Two different versions of TRACAB were validated: the 4th generation TRACAB system (Gen4, installed at a height of 36m), and the 5th generation (Gen5, installed at a height of 16m), which represents an advancement of the Gen4 system. The Gen4 system is a stereo camera system, consisting of two multi-camera units in two locations either side of the halfway line, each comprising three HD-SDI cameras with a resolution of 1920x1080 pixels (see Fig 2 ).
Each multi-camera unit provides a stitched panoramic picture, which is then used to create the stereoscopic view for triangulating the players and ball. The Gen5 system represents a distributed camera architecture, combining two stereo pairs on each side of the field as well as two monocular systems behind the goal areas. In total, the tested Gen5 systems comprised 16 IP-HD cameras with a resolution of 1920x1200 pixels. An exemplary still image of one of Gen5’s multiple-camera units is depicted inFig 3 .
It should be noted that optical tracking data is prone to errors introduced by occlusions and ID-swaps, which is why one human operator is required to correct these errors during measurement. For this reason, three types of tracking data can be distinguished: Type I data is considered ‘real-time’ data which is available immediately (<300ms), e.g. for live broadcasting, and therefore potentially contains several tracking errors. There is also Type II data, which is made available with approximately 15s delay and has less errors due to further processing and human interventions. Finally, Type III data describes the final data which is made available within several hours after the recordings. This final step comprises complete data correction and post-processing steps.
During our tests, the Gen4 system ran live (online) tracking during the recording session with a trained operator. The resulting Type II data was made available by ChyronHego immediately after the recording sessions. The Gen5 system recorded the same exercises. However, the video footage was later tracked offline using a real time tracking framework. When tracking the Gen5 footage, target identities were assigned at the beginning of each exercise. In all cases, the Gen5 system retained the correct target identities throughout the trial and no further operator interaction was required.
Accordingly, the data analysed in this study comprises exclusively Type II data, consisting of cartesian XY coordinates sampled with a frequency of 25 frames per second (25 Hz), which we from here on refer to as ‘raw’ tracking data, as no additional filtering methods or post-processing procedures have been applied.
Each multi-camera unit provides a stitched panoramic picture, which is then used to create the stereoscopic view for triangulating the players and ball. The Gen5 system represents a distributed camera architecture, combining two stereo pairs on each side of the field as well as two monocular systems behind the goal areas. In total, the tested Gen5 systems comprised 16 IP-HD cameras with a resolution of 1920x1200 pixels. An exemplary still image of one of Gen5’s multiple-camera units is depicted in
It should be noted that optical tracking data is prone to errors introduced by occlusions and ID-swaps, which is why one human operator is required to correct these errors during measurement. For this reason, three types of tracking data can be distinguished: Type I data is considered ‘real-time’ data which is available immediately (<300ms), e.g. for live broadcasting, and therefore potentially contains several tracking errors. There is also Type II data, which is made available with approximately 15s delay and has less errors due to further processing and human interventions. Finally, Type III data describes the final data which is made available within several hours after the recordings. This final step comprises complete data correction and post-processing steps.
During our tests, the Gen4 system ran live (online) tracking during the recording session with a trained operator. The resulting Type II data was made available by ChyronHego immediately after the recording sessions. The Gen5 system recorded the same exercises. However, the video footage was later tracked offline using a real time tracking framework. When tracking the Gen5 footage, target identities were assigned at the beginning of each exercise. In all cases, the Gen5 system retained the correct target identities throughout the trial and no further operator interaction was required.
Accordingly, the data analysed in this study comprises exclusively Type II data, consisting of cartesian XY coordinates sampled with a frequency of 25 frames per second (25 Hz), which we from here on refer to as ‘raw’ tracking data, as no additional filtering methods or post-processing procedures have been applied.
Free full text: Click here
