method for Multiple Pointer-type Meters Based on Machine Vision |
指導教授: |
林淵翔
Yuan-Hsiang Lin |
口試委員: |
陳維美
Wei-Mei Chen 林昌鴻 Chang-Hong Lin 沈中安 Chung-An Shen 林淵翔 Yuan-Hsiang Lin |
學位類別: |
碩士 Master |
系所名稱: |
電資學院 - 電子工程系 Department of Electronic and Computer Engineering |
論文出版年: | 2020 |
畢業學年度: | 108 |
語文別: | 中文 |
論文頁數: | 52 |
中文關鍵詞: | 影像處理 、深度學習 、儀表辨識 |
外文關鍵詞: | Image processing, Deep learning, Meters recognition |
相關次數: | 點閱:188 下載:0 |
分享至: |
查詢本校圖書館目錄 查詢臺灣博碩士論文知識加值系統 勘誤回報 |
在工廠,指針式的儀表具有容易觀察動態變化、維修成本較低、在電磁環境下無干擾等優勢。除此之外,大多數的被測量機台有未數位化、體積龐大不易更換等問題,造就工廠不願更換成電子式儀表,仍以人工的方式去讀取。為了減少人力資源,可利用攝影鏡頭取代人眼來進行即時地監控,不僅能提升效率,還可以改善人眼所造成的讀取誤差。
本論文著重於工廠常見的指針式壓力表,提出一套可同時辨識多個且多樣式之指針式壓力儀表演算法。利用YOLOv3網絡建立一套可偵測儀表位置的模型,接著使用影像處理取得儀表中的主要座標,即中心、最大刻度、最小刻度以及指針,並進行極座標轉換,再使用拉格朗日插值法來達成儀表數值的辨識。
自動讀取儀表數值在傳統上都是使用影像處理技術,精準度上有一定水準,然而此技術通常針對單個且單樣式、單個且多樣式以及多個且單樣式這三種來辨識儀表。近年來因機器學習和深度學習的崛起,能同時辨識多個且多樣式的研究也慢慢浮出,但在精準度上還有改善的空間。而本論文結合了影像處理技術與深度學習,利用兩者優點,來達到同時多個且多樣式的儀表辨識。
本論文探討了不同光照度下的儀表數值辨識誤差,在一般日光燈環境下,照度約280 lx以上時,儀表數值辨識的誤差就可以保持一定的誤差範圍內,顯示本系統應可適用於工廠內的日光燈環境下。
In factories, the advantages of pointer meters are easier to observe the dynamic changes, with lower maintenance costs, and not interfered in the electromagnetic environment. In addition, most of the machines are not digitized, and some are too bulky to replace. Therefore, many factories are unwilling to replace the old pointer meters and still read them manually. In order to save manpower, the camera can be used to replace the naked eye for real-time monitoring, which can not only improve efficiency, but also reduce reading errors caused by the naked eye.
This thesis focuses on the common pointer pressure meters in factories, and proposes an algorithm that can recognize multiple meters of various styles at the same time. This system uses the YOLOv3 network to establish a model that can detect the position of the meters. The main coordinates of the meter, namely the center coordinate, the maximum scale coordinate, the minimum scale coordinate, and the pointer coordinate, are obtained by image processing. After the polar coordinate is transformed, the Lagrange polynomial method is used to determine the values of the meters.
Traditionally the technology of image processing has been used to read the meter value automatically with a certain degree of accuracy. However, this technique usually recognizes three types of meters: single meter with single style, single meter with multiple styles, and multiple meters with single style. In recent years, due to the rise of machine learning and deep learning, research on recognizing multiple meters with multiple styles at the same time has also appeared, but the accuracy still needs to be improved. Combining image processing and deep learning, this thesis takes advantage of both technologies to achieve multiple and multi-style meters recognition at the same time.
This thesis also discusses the error of meter value recognition in different illuminance. In the fluorescent environment, when the illuminance is above 280 lx, the error can be kept within a certain range. It shows that the system can be applied under the fluorescent lamp environment of the factories.
