Patent attributes
An automatics system 100 that uses one to three grids 20cm of overhead cameras 20c to first video an event area 2. Overall bandwidth is greatly reduced by intelligent hubs 26 that extract foreground blocks 10m based upon initial and continuously updated background images 2r. The hubs also analyze current images 10c to constantly locate, classify and track in 3D the limited number of expected foreground objects 10. As objects 10 of interest are tracked, the system automatically directs ptz perspective view cameras 40c to follow the activities. These asynchronous cameras 40c limit their images to defined repeatable pt angles and zoom depths. Pre-captured venue backgrounds 2r at each repeatable ptz setting facilitate perspective foreground extraction. The moving background, such as spectators 13, is removed with various techniques including stereoscopic side cameras 40c-b and 40c-c flanking each perspective camera 40c. The tracking data 101 derived from the overhead view 102 establishes event performance measurement and analysis data 701. The analysis results in statistics and descriptive performance tokens 702 translatable via speech synthesis into audible descriptions of the event activities corresponding to overhead 102 and perspective video 202.
