Bayesian belief based tactic analysis of attack events in broadcast soccer video

Nowadays the concept of soccer technical and tactical analysis is in the center of attention for coaches and analysts as this information includes teams and player policies. Thus, designing analytical systems working without the need for human intervention is required. The objective of this paper is to perform semantic mining with the aim of inferring high level knowledge and assisting the detection of strengths and weaknesses of the teams. This work is fulfilled based on extracting attack and defense situations in broadcast soccer videos using Bayesian belief network. In the proposed method, after automatic classification of the shots and detecting replay scenes, the match sequences are extracted. Then the attack and goal situations are identified by extracting a set of audiovisual features for each sequence and their modeling by Bayesian belief network. By applying heuristic rules during a post-processing level, the attacking team and the type of attack (center, side) will be identified and presented as tactical information. The simulation performed on selected videos from 2010 UEFA championship games demonstrate promising results and precise detections.