Localization of parking spots in surveillance video

Razmatramo detekciju događaja uparkiravanja automobila u videu snimljenom nadzornom kamerom. Predlažemo metodu koja se temelji na praćenju točkastih značajki kroz slijed slika. Pokretne objekte detektiramo kao prostorno bliske nakupine piksela s približno jednakim pomakom s obzirom na prethodnu sliku. Objekte predstavljamo opisanim pravokutnikom koji je poravnat s glavnom osi objekta. Ispitni skup sastojao se od videozapisa statične nadzorne kamere iz ptičje perspektive. U implementaciji je korišten programski jezik C++ i vanjska biblioteka OpenCV koja olakšava rad sa slikama. Metodu smo evaluirali nad uparkiravanjem dvaju automobila. Postignuti rezultati su prikazani i diskutirani.We consider detection of car parking in video recorded by surveillance camera. The method which we propose is based on point features tracking through sequence of photos. We detect moving objects like close clusters of pixels with about same shift according to previous photo. Objects were presented with circumscribed rectangle which is aligned with main axis of object. Test set contains videos recorded by static surveillance camera with bird’s-eye view. Solution was implemented in C++ with OpenCV library for easy work with photos. We evaluated method on two cars parking. Achieved results were displayed and discussed