Basic Search / Detailed Display

Author: 陳凱呈
KAI-CHENG CHEN
Thesis Title: 應用影片分析生成足球比賽精華片段
Applying Video Analysis to Generate The Highlight of The Soccer Match
Advisor: 楊傳凱
Chuan-Kai Yang
Committee: 羅乃維
林伯慎
Degree: 碩士
Master
Department: 管理學院 - 資訊管理系
Department of Information Management
Thesis Publication Year: 2021
Graduation Academic Year: 109
Language: 中文
Pages: 58
Keywords (in Chinese): 鏡頭分類標誌偵測直線偵測角落偵測精華產生
Keywords (in other languages): Shot Classification, Logo Detection, Corner Detection, Straight Line Detection, Highlight Generation
Reference times: Clicks: 235Downloads: 0
Share:
School Collection Retrieve National Library Collection Retrieve Error Report

一場足球比賽耗時 90 分鐘,如果觀看賽後精華就可以跳過大部分的比賽內容,可以更快速的了解整場比賽的賽事走向。如果透過自動剪輯精華片段的技術,或許就可以更方便且更省人力的提供給使用者。本論文的系統先輸入比賽時的影片,透過鏡頭分類、Logo 偵測、角落偵測和直線偵測的方法,分析出影片裡潛藏的資料。當中結合了深度學習的技術和型態學(Morphology)的方法,來找到這些有助於分析精華片段的資料。

本論文在鏡頭分類在訓練集和其母群體的場次中的準確度都高於97%,測試集的場次準確度高於 85%。Logo 偵測準確度大於96%,角落偵測準確度也大於98%。在直線偵測上,能順利判斷出場地直線的位置。最後結合上述資料使用精華規則來自動剪輯出精華片段。其結果與官方的精華比較後,證明本論文能有效的自動剪輯出一些精華的片段。


A soccer match takes 90 minutes. If we can obtain the highlight of a game, we
can skip most of the game, and you can more quickly understand the entire game.
If there is technology that can automatically extract the highlights, it may be more
convenient and labor-saving for users. The system of this paper first loads the video
of a match, after shot classification, logo detection, corner detection and straight
line detection methods to analyze the image information hidden in the film, it then
combines deep learning techniques and morphology methods to find the components
that help analyze the highlights of the game.
The accuracy of shot classification is higher than 97% in the training set and
its original dataset, and the accuracy of the test set is higher than 85%. The logo
detection accuracy is greater than 96%, and the corner detection accuracy is also
greater than 98%. For the straight line detection, the straight line position of the
field can be judged smoothly. Finally, we combine the aforementioned techniques
and use the highlight rules to automatically extract the highlights clips. After comparing with the official highlights, this paper proves that it can effectively generate
some highlight clips.

中文摘要.................................................................. II 英文摘要.................................................................. III 誌謝 ...................................................................... IV 目 錄 ..................................................................... V 圖目錄 .................................................................... VII 表目錄 .................................................................... IX 第一章 緒論 .............................................................. 1 1.1 研究背景 .................................................................... 1 1.2 研究動機與目的 ............................................................ 1 1.3 論文架構 .................................................................... 2 第二章 文獻探討 ......................................................... 3 2.1 場地與直線偵測 ............................................................ 4 2.2 事件或動作判定 ............................................................ 6 2.3 鏡頭分類 .................................................................... 6 2.4 Logo 偵測 ................................................................... 8 第三章 演算法設計與系統實作............................................ 9 3.1 系統流程 .................................................................... 9 3.2 Shot Classification .......................................................... 10 3.2.1 Shot Classification model . . . . . . . . . . . . . . . . . . . . . 13 3.2.2 Logo detection . . . . . . . . . . . . . . . . . . . . . . . . . . 17 3.3 場地位置偵測 ............................................................... 19 3.3.1 Straight line detection . . . . . . . . . . . . . . . . . . . . . . 20 3.3.2 Corner detection . . . . . . . . . . . . . . . . . . . . . . . . . 21 3.4 精華規則 .................................................................... 22 第四章 結果展示與評估................................................... 23 4.1 系統環境 .................................................................... 23 4.2 資料集....................................................................... 24 4.3 鏡頭分類模型實驗結果..................................................... 26 4.3.1 三方法比較結果 . . . . . . . . . . . . . . . . . . . . . . . . . . 27 4.3.2 不同pooling比較結果 . . . . . . . . . . . . . . . . . . . . . . . 30 4.3.3 不同optimizer比較結果 . . . . . . . . . . . . . . . . . . . . . . 33 4.3.4 本系統和其他論文比較結果 . . . . . . . . . . . . . . . . . . . 35 4.4 Logo偵測實驗結果.......................................................... 37 4.5 場地位置實驗結果.......................................................... 38 4.5.1 直線偵測比較結果 . . . . . . . . . . . . . . . . . . . . . . . . 38 4.5.2 角落偵測比較結果 . . . . . . . . . . . . . . . . . . . . . . . . 40 4.6 精華剪輯結果 ............................................................... 40 第五章 結論與未來展望................................................... 43 參考文獻.................................................................. 44

