Detecting Multiple Indian Licence plates from Real time Videos using YOLOv4 and CNN
Author(s):
Abhishek Wakchaure, Dr. Rajashree Shettar
Keywords:
Automatic Licence Plate Recognition, complex environments, Real time, Optical Character Recognition, YOLOv4, CNN, UFPR-ALPR.
Abstract
An Automatic Licence Plate Recognition (ALPR) system is able to read the characters on the number plate without human interference. It has been an area of continuous research due to a variety of practical applications but they are still dependent on many constraints. Indian licence plates are even tougher because of the density of cars on the roads. In this paper, we present a deep learning based approach which treats ALPR as two separate problems, an object detection using YOLO v4 and a character recognition problem using a Convolutional Neural Network model (CNN). The YOLO v4 model is custom trained to detect multiple licence plates in complex environments in real time and the CNN model is fine tuned so it is robust enough to recognize characters even in improper lighting conditions or low clarity. The approach yielded impressive results on two separate datasets, an UFPR-ALPR dataset which contains 150 videos and 4500 frames and an Indian Licence plate dataset with 640 images with different colour and lighting conditions.
Article Details
Unique Paper ID: 152251
Publication Volume & Issue: Volume 8, Issue 2
Page(s): 688 - 694
Article Preview & Download
Share This Article
Conference Alert
NCSST-2021
AICTE Sponsored National Conference on Smart Systems and Technologies
Last Date: 25th November 2021
SWEC- Management
LATEST INNOVATION’S AND FUTURE TRENDS IN MANAGEMENT