LICENSE PLATE RECOGNITION ON MOBILE PHONE WITH OPERATING SYSTEM ANDROID

V tem diplomskem delu smo se ukvarjali z razpoznavo registrskih tablic na mobilnem telefonu z operacijskim sistemom Android. Kot izhodiščno metodo smo uporabili detekcijo s konturami. Z uporabo različnih morfoloških operacij smo izboljšali natančnost razpoznave, vključili pa smo še sistem za optično razpoznavo znakov. Nastala je aplikacija, s katero med vožnjo iz video toka zajemamo slike ter na njih lociramo registrske tablice. Iz njih nato izločimo posamezne znake in jih pretvorimo v tekst. V zadnjem koraku enake registrske tablice združimo s pomočjo primerjave histogramov in na ta način še izboljšamo končne rezultate. Uspešnost postopka smo preverjali s pomočjo predhodno zajetega video materiala. Ocenjevali smo uspešnost posameznih faz ter uspešnost celotnega postopka. Rezultati kažejo, da smo glede na metodo, iz katere smo izhajali, uspešnost razpoznave registrskih tablic bistveno izboljšali. Odstotek pravilno razpoznanih registrskih tablic namreč presega 80 %.In the diploma thesis, we have been dealing with recognition of licence plates on a mobile phone with operating system Android. We have used detection with contours as a starting point method. We used different morphological operations and have improved precision of recognition, including a system for optical sign recognition. We have made an application which can be used for gathering pictures during driving and locating licence plates on the pictures. We can exclude individual signs and change them into the text. The last step is when we join the same licence plates with the help of histogram comparison and improve final results this way. Success of the procedure has been checked with the help of previously gathered video material. We have graded successfulness of individual phases and successfulness of the whole procedure. The results have shown that we have, according to the method we have used, fundamentally improved successfulness of licence plate recognition. The percentage of correctly recognised licence plates exceeds 80%