[1] Thanaphon Rianthong, Srisupang Thewsuwan, Theekapun Charoenpong, and
Kanjana Pattanaworapan. A method for detecting lines on soccer field by color
of grass variation. In 2020 12th International Conference on Knowledge and
Smart Technology (KST), pages 131–134. IEEE.
[2] Nan Nan, Guizhong Liu, Xueming Qian, and Chen Wang. An svm-based soccer
video shot classification scheme using projection histograms. In Pacific-Rim
Conference on Multimedia, pages 883–886. Springer, 2008.
[3] Mehran RastegarSani and Amir Farid Aminian Modarres. Playfield extraction
in soccer video based on lab color space classification. In 2019 27th Iranian
Conference on Electrical Engineering (ICEE), pages 1999–2003. IEEE, 2019.
[4] Li Sun and Guizhong Liu. Field lines and players detection and recognition in
soccer video. In 2009 IEEE International Conference on Acoustics, Speech and
Signal Processing, pages 1237–1240. IEEE, 2009.
[5] Rahul Anand Sharma, Vineet Gandhi, Visesh Chari, and CV Jawahar. Automatic analysis of broadcast football videos using contextual priors. Signal,
Image and Video Processing, 11(1):171–178, 2017.
[6] Saikat Sarkar, Sazid Ali, and Amlan Chakrabarti. Shot classification and replay
detection in broadcast soccer video. In Advanced Computing and Systems for
Security, pages 57–66. Springer, 2020.
[7] Ali Bagheri-Khaligh, Ramin Raziperchikolaei, and M Ebrahimi Moghaddam. A
new method for shot classification in soccer sports video based on svm classifier.
In 2012 IEEE Southwest Symposium on Image Analysis and Interpretation,
pages 109–112. IEEE, 2012.
[8] Laws of the game - fifa. https://img.fifa.com/image/upload/
datdz0pms85gbnqy4j3k.pdf . Accessed on 07.07.2021.
[9] Market size of the global sports market from 2011 to 2018. https:
//www.statista.com/statistics/1087391/global-sports-market-size/ .
Accessed on 07.07.2021.
[10] Kausik Bandyopadhyay. Legacies of great men in world soccer: heroes, Icons,
Legends. Routledge, 2017.
[11] Sports technology market size, share trends analysis report by technology
(devices, smart stadium), by sports (basketball, soccer, tennis, golf), by region (apac, europe), and segment forecasts, 2021 - 2028. https://www.
grandviewresearch.com/industry-analysis/sports-technology-market .
Accessed on 07.07.2021.
[12] Silvio Giancola, Mohieddine Amine, Tarek Dghaily, and Bernard Ghanem. Soccernet: A scalable dataset for action spotting in soccer videos. In Proceedings of
the IEEE Conference on Computer Vision and Pattern Recognition Workshops,
pages 1711–1721, 2018.
[13] Wook-Sung Yoo, Zach Jones, Henok Atsbaha, and David Wingfield. Painless
tennis ball tracking system. In 2018 IEEE 42nd Annual Computer Software
and Applications Conference (COMPSAC), volume 1, pages 783–784. IEEE,
2018.
[14] Binren Tian, Debing Zhang, and Chun Zhang. High-speed tiny tennis ball
detection based on deep convolutional neural networks. In 2020 IEEE 14th
International Conference on Anti-counterfeiting, Security, and Identification
(ASID), pages 30–33. IEEE.
[15] Rong Ji. Research on basketball shooting action based on image feature extraction and machine learning. IEEE Access, 8:138743–138751, 2020.
