Institute of Cartography and Geoinformatics | Leibniz Universität Hannover Trajectory Analysis Analyzing Trajectories in a Soccer Context M.Sc. Udo Feuerhake [email protected] Outline ► Motivation ► The Tool ► Basic Analysis Tasks ► Advanced Analysis Tasks ► Conclusion & Outlook Move – Final Conference2013 Udo Feuerhake| 2 Motivation and Application Scenarios ► Application scenarios: Monitoring of performance in the training/competition • Enables an adjusted training and better performance of the individual player and the whole team Analysis of the opponent • Better/easier preparation of the competition ► Existing services/applications (especially in soccer domain) provide just the basic analysis tasks Move – Final Conference2013 Udo Feuerhake| 3 The Tool ► ► Implemented in Java, at the moment extension to a framework Purposes: Testing Visualization of the results Comparison of results Move – Final Conference2013 Udo Feuerhake| 4 Basic Analysis Tasks ► Determination (measurement) of basic statistical values of a player or a whole team Total covered distance (Distribution of) velocities / accelerations Min./mean/ max. values Heat/intensity maps Move – Final Conference2013 Udo Feuerhake| 5 Basic Analysis Tasks ► Use of event-based approach ► Different kinds of events ‘Game events’ may be given attached to the dataset (annotations) • Match is started / interrupted / finished • Control of movement observer ‘Movement events’ are generated by the observer from the data Game Start Event Game Interruption Event Game Resume Event Movement observer Active Inactive t Movement Events Move – Final Conference2013 Udo Feuerhake| 6 Basic Analysis Tasks ► Determining the ball possession (per team) Nearest player (body part) is possessor (up to an upper boundary) • E.g. 0.3m (depends on the data accuracy) Ball possession change event, if possessor changes Possession time = time between two possession events t Team A in possession Ball Possession Change Event Team B in possession Ball is free Move – Final Conference2013 Udo Feuerhake| 7 Basic Analysis Tasks ► Detection of passes Framed by a ‘ball kick event’ and a ‘ball stop event’ a_ball Ball possessing players are sender and receiver Completed pass Bad pass Bad passes have no or wrong receiver Whole team Move – Final Conference2013 One player Udo Feuerhake| 8 Basic Analysis Tasks ► Further tasks are solved similarly: Goals Sprints Ball contacts Move – Final Conference2013 Udo Feuerhake| 9 Advanced Analysis Tasks ► ‚Pass graph‘ Generation of a graph structure • Nodes players • Edges passes • Edge weight frequency of passes between pair of players Visual analysis is possible via the stroke width of the edges Analysis via graph based algorithms, e.g. frequent pass sequences Move – Final Conference2013 Udo Feuerhake| 10 Advanced Analysis Tasks ► Extraction of group movement patterns Approach is based on constellations (vector of relative player positions) Sequence of constellations is recorded during the observation time Clustering of constellations to determine their similarities Use of sequence mining algorithm to extract patterns from the sequence of clusters (clustered constellations) Example pattern (occurred twice during the observation time): time step: subsequence subsequence Move – Final Conference2013 Udo Feuerhake| 11 Advanced Analysis Tasks Move – Final Conference2013 Udo Feuerhake| 12 Conclusion ► ► Tool for observing and analyzing trajectories in a soccer context Basic analysis tasks basic statistical values, hotspots Ball possession, contacts Passes, goals, sprints ► Advanced analysis tasks Passes graph Group movement pattern recognition Move – Final Conference2013 Udo Feuerhake| 13 Outlook ► Further planned features: Detection of goal kicks (distinction of kicks and passes) Detection of corner kicks, free kicks, penalties, throw-ins Detection of physical interactions of players (e.g. fouls) ► Implementation of graph analysis methods for the pass graph ► Extension of the pattern recognition approach Use of more detailed and specific knowledge Use of a database for comparison issues ► !STRONG NEED FOR DATASETS! Move – Final Conference2013 Udo Feuerhake| 14 Thank you for your attention! Move – Final Conference2013 Udo Feuerhake| 15
© Copyright 2026 Paperzz