The new algorithms we develop in this area cover the stages of perception from relevant sensors and video cameras, as well as cognitive stages to build models for interpreting actions and sports activities, making it possible to perform an automatic analysis of matches and propose improvements in the game, objectively and concisely.
ACE seeks to automatically extract the tactics and strategies of sports games based on television videos. In the case of football, it is possible to locate the visible players on the field, and identify tactics collecting statistical data. We currently work with football matches, generating statistical, tactical and strategic analysis.
Temporal and spatial segmentation
To automatically define which portions of video provide useful information for analysis, detecting outages and scene types. Furthermore, using image processing techniques we can detect image regions corresponding to the players.
Calibrating the camera model
Approximation of camera parameters to reconstruct the scene in a "3D" game from the traced points.
Classification of actions and strategies