Tactic Analysis in Football Instructors: Nima Najafzadeh Mahdi Oraei Spring
Out line Introduction Framework Related Works Tactic Analysis Advantages and disadvantages Challenges Future Works References 2
Introduction Using digital Videos with high quality for analyzing Some Services: Highlight Replay in‐game Statistics Pattern Analysis Tracking Semantic Analysis 3
Framework Low-level processing based analysis Mid-level representation based analysis High-level analysis based on multimodality 4
Framework 5 Source Video Low level Features Visual Features: Color Shape Audio FeaturesText Features
Framework 6 Visual Model: Motion Ball and player trajectories Mid-Level Model Audio Model: Audio keyword model Text models: Text keywords model Model selections Domain knowledge & Machine learning
Framework 7 Semantics concept Event & highlight extraction Tactic analysis Tracking In game statistics
Related Works 8 Existing approaches for soccer video analysis were mostly for event-driven indexing of video content, which cannot provide detailed tactic information used in the game.
Tactic Analysis 9
10 GOAL EVENT EXTRACTION Web-Casting Text Analysis
Tactic Analysis 11 GOAL EVENT EXTRACTION Game Time Recognition Video/Text Alignment
Tactic Analysis 12
Tactic Information Extraction and Representation 13 Multi-Object Trajectories Acquisition Ball Detection and Tracking Player Detection and Tracking
Ball Detection and Tracking 14
Player Tracking 15
TACTIC INFORMATION EXTRACTION AND REPRESENTATION 16 Aggregate Trajectory Computation Mosaic Trajectory Computation Temporal and Spatial Interaction Analysis
TACTIC INFORMATION EXTRACTION AND REPRESENTATION 17 Play Region Identification
Tactic Analysis 18
Tactic Analysis 19 TACTIC PATTERN ANALYSIS Route Pattern Recognition Interaction Pattern Recognition TACTIC MODE PRESENTATION Data must be clearly and concisely and easy to understand Usable information like ball tracking, player tracking, etc.
Advantages & Disadvantages Pros: Event-driven plus tactic analysis Effective performance in ball and players tracking Good performance in 2006 world cup Cons: human-labeled: Web-Casting Text Weak machine learning in use 20
Challenges the ball becomes a long blurred strip when it moves fast the ball is sometimes occluded by players, merged with lines, or hidden in the auditorium many other objects are similar to the ball. 21
Future Works Using sensor for players and balls Using online statistics of match for coaching Develop this method for other sports 22
Reference G. Zhu, Q. Huang, C. Xu, Y. Yui, S. Jiang,W. Gao, and H. Yao. Trajectory based event tactics analysis in broadcast sports video. In 15th Int. Conf. on Multimedia, pages 58– 67, 2007 Sports Video Analysis: Semantic Extraction, Editorial Content Creation and Application, Changsheng Xu, 2009 Survey of Sports Video Analysis: Research Issues and Applications, J. R. Wang,
Any Question? 24