臺灣博碩士論文加值系統

因為籃球影片的大量增加，影片分析在現今扮演著一個重要的角色，相機校正也一樣很重要，因為它是影片分析的前處理，籃球影片的相機校正被用在將影格中的球場校正到標準模型的球場，因為籃球場是一個平面，大部分的方法利用這個性質去解出單應性矩陣來達成相機校正，為了得到單應性矩陣，需要從籃球影格中偵測至少四個具有意義的點，然後再去解線性系統，事實上，任兩條非平行線可以相交出一個點，所以這個偵測點的問題可以轉換成線的偵測，在籃球影片中，我們會偵測禁區線，因為他在每個影格中是最常出現且最明顯的線，然而因為現代的球場有很多樣式且顏色都不同，所以這會嚴重的影響偵測禁區線的正確率，於是我們提出了一個可以有效的適用在多樣性籃球場的相機校正系統，在這篇論文中，首先取出了主宰顏色圖與邊緣圖，然後應用了計分板偵測與球員偵測，為了除去邊緣圖的雜訊像素，使用了前面得到的資訊去設計出四個遮罩，在這之後先偵測了底線與邊線，並且將邊緣圖上球場外沒用的像素過濾掉，接著取出禁區線的候選人，然後使用兩個限制與兩個分數決定出最終的禁區線，最終使用內插法來提煉所有的線，並且計算出單應性矩陣來做相機校正，在實驗結果中，我們的方法會與[4]做比較，並實驗在所有的NBA球場做驗證。

Due to a massive growth of basketball videos, the video analysis plays an important role nowadays. Camera calibration is also important since it has been used to preprocess for video analysis. Camera calibration is used to calibrate the court in the frame to the court in the standard basketball court. Because a basketball court is actually a plane, most of the methods use this property to solve a homography to achieve camera calibration. In order to obtain the homography, at least four meaningful points should be detected from the frames of the basketball videos and then used to solve a linear system. In fact, two nonparallel lines result in an intersection point, so the problem of detecting points can turn into line detection. In a basketball video, what we detect are paint lines since they are obvious and often appear in each frame. However, due to various style of modern basketball courts, detecting accurate paint lines are heavily affected. Hence, we propose a camera calibration system which is robust for various style of basketball courts. In this paper, both the dominant color map and the edge map are extracted first, and then scoreboard detection, as well as player detection, are applied. Using above information to design four masks is necessary to remove noisy pixels from the edge map. After that, baseline and sideline are detected and the useless pixels outside the court are filtered out in the edge map. Next, paint line candidates are extracted. Then two constraints and two scores are applied for deciding the final paint lines. Finally, all the lines are refined by means of interpolation and the homography can be calculated for camera calibration. In experiment results, our method is compared with [4] and applied to all NBA courts for evaluation.

誌謝 i
中文摘要 ii
ABSTRACT iii
CONTENTS iv
LIST OF FIGURES vi
Chapter 1 Introduction 1
1.1 Introduction of Camera Calibration in Broadcast Basketball Videos 1
1.2 Thesis Organization 5
Chapter 2 Related Work 6
2.1 Camera Calibration in Basketball Videos Overview 6
2.2 Paint Detection for Camera Calibration in Basketball Videos 7
2.3 Court Reconstruction for Camera Calibration in Basketball Videos 10
Chapter 3 Method 12
3.1 System Overview 12
3.2 Preprocessing and Feature Extraction 13
3.2.1 Scoreboard and Player Detection 13
3.2.2 Dominant Color Map Extraction 16
3.2.3 Mask Generation 20
3.3 Paint Line Extraction 22
3.3.1 Court Line Detection 22
3.3.2 Paint Line Candidate Extraction 26
3.3.3 Paint Line Determination 27
3.3.4 Paint Line Refinement 32
3.3.5 Homography Calculation 34
Chapter 4 Experiment Results 35
4.1 Statistical Results 35
4.2 Visualizing Results 38
Chapter 5 Conclusion and Future Work 43
5.1 Conclusion 43
5.2 Future Work 43
REFERENCE 45

[1]W. L. Lu, J. A. Ting, J. J. Little, and K. P. Murphy. Learning to Track and Identify Players from Broadcast Sports Videos. IEEE Transactions on Pattern Analysis and Machine Intelligence, 35(7), 1704-1716, 2013.
[2]P. C. Wen, W. C. Cheng, Y. S. Wang, H. K. Chu, N. C. Tang, ad H. Y. M. Liao. Court Reconstruction for Camera Calibration in Broadcast Basketball Videos. IEEE Transactions on Visualization and Computer Graphics, 22(5), 1517-1526, 2016.
[3]M. C. Hu, M. H. Chang, J. L. Wu, and L. Chi. Robust Camera Calibration and Player Tracking in Broadcast Basketball Video. IEEE Transactions on Multimedia, 13(2), 266-279, 2011.
[4]H. T. Chen, C. L. Chou, T. S. Fu, S. Y. Lee, and B. S. P. Lin. Recognizing Tactic Patterns in Broadcast Basketball Video Using Player Trajectory. Journal of Visual Communication and Image Representation, 23(6), 932-947, 2012
[5]D. Farin, J. Han, and P. H. de With. Fast Camera Calibration for the Analysis of Sport Sequences. IEEE International Conference in Multimedia and Expo on (pp. 4-pp). IEEE, 2005.
[6]Y. Liu, S. Jiang, Q. Ye, W. Gao, and Q. Huang. Playfield Detection Using Adaptive GMM and Its Application. IEEE International Conference in Acoustics, Speech, and Signal Processing on (Vol. 2, pp. ii-421), 2005.
[7]C. M. Chen, and L. H. Chen. A Novel Method for Slow Motion Replay Detection in Broadcast Basketball Video. Multimedia Tools and Applications, 74(21), 9573-9593, 2015.
[8]P. F. Felzenszwalb, R. B. Girshick, D. McAllester, and D. Ramanan. Object Detection with Discriminatively Trained Part-based Models. IEEE Transactions on Pattern Analysis and Machine Intelligence, 32(9), 1627-1645, 2010.
[9]R. Girshick, J. Donahue, T. Darrell, and J. Malik. Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation. IEEE Computer Vision and Pattern Recognition (pp. 580-587), 2014.
[10]R. Girshick, Fast R-cnn. IEEE International Conference on Computer Vision (pp. 1440-1448), 2015.
[11]S. Ren, K. He, R. Girshick, and J. Sun. Faster R-CNN: Towards Real-time Object Detection with Region Proposal Networks. In Advances in Neural Information Processing Systems (pp. 91-99), 2015.
[12]V. Ramanathan, J. Huang, S. Abu-El-Haija, A. Gorban, K. Murphy, and L. Fei-Fei. Detecting Events and Key Actors in Multi-person Videos. IEEE Computer Vision and Pattern Recognition (pp. 3043-3053), 2016.
[13]A. Krizhevsky, I. Sutskever, and G. E. Hinton. Imagenet Classification with Deep Convolutional Neural Networks. In Advances in Neural Information Processing Systems (pp. 1097-1105), 2012.