Character recognition of Malaysian vehicle license plate with deep convolutional neural networks

Research output: Chapter in Book/Report/Conference proceedingConference contribution

4 Citations (Scopus)

Abstract

This paper presents a vehicle license plate recognition method using deep convolutional neural networks. The focus of this paper is placed on the recognition of segmented characters of vehicle license. The deep convolutional neural network is able to distinguish numbers (0 to 9), alphabets (A to Z) and background image from one another. We show that the neural networks trained on computer fonts and natural images can be used to recognize the characters and non-characters on the vehicle license plates. In our experiments, we compared several models of the deep learning model and measure the performance of each model. We find that deeper models of neural networks yield better recognition results. What also find that the deep convolutional neural network is much more robust at the task of character recognition compared to the deep multilayer perceptron. With approximately equal amount of weights and biases parameters, the deep convolutional neural network outperforms all other models on the same task. Our best model using deep convolutional network, can achieve 95.89% correct classification of real license plate characters when even though the network is only trained on computer fonts (from Chars74K dataset) and natural images (from CIFAR10 dataset). No data augmentation is performed during the training.

Original languageEnglish
Title of host publicationIRIS 2016 - 2016 IEEE 4th International Symposium on Robotics and Intelligent Sensors
Subtitle of host publicationEmpowering Robots with Smart Sensors
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages1-5
Number of pages5
ISBN (Electronic)9781509060849
DOIs
Publication statusPublished - 11 Oct 2017
Event4th IEEE International Symposium on Robotics and Intelligent Sensors, IRIS 2016 - Tokyo, Japan
Duration: 17 Dec 201620 Dec 2016

Publication series

NameIRIS 2016 - 2016 IEEE 4th International Symposium on Robotics and Intelligent Sensors: Empowering Robots with Smart Sensors

Other

Other4th IEEE International Symposium on Robotics and Intelligent Sensors, IRIS 2016
CountryJapan
CityTokyo
Period17/12/1620/12/16

Fingerprint

character recognition
Character recognition
Character Recognition
license
neural network
vehicles
Neural Networks
Neural networks
Model
Data Augmentation
Approximately equal
Multilayer neural networks
alphabets
Perceptron
self organizing systems
Multilayer
learning
education
experiment
trend

All Science Journal Classification (ASJC) codes

  • Artificial Intelligence
  • Computer Science Applications
  • Control and Optimization
  • Instrumentation
  • Social Sciences (miscellaneous)

Cite this

How, D. N. T., & Sahari, K. S. M. (2017). Character recognition of Malaysian vehicle license plate with deep convolutional neural networks. In IRIS 2016 - 2016 IEEE 4th International Symposium on Robotics and Intelligent Sensors: Empowering Robots with Smart Sensors (pp. 1-5). (IRIS 2016 - 2016 IEEE 4th International Symposium on Robotics and Intelligent Sensors: Empowering Robots with Smart Sensors). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/IRIS.2016.8066057
How, Dickson Neoh Tze ; Sahari, Khairul Salleh Mohamed. / Character recognition of Malaysian vehicle license plate with deep convolutional neural networks. IRIS 2016 - 2016 IEEE 4th International Symposium on Robotics and Intelligent Sensors: Empowering Robots with Smart Sensors. Institute of Electrical and Electronics Engineers Inc., 2017. pp. 1-5 (IRIS 2016 - 2016 IEEE 4th International Symposium on Robotics and Intelligent Sensors: Empowering Robots with Smart Sensors).
@inproceedings{2b4cca9763174e308888e7b08171c36e,
title = "Character recognition of Malaysian vehicle license plate with deep convolutional neural networks",
abstract = "This paper presents a vehicle license plate recognition method using deep convolutional neural networks. The focus of this paper is placed on the recognition of segmented characters of vehicle license. The deep convolutional neural network is able to distinguish numbers (0 to 9), alphabets (A to Z) and background image from one another. We show that the neural networks trained on computer fonts and natural images can be used to recognize the characters and non-characters on the vehicle license plates. In our experiments, we compared several models of the deep learning model and measure the performance of each model. We find that deeper models of neural networks yield better recognition results. What also find that the deep convolutional neural network is much more robust at the task of character recognition compared to the deep multilayer perceptron. With approximately equal amount of weights and biases parameters, the deep convolutional neural network outperforms all other models on the same task. Our best model using deep convolutional network, can achieve 95.89{\%} correct classification of real license plate characters when even though the network is only trained on computer fonts (from Chars74K dataset) and natural images (from CIFAR10 dataset). No data augmentation is performed during the training.",
author = "How, {Dickson Neoh Tze} and Sahari, {Khairul Salleh Mohamed}",
year = "2017",
month = "10",
day = "11",
doi = "10.1109/IRIS.2016.8066057",
language = "English",
series = "IRIS 2016 - 2016 IEEE 4th International Symposium on Robotics and Intelligent Sensors: Empowering Robots with Smart Sensors",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
pages = "1--5",
booktitle = "IRIS 2016 - 2016 IEEE 4th International Symposium on Robotics and Intelligent Sensors",
address = "United States",

