Automatic Recognition System of Tunisian License Plates for Car Park Access Control

This paper aims at highlighting the automatic approaches of recognition system of license plates to assure an appropriate access control of car park. After extracting the region of interest by the use of CNN-Mobile-Net-V2 method, a pretreatment step is based on Gaussian Blur filter for enhancing the quality of plate image. The next step is reserved to the characters segmentation of the extracted plate and then decomposing the binary image into sub-images each admitting a digit. The third phase addresses character recognition while using a set of training samples containing many models. For the characters prediction of the tested plate, the supervised learning approach support vector machine gives the greater precision when compared to other computer aided methods. Transforming the decomposed sub-images into alphanumeric information containing the plate number present the last phase. The proposed approach shows an important part in enhancing the parking process of control access.