[1]聯合新聞網. Accessed: Apr. 6, 2020. [Online]. Available:https://udn.com/news/story/7238/4116963
[2]自由時報. Accessed: Apr. 6, 2020. [Online]. Available:https://news.ltn.com.tw/news/life/breakingnews/2940681
[3]X. Ye, D. Xie and S. Tao, “Automatic Value Identification of Pointer-Type Pressure Gauge Based on Machine Vision,” Journal of Computers, vol. 8, no. 5, May 2013, pp. 1309-1314
[4]B. Yang, G. Lin and W. Zhang, “Auto-recognition Method for Pointer-type Meter Based on Binocular Vision,” Journal of Computers, vol. 9, no. 4, Apr. 2014, pp. 787-793
[5]L. Zhang et al., “Pointer-type meter automatic reading from complex environment based on visual saliency,” in Proc. ICWAPR., Jul. 2016, pp. 264-269
[6]Y. Ma et al., “An automatic reading method of pointer instruments,” in Proc. CAC., Oct. 2017, pp. 1448-1453
[7]W. Zheng et al., “Development of an automatic reading method and software for pointer instruments,” in Proc. EIIS., Jun. 2017, pp. 1-6
[8]C. Yang et al., “Design and Implementation of Pointer-Type Multi Meters Intelligent Recognition Device Based on ARM Platform,” Journal of Physics: Conference Series, vol. 960, Jan. 2018, pp. 1-7
[9]Y. Ma and Q. Jiang, “A robust and high-precision automatic reading algorithm of pointer meters based on machine vision,” Measurement Science and Technology, vol. 30, no. 1, Dec. 2018, pp. 1-11
[10]Y. Rao et al., “A Real-time Auto-recognition Method for Pointer-meter under Uneven Illumination,” in Proc. RCAR., Aug. 2019, pp. 890-894
[11]Y. Zhang et al., “Reading recognition of pointer meter based on pattern recognition and dynamic three-points on a line,” in Proc. ICMV., Mar. 2017, pp. 101-106
[12]M. Yi et al., “A clustering-based algorithm for automatic detection of automobile dashboard,” in Proc. IECON 2017 - 43rd Annual Conference of the IEEE Industrial Electronics Society, Nov. 2017, pp. 3259-3264
[13]R. A. Sowah et al., “Intelligent Instrument Reader Using Computer Vision and Machine Learning,” in Proc. IAS., Sep. 2018, pp. 1-7
[14]C. Dai et al., “Intelligent Ammeter Reading Recognition Method Based on Deep Learning,” in Proc. ITAIC., May 2019, pp. 25-29
[15]P. He et al., “A Value Recognition Algorithm for Pointer Meter Based on Improved Mask-RCNN,” in Proc. ICIST., Aug. 2019, pp. 108-113
[16]Y. Fang et al., “A Mask RCNN based Automatic Reading Method for Pointer Meter,” in Proc. CCC., Jul. 2019, pp. 8466-8471
[17]韓思之, “基於機器學習的數字儀表辨識方法,” in 國立臺灣科技大學, Aug. 2019
[18]J. Redmon and A. Farhadi, “Yolov3: An Incremental Improvement,” Accessed: Apr. 7, 2020. [Online]. Available: https://arxiv.org/abs/1804.02767
[19]J. Redmon et al., “You Only Look Once: Unified, Real-Time Object Detection,” in Proc. CVPR., Jun. 2016, pp. 779-788
[20]J. Redmon and A. Farhadi, “YOLO9000: Better, Faster, Stronger,” in Proc. CVPR., Jul. 2017, pp. 6517-6525
[21]T. Lin et al., “Feature Pyramid Networks for Object Detection,” in Proc. CVPR., Jul. 2017, pp. 936-944
[22]C. Tomasi and R. Manduchi, “Bilateral filtering for gray and color images,” in Proc. Sixth International Conference on Computer Vision (IEEE Cat. No.98CH36271), Jan. 1998, pp. 839-846
[23]賴屹民, “OpenCV3學習手冊,” in 碁封資訊股份有限公司, Nov. 2017, ch.10, pp.258-259
[24]賴屹民, “OpenCV3學習手冊,” in 碁封資訊股份有限公司, Nov. 2017, ch.10, pp.250-252
[25]鐘國亮, “影像處理與電腦視覺,” in 臺灣東華書局股份有限公司, Sep. 2016, ch.2, pp.32-39
[26]賴屹民, “OpenCV3學習手冊,” in 碁封資訊股份有限公司, Nov. 2017, ch.10, pp.264-273
[27]T. Y. Zhang and C. Y. Suen, “A Fast Parallel Algorithm forThinning Digital Patterns,” Communications of the ACM, vol. 27, no.3, Mar. 1984, pp. 236-239
[28]K. Wu, E. Otoo and A. Shoshani, “Optimizing connected component labeling algorithms,” in Proc. Medical Imaging 2005: Image Processing, 2005, pp. 1965-1976
[29]林紀瑋, “影像處理的13堂課,” in 博碩文化股份有限公司, Oct. 2011, ch.10, pp.(10-20)-(10-22)
[30]W. Sanford, “Applied linear regression,” in John Wiley & Sons Inc., Dec. 2013, ch.2, pp.24-26
[31]D. H. Ballard, “Generalizing the Hough transform to detect arbitrary shapes,” Pattern Recognition, vol. 13, no.2, 1981, pp. 111-122
[32]賴屹民, “OpenCV3學習手冊,” in 碁封資訊股份有限公司, Nov. 2017, ch.10, pp.339-342
[33]莊紹容, 楊精松, “微積分觀念與解析,” in 東華書局股份有限公司, May 2016, ch.7, pp.367
[34]吳俊諆, “應用數值方法:使用MATLAB,” in 美商麥格羅希爾國際股份有限公司, Jan. 2018, ch.17, pp.379-380