Detection of faces in video content and their classification

Namen magistrskega dela je razviti sistem, ki zna učinkovito zaznati obraz v video vsebinah. Izdelali smo sistem, ki zazna obraz in ga klasificira za nadaljnjo obravnavo v drugih sistemih. Razložili smo proces zaznave in prepoznave obraza, napisali kratek pregled tehnologij in algoritmov, ki se uporabljajo danes. Predstavili smo princip zaznave obraza v video vsebinah, ki je neodvisen od vira video vsebine. Uporabljeni metodi za zaznavo obrazov sta analiza glavnih komponent in razvrščanje značilk s Haarovimi razvrščevalniki. Pri izvedbi sistema zaznave obrazov v video vsebinah smo uporabili funkcije iz knjižnice OpenCV. Sistem je bil razvit v programskem jeziku Python, saj smo želeli ustvariti sistem, ki je med drugim tudi prenosljiv med platformami. Sistem se je med testiranjem izkazal za učinkovitega, z določenimi izboljšavami pa bo koristna osnova za nadaljnjo procesiranje in prepoznavo obrazov.The purpose of the master thesis is to develop a system that is able to effectively detect a face in a video content. We developed a system that detects a face and classifies it for further processing in other systems. We explained the process of perception and facial recognition, wrote a brief overview of technologies and algorithms we use today. We presented the principle of Face Detection in video content that is independent of the video content stream. Used methods for face detection are principal component analysis and classification with Haar cascade classifiers. To create the system for face detection, we used the functions from the OpenCV library. The system was developed in the Python programming language, as we wanted to create a system that is portable between platforms. The experimental results showed that the developed system is efficient and with certain improvements it will be a useful basis for further processing and face recognition