Detection and Tracking of Players in Football Game Video

U radu je opisan jedan pristup u izradi sustava koji može pronaći i pratiti nogometne igrače na snimci utakmice. Sustav koristi razne algoritme računalnoga vida. Na samome početku potrebno je odrediti područje terena na snimci. Nakon toga se pronalaze početne pozicije igrača na terenu koji se zatim prate. Za zadatak praćenja je isprobano nekoliko algoritama. Na samome kraju dobivaju se trajektorije igrača. Trajektorije se zatim dodatno analiziraju i spremaju u datoteku za buduću upotrebu. Korišteni algoritmi praćenja su uspoređeni na temelju uspješnosti praćenja i vremena izvođenja. Sustav je implementiran u programskom jeziku C++ pomoću biblioteke OpenCV.In this paper an approach in building a system that can find and track football players in a video of a game is described. The system uses different computer vision algorithms. At the beginning it is necessary to determine the position of the pitch in the video. After that the positions of players are found and they are tracked after that. For the task of tracking a few algorithms have been tested. At the end the trajectories of the players are acquired. The trajectories are then further analysed and saved in a file for future use. The used tracking algorithms are compared based on tracking performance and running time. The system was implemented in C++ programming language and with OpenCV library