Detection and Person Tracking in videos based on characteristic points

U ovom radu smo opisali metodu praćenja osobe pomoću karakterističnih točaka te detaljno objasnili njenu implementaciju. Prošli smo mane i prednosti naše metode te predložili poboljšanja koja bi trebala omogućiti robusno praćenje osoba čak i u gužvi. Također smo pokazali kako mnogo toga može utjecati na kvalitetu videa (npr. kiša ili jako sunce) te kakve probleme različiti broj ljudi i različita područja (npr. radi mnogih znakova / zaklona) mogu izazvati. Stoga vrstu metode koju ćemo koristiti potrebno je pažljivo izabrati ovisno o našim potrebama i uvjetima s kojima će se morati nositi. Idealno bi bilo moći implementirati posebnu metodu za svaki slučaj što je na žalost financijski ne izvedivo. Još smo diskutirali o manama raznih metoda te dokazali da je u većini slučajeva hibridna vrsta metoda najefikasnija i najuspješnija radi sadržavanja beneficija detekcije i praćenja te umanjivanja njihovih mana.In this paper we have described a method for person detection and tracking based on extraction of characteristic points with Harris corner detection. We have explained its implementation and its strengths and flaws as well as some improvements to make it more robust during tracking of crowded areas. Importance of video quality and the impact various situations on it has been shown among them crowdedness, obstacles and even weather. Because of that the choice of detection and tracking method should be carefully considered to ensure maximum quality. It would be best if we could make custom detection and tracking method but that’s not really economically doable. Finally, we have discussed flaws of various methods and proven that hybrid methods are in most cases most efficient and most successful since they contain all the benefits of detection and tracking methods while considerably reducing their flaws