}

How, DNT & Sahari, KSM 2017, Character recognition of Malaysian vehicle license plate with deep convolutional neural networks. in IRIS 2016 - 2016 IEEE 4th International Symposium on Robotics and Intelligent Sensors: Empowering Robots with Smart Sensors. IRIS 2016 - 2016 IEEE 4th International Symposium on Robotics and Intelligent Sensors: Empowering Robots with Smart Sensors, Institute of Electrical and Electronics Engineers Inc., pp. 1-5, 4th IEEE International Symposium on Robotics and Intelligent Sensors, IRIS 2016, Tokyo, Japan, 17/12/16. https://doi.org/10.1109/IRIS.2016.8066057

Character recognition of Malaysian vehicle license plate with deep convolutional neural networks. / How, Dickson Neoh Tze; Sahari, Khairul Salleh Mohamed.

IRIS 2016 - 2016 IEEE 4th International Symposium on Robotics and Intelligent Sensors: Empowering Robots with Smart Sensors. Institute of Electrical and Electronics Engineers Inc., 2017. p. 1-5 (IRIS 2016 - 2016 IEEE 4th International Symposium on Robotics and Intelligent Sensors: Empowering Robots with Smart Sensors).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Character recognition of Malaysian vehicle license plate with deep convolutional neural networks

AU - How, Dickson Neoh Tze

AU - Sahari, Khairul Salleh Mohamed

PY - 2017/10/11

Y1 - 2017/10/11

N2 - This paper presents a vehicle license plate recognition method using deep convolutional neural networks. The focus of this paper is placed on the recognition of segmented characters of vehicle license. The deep convolutional neural network is able to distinguish numbers (0 to 9), alphabets (A to Z) and background image from one another. We show that the neural networks trained on computer fonts and natural images can be used to recognize the characters and non-characters on the vehicle license plates. In our experiments, we compared several models of the deep learning model and measure the performance of each model. We find that deeper models of neural networks yield better recognition results. What also find that the deep convolutional neural network is much more robust at the task of character recognition compared to the deep multilayer perceptron. With approximately equal amount of weights and biases parameters, the deep convolutional neural network outperforms all other models on the same task. Our best model using deep convolutional network, can achieve 95.89% correct classification of real license plate characters when even though the network is only trained on computer fonts (from Chars74K dataset) and natural images (from CIFAR10 dataset). No data augmentation is performed during the training.

AB - This paper presents a vehicle license plate recognition method using deep convolutional neural networks. The focus of this paper is placed on the recognition of segmented characters of vehicle license. The deep convolutional neural network is able to distinguish numbers (0 to 9), alphabets (A to Z) and background image from one another. We show that the neural networks trained on computer fonts and natural images can be used to recognize the characters and non-characters on the vehicle license plates. In our experiments, we compared several models of the deep learning model and measure the performance of each model. We find that deeper models of neural networks yield better recognition results. What also find that the deep convolutional neural network is much more robust at the task of character recognition compared to the deep multilayer perceptron. With approximately equal amount of weights and biases parameters, the deep convolutional neural network outperforms all other models on the same task. Our best model using deep convolutional network, can achieve 95.89% correct classification of real license plate characters when even though the network is only trained on computer fonts (from Chars74K dataset) and natural images (from CIFAR10 dataset). No data augmentation is performed during the training.

UR - http://www.scopus.com/inward/record.url?scp=85050190029&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85050190029&partnerID=8YFLogxK

U2 - 10.1109/IRIS.2016.8066057

DO - 10.1109/IRIS.2016.8066057

M3 - Conference contribution

AN - SCOPUS:85050190029

T3 - IRIS 2016 - 2016 IEEE 4th International Symposium on Robotics and Intelligent Sensors: Empowering Robots with Smart Sensors

SP - 1

EP - 5

BT - IRIS 2016 - 2016 IEEE 4th International Symposium on Robotics and Intelligent Sensors

PB - Institute of Electrical and Electronics Engineers Inc.

ER -

How DNT, Sahari KSM. Character recognition of Malaysian vehicle license plate with deep convolutional neural networks. In IRIS 2016 - 2016 IEEE 4th International Symposium on Robotics and Intelligent Sensors: Empowering Robots with Smart Sensors. Institute of Electrical and Electronics Engineers Inc. 2017. p. 1-5. (IRIS 2016 - 2016 IEEE 4th International Symposium on Robotics and Intelligent Sensors: Empowering Robots with Smart Sensors). https://doi.org/10.1109/IRIS.2016.8066057