跳到主要內容

臺灣博碩士論文加值系統

(44.222.104.206) 您好!臺灣時間:2024/05/28 11:44
字體大小: 字級放大   字級縮小   預設字形  
回查詢結果 :::

詳目顯示

我願授權國圖
: 
twitterline
研究生:林鼎智
研究生(外文):Ting-Chih Lin
論文名稱:建構於視訊運動特徵分析之棒球比賽事件偵測及其語意式分類
論文名稱(外文):Motion-based event detection and semantic classification in baseball sports videos
指導教授:賴文能賴文能引用關係
指導教授(外文):Wen-Nung Lie
學位類別:碩士
校院名稱:國立中正大學
系所名稱:電機工程研究所
學門:工程學門
學類:電資工程學類
論文種類:學術論文
論文出版年:2002
畢業學年度:90
語文別:中文
中文關鍵詞:語意式分類事件偵測運動特徵為主棒球比賽
外文關鍵詞:semantic classificationevent detectionmotion-basedbaseball games
相關次數:
  • 被引用被引用:4
  • 點閱點閱:278
  • 評分評分:
  • 下載下載:0
  • 收藏至我的研究室書目清單書目收藏:2
在本論文中,我們對棒球比賽的視訊內容進行分析。由於運動類型節目具有豐富的運動資訊,因此我們利用棒球視訊中之運動特徵量,進行棒球比賽中事件之偵測及其語意式分類。
我們以video raw data為輸入,以常用的 block matching 方法來估測運動向量。由於 block matching 方法並不保障所估測運動向量的正確性,我們開發了一種評估運動向量可信度的法則。去除不可信任運動向量後我們利用剩餘者進行攝影機運動參數的估測,並經攝影機運動補償後求出實際因於物體的運動向量及每個畫面運動向量的平均大小,作為所採用的運動量 (motion activity)。在本論文中,我們定義從某一個投手投球的鏡頭開始至下一個投球鏡頭之前,稱為一個事件。我們從每一個事件中找出其關鍵鏡頭 (key-shot),並以其各時間段落 (temporal segment) 的運動特徵量為特徵,利用類神經網路進行 supervised 語意式分類。在本文中事件種類分為三類,分別為 (1) 未擊出球型 (2) 內野擊出球型 (3) 外野擊出球型。以本文所收集的 237個棒球視訊事件為實驗,我們的關鍵鏡頭選取法與類神經網路可以達到 86% 的正確分類。
在實際的應用上,以視訊摘要而言,由於整場棒球比賽是由許多的事件組合而成,而關鍵鏡頭又代表了每一事件最重要的部分,因此將每一個事件之關鍵鏡頭串連起來即可視為此場比賽的摘要。在搜尋方面,球迷若希望觀賞屬於某一特定類型的事件,我們可從棒球視訊中搜尋出屬於此類型之事件。
In this thesis, we analyze the content of baseball videos. Due to the abundant motion information in sports videos, we utilize motion features to detect events in baseball games and semantically classify them into three categories.
We take raw video data as input and general block matching is the criterion for motion estimation. Because the using of block matching for motion estimation doesn’t guarantee that the estimated motion vectors are correct, we proposed a method to evaluate the validation of motion vectors. By eliminating incorrect motion vectors, the remained validated ones are used for camera motion parameters estimation. The camera motion parameters and the average magnitude of object motion vectors in each frame are then defined as motion activities. Considering the domain-specific knowledge of baseball games, an event is defined as the time interval from one pitching shot and before the next one. We find the key-shot of each event and its motion features are feed into the neural network to proceed with semantic classification. Our categories include Non-hitting, Infield, and Outfield. In our experiments, more than 200 events are collected for classification. We can achieve 86% classification rate after the key-shot selection and classification.
Considering the real-world applications, video summary is possible, since a baseball game is composed of many events and the key-shot is the representative shot of each event, cascaded key-shots can be considered as the summary of the whole game. For video retrieval, if fans want to watch events belonging to a particular category, we can pick them out of baseball games easily.
一. 緒論
1.1背景和簡介………………………………………………………………1
1.2文獻回顧…………………………………………………………………2
1.3本文研究方法與應用……………………………………………………5
二. 視訊運動特徵之擷取
2.1運動向量估測……………………………………………………………7
2.2運動向量可信度評估……………………………………………………10
2.3攝影機運動參數估測……………………………………………………13
三. 棒球比賽視訊事件之語意式分類
3.1 事件之偵測……………………………………………………………17
3.1.1 事件定義………………………………………………………17
3.1.2 事件偵測法則…………………………………………………18
3.2 事件之語意式分類……………………………………………………22
3.2.1 分類種類………………………………………………………22
3.2.2 關鍵鏡頭的選取………………………………………………23
3.2.3 特徵選取………………………………………………………24
3.2.4 類神經網路分類法則…………………………………………25
四. 實驗結果與討論
4.1 運動向量可信度評估實驗結果………………………………………27
4.2 事件偵測實驗結果……………………………………………………30
4.3 關鍵鏡頭中攝影機運動參數估測……………………………………32
4.4 事件分類結果…………………………………………………………35
4.4.1 One-stage分類…………………………………………………35
4.4.2 Two-stage分類…………………………………………………37
五. 結論與未來研究方向…………………………………………………39
參考文獻……………………………………………………………………40
[1]J. N. Kim, D.K. Kang, S. C. Byun, I.L. Lee and B. H. Ahn, “ A fast full-search motion-estimation algorithm using representative pixels and adaptive matching”, Transactions on Broadcasting, Volume: 48 Issue: 1 , March 2002
