There are thousands of football games played all over the world every day. Most of them are between amateur teams. recording the games as videos for later analysis enables systematic performance improvements. But due do the lack of dedicated film makers and cameras the task is intractable for most of the teams and clubs. Thus, the idea of autonomous camera for football games emerged. The challenges are the small ball and the large distance of the camera from the field.
The presented approach relies on an algorithm implemented on a smart phone, which can automatically track and film the game. The system estimates the direction of the player motion on the pitch and rotates the phone via the attached motor accordingly. The assumption is that the direction where the most players move to, is the direction of the game progress.
The presented work is a framework incorporating multiple computer vision techniques such as corner detection. Optical Flow and robust filtering implemented on an Android smart phone. The technique estimates the overall player movement on the field and sends commands to the motor via a serial interface.
The player tracking algorithm was validated with test pictures and the system was tested on several football games of Berlin amateur teams. The results show, that the approach developed in this thesis can be used for autonomous football game tracking and filming using a standard smart phone.