Tracking of Players in Football Game Video

Danas postoje razne metode praćenja objekata koje na razne načine mogu brzo i efikasno pomoći treneru i stručnom stožeru analizirati igrače unutar utakmice na temelju čega mogu doći do raznih strategija kako igrati protiv drugih timova, primjerice koji igrači bi bili prikladni za igru, koji ne, te sam način igre. Sve se to dobiva na temelju puta kojeg prolazi igrač, prostora u kojem se kreće te svega ostalog što se dade zaključiti iz praćenja igrača. U ovom radu opisuje se klasično praćenje objekata koristeći algoritam Mean-shift. Kako bismo primjenili Mean-shift moramo koristiti neku metodu obrade slike kako bismo dobili dobru distribuciju s kojom radi Mean-shift. Modelira se histogram na temelju kojega se opisuje vjerojatnost da li se igrač nalazi na tom mjestu u terenu. Algoritam je testiran na sekvencama nogometnih utakmica, a radi boljeg praćenja i pokazivanja što sve poboljšava praćenje, implementacija je primjenjiva na specifičan skup video-sekvenci.Today there are many different object tracking methods which can be helpful to coach and his expert team for efficient and fast analysis of players in the game . They can decide which strategy to use versus opponent teams, for instance, which players would be suitable for playing, which not, or how to play the whole game generally. We get all of that based on player’s path, space where he moves and all other things which can be deducted from players tracking. In this thesis is described classic object tracking using Mean-shift algorithm. In order to apply Mean-shift, we must use some specific image processing method, so we could get good density distribution, which Mean-shift use as input. We are modelling histogram which is used for describing probability, is player placed on a specific dot in football field. Mean-shift is tested on video-sequences of football games. For better tracking and indicating what does improve tracking, given implementation is effective on specific set of video-sequences