[2]Sylvie Jeannin, Radu Jasinschi, Alfred She, Thumpudi Naveen, Benoit Mory, and Ali Tabatabai, “Motion descriptor for content-based video representation”, Signal Processing: Image Communication 16, 2000, pp59-85.
[3]Yap-Peng Tan, Sanjeev R. Kulkarni and Peter J. Ramadge, “A new method for camera motion parameter estimation”, in Proc.1995 IEEE Int. Conf. Image Processing, Vol.1, Oct.1995, pp.406-409.
[4]Gagan B. Rath and Anamitra Makur, “ Iterative least squares and compressed based estimation for a Four-Parameter linear global motion model and global motion compensation” Circuits and Systems for Video Technology, IEEE Transactions on Volume: 9 7 , Oct. 1999 , pp.1075-1099.
[5]蕭偉權,建構於物體運動軌跡特徵之視訊搜尋研究,碩士論文.
[6]Richard O. Duda, Peter E. Hart, and David G. Stork, Pattern Classification, Wiley-Interscience, 2001
[7]B. Adams, C. Dorai, and S. Venkatesh, “Automated Film Rhythm Extraction for Scene Analysis,” ICME 2001.
[8]C. Colombo, A. D. Bimbo, and P. Pala, “Retrieval of Commercials by Video Semantic,” Proceedings. Of 1998 IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 1998.
[9]Tianli Yu and Yujin Zhang, “Retrieval of video clips using global motion information,” Electronics Letters , Volume: 37 Issue: 14 , 5 July 2001, Page(s): 893 —895
[10]Peker, K.A., Alatan, A.A., and Akansu, A.N., “Low-level motion activity features for semantic characterization of video,”, Proc. Of IEEE International Conference on Multimedia and Expo, Volume: 2 , 2000, Page(s): 801 -804
[11]Yoshitaka, A., Hosoda, Y., Hirakawa, M., and Ichikawa, T., “Content-based retrieval of video data based on spatiotemporal correlation of objects,”, Proceedings. of IEEE International Conference on Multimedia Computing and Systems , 1998, Page(s): 208 —213
[12]V. Tovinkere, R.J. Qian, “Detecting Semantics Events in Soccer Games: Towards A Complete Solution,” ICME-2001
[13]A. Bonzanini, R. Leonardi, P. Migliorati, “Event Recognition in Sports Programs Using Low-Level Motion Indices,” ICME-2001
[14]D. Zhong, S. H. Chang, “Structure Analysis of Sports Video Using Domain Models,” ICME-2001
[15]Y. F. Ma, H. J. Zhang, “Detecting Motion Object by Spatio-Temporal Entropy,” ICME-2001
[16]Kittler, J., Messer, K., Christmas, W.J., Levienaise-Obadia, B., and Koubaroulis, D., “Generation of semantic cues for sports video annotation,” Proceedings. 2001 International Conference on Image Processing, Volume: 3 , 2001, Page(s): 26 —29
[17]H.R. Naphide, T.S. Huang. “A probabilistic framework for semantic video indexing, filtering, and retrieval,” IEEE Trans. On Multimedia, vol. 3, no. 1, March 2001, pp.141-151
[18]T. Sikora, “The Mpeg-7 Visual Standard for Content Description-An Overview,” IEEE Transactions on Circuits and Systems for Video Technology,” vol. 11, no. 6, June 2001.
[19]S. Jeannin, A. Divakaran,“Mpeg-7 Visual Motion Descriptor,” IEEE Transactions on Circuits and Systems for Video Technology,” vol. 11, no. 6, June 2001.
[20]M. Bober, “Mpeg-7 Visual Shape Descriptor,” IEEE Transactions on Circuits and Systems for Video Technology,” vol. 11, no. 6, June 2001.
[21]Chia-Han Lin and Arbee L.P. Chen, “Processing Concept Queries with Object Motions in Video Databases,” pp.641-644, ICIP-2001.
[22]張志宇,使用隱藏式馬可夫模型來做棒球節目視訊片段之分類系統,碩士論文.
[23]B. Li and M. I. Sezan, “Event Detection and Summarization in Sports Video,” Content-Based Access of Image and Video Libraries. IEEE Workshop on, 2001, pp132-138
[24]A.N. Chang and C.L. Huang, “Baseball Game Video Shot Analysis and Semantic Classification for Video Indexing System,” Image & Recognition, vol. 5, no. 2, pp. 39-51, 1999.
QRCODE
 
 
 
 
 
                                                                                                                                                                                                                                                                                                                                                                                                               
第一頁 上一頁 下一頁 最後一頁 top