輸電系統是發電廠與用戶端傳遞的橋樑，而絕緣礙子是連接鐵塔與傳輸高壓導線之間重要的絕緣裝置。如果絕緣礙子有損壞的情形，會影響到供電安全及品質，所以能更快速檢測絕緣礙子的狀態是我們的目標。本系統是將絕緣礙子運用深度學習中物件偵測的YOLO(You Only Look Once)演算法，建立絕緣礙子的辨識模型，可檢測與識別絕緣礙子的圖片檔、視訊檔及即時影像辨識。經由目標檢測系統能將閃絡、破損、汙染等有劣化的絕緣礙子快速被偵測出來。絕緣礙子物件偵測與識別系統可使維護人員不需每座鐵塔都登塔檢視以及縮短確認視訊、圖片時間並可快速找到弱點，可提升人員工作安全及工作效率，進而提供更安全的供電品質。

The transmission system is the bridge between the power plant and the customer, and the insulator is an important insulation device that connects the tower and the transmission high-voltage conductor. If the insulator is damaged, it will affect the safety and quality of power supply, so it is our goal to detect the status of the insulation barrier more quickly. The system uses the YOLO (You Only Look Once) algorithm for object detection in deep learning to establish an insulator recognition model, which can detect and identify picture files, video files and real-time image recognition of insulator. Through the object detection system, it can quickly detect the degradation of insulation barriers such as flashover, damage, and pollution. Insulation insulator object detection and identification system allows maintenance personnel to check the tower without having to check each tower and shorten the time to confirm video and pictures quality.

摘要......i
Abstract......ii
誌謝......iii
目錄......iv
表目錄......vi
圖目錄......vii
第一章 緒論......1
1.1研究背景與動機......1
1.2 研究目的與方法......1
1.3 論文架構......2
第二章 電力系統與架空輸電線路介紹......3
前言......3
2.1 電力系統簡介......3
2.2架空輸電線路組成介紹[2]......4
2.2.1支持物......4
2.2.2 架空導線......6
2.2.3 架空地線......7
2.2.4 絕緣礙子......7
2.2.5 其他配件......10
2.3 架空輸電線路維護[4]......11
第三章 深度學習之物件偵測與識別介紹......13
前言......13
3.1 深度學習介紹......13
3.2 物件偵測與識別介紹[7]......13
3.2.1 物件偵測應用與種類[8]......14
3.2.2 物件偵測測常用指標[10]......15
3.2.3 Faster R-CNN、SSD、YOLO演算法介紹......17
第四章 建構絕緣礙子辨識模型......27
前言......27
4.1 系統設備與軟體設置[9]......27
4.2 資料集準備與建立辨識模型[27]......28
4.2.1 訓練圖片收集......28
4.2.2 建立相關資料夾......30
4.2.3 標記圖片......30
4.2.4 訓練數據集準備[25]......33
4.2.5 建立辨識模型......38
4.3 結果測試與分析......40
第五章 結論......51
參考文獻......52

