Selvadurai, Vignesh (2024) “Real-time automatic number plate recognition system usingdeep learning approach”. Project Report. Universiti Teknikal Malaysia Melaka, Melaka, Malaysia. (Submitted)
![]() |
Text (Full Text)
“Real-time automatic number plate recognition system usingdeep learning approach”.pdf Download (3MB) |
Abstract
Automatic Number Plate Recognition (ANPR) technology is important and widely used in parking control, traffic monitoring and security surveillance. The current existing ANPR camera faces issues of higher costing, and the process of deployment and maintenance is complex since it requires more advanced technology and higher computational power. Hence, there is a need to develop a simpler but efficient ANPR system using simple devices where it can be applied easily. This project aims to develop an automatic number plate recognition system using a webcam on Predator Helios 300, analyze model accuracy through YOLOv7 algorithms, and examine the accuracy of the ANPR model under different sun ray conditions. This study focuses on the practical challenges of recognizing characters on number plates in the context of developing a real-time automatic number plate recognition (ANPR) system using a deep learning approach. Localization of number plates is the initial and essential step in automatic number plate recognition, which is then followed by character recognition. In this study, the number plate is first identified using a preprocessing technique to extract its features. It then goes through fully connected layers to classify and predict grid and cell division. A post-processing phase called non-maximum suppression (NMS) is used to reduce the number of overlapping candidate regions of interest (ROIs) and replace them with a single, more accurate detection. Combining YOLOv7, a powerful convolutional neural network, with Tesseract-OCR, a highly advanced optical character recognition engine, is crucial for achieving precise and instantaneous ANPR results. The study highlights the plate localization and character recognition, demonstrating the exceptional performance metrics of the YOLOv7 model: mean average precision of 0.99405, precision of 0.994168, and recall of 0.988764. This output is expected to beneficial to smart parking systems, stolen vehicle identification, and automatic enforcement systems. It provides a reliable and efficient method for accurately recognizing license plates in real-time situations.
Item Type: | Final Year Project (Project Report) |
---|---|
Uncontrolled Keywords: | Real-time, Number Plate , Deep Learning |
Subjects: | T Technology > T Technology (General) T Technology > TA Engineering (General). Civil engineering (General) |
Divisions: | Library > Final Year Project > FTKE |
Depositing User: | Norfaradilla Idayu Ab. Ghafar |
Date Deposited: | 21 Oct 2024 07:20 |
Last Modified: | 20 Nov 2024 07:21 |
URI: | http://digitalcollection.utem.edu.my/id/eprint/33760 |
Actions (login required)
![]() |
View Item |