A System Design for License Plate Recognition by Using Edge Detection and Convolution Neural Network

• Published in: 2018 International Conference on Computer, Communication, Chemical, Material and Electronic Engineering (IC4ME2) pp. 1 - 4
• Main Authors: Dhar, Prashengit, Guha, Sunanda, Biswas, Tonoy, Abedin, Md. Zainal
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.02.2018
• Subjects: Character recognition; Character Segmentation; Convolution Neural Network (CNN); Distance to Borders (DtBs); Edge detection; Feature extraction; Filtering; Image edge detection; Image segmentation; License plate recognition; License Plate Recognition (LPR)
• DOI: 10.1109/IC4ME2.2018.8465630

Intelligent Transport System (ITS) shows significant contributions in smart system applications .Automatic License Plate Recognition(LPR) system is a fascinating component of ITS that can be used in many real life applications such as surveillance, traffic flow monitoring, tracking stolen car, parking lot maintenance. This paper focuses a system design for LPR implementation of Bangladeshi License Plate. To detect the plate from the car image is the first stage of the system. The detection consists of the preprocessing, edge detection, morphological dilation and filtering region properties of the input image. Secondly, shape of the plate is verified by using robust distance to borders (DtBs) method. Thirdly, extrema points are used to correct horizontal tilt. Then, to extract the plates object, character segmentation is executed based on region properties and morphological operation. Finally, the extracted characters are recognized through automatic feature extraction with Convolution Neural Networks. The simulation results illustrate that the accuracy is quiet remarkable.