[16] Nur Azmina Rahmad, Nur Anis Jasmin Sufri, Nurul Hamizah Muzamil, and
Muhammad Amir As’ari. Badminton player detection using faster region
convolutional neural network. Indonesian Journal of Electrical Engineering
and Computer Science, 14(3):1330–1335, 2019.
[17] Anant Baijal, Jaeyoun Cho, Woojung Lee, and Byeong-Seob Ko. Sports highlights generation bas ed on acoustic events detection: A rugby case study. In
2015 IEEE International Conference on Consumer Electronics (ICCE), pages
20–23. IEEE, 2015.
[18] Jinjun Wang, Changsheng Xu, Engsiong Chng, and Qi Tian. Sports highlight
detection from keyword sequences using hmm. In 2004 IEEE International
Conference on Multimedia and Expo (ICME)(IEEE Cat. No. 04TH8763), volume 1, pages 599–602. IEEE, 2004.
[19] Yueying Zhang, Xiaochun Cao, Dao Wu, and Tao Li. Weakly-supervised tv
logo detection. In 2017 32nd Youth Academic Annual Conference of Chinese
Association of Automation (YAC), pages 1031–1036. IEEE, 2017.
[20] Ngoc Nguyen and Atsuo Yoshitaka. Shot type and replay detection for soccer
video parsing. In 2012 IEEE International Symposium on Multimedia, pages
344–347. IEEE, 2012.
[21] François Chollet. Xception: Deep learning with depthwise separable convolutions. In Proceedings of the IEEE conference on computer vision and pattern
recognition, pages 1251–1258, 2017.
[22] Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer
vision and pattern recognition, pages 770–778, 2016.
[23] Blurred region detection using singular value decomposition (svd). https:
//github.com/fled/blur_detection . Accessed on 07.07.2021.
[24] Bolan Su, Shijian Lu, and Chew Lim Tan. Blurred image region detection
and classification. In Proceedings of the 19th ACM international conference on
Multimedia, pages 1397–1400, 2011.
[25] A pytorch implementation of a yolo v3 object detector. https://github.com/
ayooshkathuria/pytorch-yolo-v3 . Accessed on 07.07.2021.
[26] Joseph Redmon and Ali Farhadi. Yolov3: An incremental improvement. arXiv
preprint arXiv:1804.02767, 2018.
[27] Ngoc Nguyen and Atsuo Yoshitaka. Soccer video summarization based on cinematography and motion analysis. In 2014 IEEE 16th International Workshop
on Multimedia Signal Processing (MMSP), pages 1–6. IEEE, 2014.
[28] Rockson Agyeman, Rafiq Muhammad, and Gyu Sang Choi. Soccer video summarization using deep learning. In 2019 IEEE Conference on Multimedia Information Processing and Retrieval (MIPR), pages 270–273. IEEE, 2019.
[29] Mexico 2 -1 brazil highlight. https://www.youtube.com/watch?v=
7ze8AgkB64U . Accessed on 07.07.2021.
[30] Uae 1-2 uruguay highlight. https://www.youtube.com/watch?v=HGRKSRy7cxA
. Accessed on 07.07.2021.
[31] Great britain 1-1 senegal highlight. https://www.youtube.com/watch?v=
0PPeasuJsiI . Accessed on 07.07.2021.
[32] Korea 2-0 japan highlight. https://www.youtube.com/watch?v=F8-4XKtoYB8
. Accessed on 07.07.2021.
[33] Korea defeat great britain highlight. https://www.youtube.com/watch?v=
eQyfuNFyVLA . Accessed on 07.07.2021.

無法下載圖示 Full text public date 2024/08/10 (Intranet public)
Full text public date 2029/08/10 (Internet public)
Full text public date 2029/08/10 (National library)
QR CODE