AMLPDS: An Automatic Multi-Regional License Plate Detection System based on EasyOCR and CNN Algorithm

Abstract

Automatic License Plate Recognition (ALPR) System detects License Plate (LP) of a vehicle. The computer vision zone considers ALPR system as a resolved issue. However, the majority of current ALPR research is based on LP from specific countries and employs country-specific data. Therefore, the proposed methodology deals with the LP which will work on the regions in & around India. The algorithm applied in the proposed methodology is Convolution Neural Network (CNN). The proposed methodology comprises three major steps: Firstly, License plate detection which uses Single Shot Detector (SSD) which divides the image into grid cells, with each grid cell being in charge of detecting objects in that area. Secondly, Unified character recognition which uses easyOCR (Optical Character Recognition) has the ability to deal with multi scale and small objects. Finally, Multi-regional layout detection extracts the correct order of the license plate. The dataset is collected from which is “Indian License Plate Dataset”. Experiment results outperform the existing mechanisms in terms of time conception accuracy of LP recognition, end to end recognition and average execution time.