A robust video based license plate recognition system

Abstract

License plate recognition system (LPRS) is a difficult problem in the field of machine vision and is of substantial interest because of its applications to areas such as cross border security, law enforcement and various other automation applications. Previous methods used plate specific details such as aspect ratio, color or dimensions of the plate in the complex task of plate localization. In this paper, license plate localization process is carried out by weight based density map (WBDM) method, which is independent of such constraints. The proposed method also relaxes constraints in lighting conditions. The robustness of this method is well suited for application in countries like India where the appearance of plates varies widely. 1750 images were captured in different background scenes under various lighting conditions. This method has been tested with different types of vehicles such as buses, care, trucks etc. The overall identification rate of this algorithm is 98.63% in different conditions.