License Plate Recognition using CNN and OCR on SMQT Images Obtained from RANSAC Algorithm

Abstract

Now a days the count of vehicles increases day by day. Due to different plate layout and fonts recognition is difficult in some situation of the identification of license plaques plays an important role in transport systems. The License plate region is localized by Random Sample Consensus (RANSAC) algorithm. Then the Successive Mean Quantization Transform (SMQT) is applied to the localized region to focus on the details of the region. The SMQT image is supplied the Convolutionary Neural Network (CNN) to classify the plates, whether or not they are readable. After classification the license plate is recognized using Optical Character Recognition. The main purpose of this paper is to identify with maximum efficiency various plate formats and fonts of different state license plates.