[1] 電力公司 “輸電線路維護養成班專業科學教材 第三冊” 訓練所講義
[2] 電力公司，1979年，〝輸電線路維護手冊〞，電力公司供電處，4月
[3] 電力公司編，〝輸電工程作業手冊〞，電力公司。
[4] 台電公司供電處編，維護作業程序工作指導書，2016年。
[5] A. Krizhevsky, I. Sutskever, and G. E.Hinton, “ImageNet Classificationwith
Deep Convolutional Neural Networks,” NIPS, 2012,doi:10.1201/9781420010749.
[6] A. Esmaeilzehi, M. O. Ahmad, and M. N. S. Swamy, ‘‘Compnet: A newscheme
for single image super resolution based on deep convolutional neural
network,’’IEEE Access, vol. 6, pp. 59963–59974, 2018.
[7] Z. Q. Zhao, P. Zheng, S. T. Xu, and X. Wu, “Object Detection with Deep
Learning: A Review,” IEEE Trans. Neural Networks Learn. Syst.,vol. 30,
no.11, pp. 3212–3232.
[8] 繆鵬 “CV+深度學習：AI最完整的跨套件Python人工智慧電腦視覺” 深智數位股份有限公司
2019.07
[9] 文淵閣工作室 “Pthon機器學習與深度學習特訓班” 碁峯資訊股份有限公司
[10] Tarang Shah , Measuring Object Detection models - mAP - What is Mean
Average Precision?, “https://tarangshah.com/blog/2018-01- 27/what-is-map-
understanding-the-statistic-of-choice-for- comparing-object-detection-
models/”, JAN .2018
[11] R. Girshick, ‘‘Fast R-CNN,’’ in Proc. IEEE Int. Conf. Comput. Vis.,Dec.
2015, pp. 1440–1448.
[12] S. Ren, K. He, R. Girshick, and J. Sun, ‘‘Faster R-CNN: Towards real-time
object detection with region proposal networks,’’ in Proc. Adv. Neural
Inf. Process. Syst., 2015, pp. 91–99
[13] S. Ren, K. He, R. Girshick, and J. Sun, “Faster R-CNN: Towards RealTime
Object Detection with Region Proposal Networks,” IEEE Trans.Pattern Anal.
Mach. Intell., vol. 39, no. 6, pp. 1137–1149.
[14] W. Liu, D. Anguelov, D. Erhan, C. Szegedy, S. Reed, C.-Y. Fu, and A. C.
Berg, ‘‘SSD: Single shot multibox detector,’’ in Proc. Eur. Conf. Comput.
Vis. (ECCV). Cham, Switzerland:Springer, 2016, pp. 21–37.
[15] 王得懿, SSD: Single Shot MultiBox Detector 介紹
“https://medium.com/@bigwaterking01/ssd-single-shot-multibox- detector-介
紹-1fe95073c1a3”, May 28, 2018
[16] J. Redmon, S. Divvala, R. Girshick, and A. Farhadi, “You only look once:
Unified, real- time object detection,” IEEE Comput. Soc. Conf.Comput.
Visr. Pattern Recognit., vol. 2016-Decem, pp.779–788.
[17] J. Redmon and A. Farhadi, ‘‘YOLOv3: An incremental improvement,’’Apr.
2018, arXiv:1804.02767. [Online]. Available:
https://arxiv.org/abs/1804.02767
[18] H. Yang, L. Chen, M.Chen, Z. Ma, F. Deng, M. Li, X. Li "Tender Tea Shoots
Recognition and Positioningfor Picking Robot Using Improved YOLO-V3 Model"
IEEE Int. Conf. Comput. Vis.11 December 2019,PP. 180998 – 181011.
[19] Tommy Huang https://medium.com/@chih.sheng.huang821/深度學習-物 件偵測
yolov1-yolov2和yolov3-cfg-檔解讀-75793cd61a01 Oct 23,2018
[20] 台部落山水之間, 目標檢測之YOLOv1到YOLOv3系列簡介,
“https://www.twblogs.net/a/5bb2e41e2b71770e645e2289”
[21] Mr. Opengate, Yolo：基於深度學習的物件偵測 (含YoloV3)
“https://mropengate.blogspot.com/2018/06/yolo-yolov3.html”
[22] Steven Shen, 關於影像辨識，所有你應該知道的深度學習模型,
“https://medium.com/cubo-ai/物體偵測-object-detection-740096ec4540” .
[23] 許哲豪, 如何以YOLOv3訓練自己的資料集-以小蕃茄為例,
“https://makerpro.cc/2019/12/train-your-dataset-with-yolov3/”
[24] 銀桑的雜貨鋪發表, 機器學習：聊聊機器學習模型中訓練集、驗證集和測試集 的關係,
“https://kknews.cc/news/b4n3bnm.html”
[25] Phinoo, YOLOV3實戰4：Darknet中cfg文件說明和理解,
“https://blog.csdn.net/phinoo/article/details/83022101”
[26] 安小漁兒發表, 如何用自己的數據訓練YOLOv3, “https://kknews.cc/zh
tw/tech/mpnq8yz.html”
[27] CH. Tseng, 建立自己的YOLO辨識模型– 以柑橘辨識為例,
“https://chtseng.wordpress.com/2018/09/01/建立自己的yolo辨識模 型-以柑橘辨
識為例/”.
[28] 爽朗, YOLOv3批量測試圖片並保存在自訂資料夾下,
“https://zhuanlan.zhihu.com/p/43824197”
[29] J. Redmon, YOLO:Real-Time Object Detection,
“https://pjreddie.com/darknet/yolo/”.
[30] Haoqimao_hand, YOLOV3+OpenCV實現目標檢測,
“https://blog.csdn.net/haoqimao_hard/article/details/82081285”