CSEIT206396 | Accepted : 18 June 2020 | Published : 25 June 2020 | May-June-2020 [ 6 (3) : 896-902 ]
International Journal of Scientific Research in Computer Science, Engineering and Information Technology
© 2020 IJSRCSEIT | Volume 6 | Issue 3 | ISSN : 2456-3307
DOI : https://doi.org/10.32628/CSEIT206396
896
Handwritten Digit Recognition using Image Preprocessing and CNN
Anshul Dubey, Ashley Lazarus, Dharmendra Mangal
Computer Science Department, MediCaps University, Indore, Madhya Pradesh, India
ABSTRACT
Handwritten digit recognition, is a technique of identifying and enlisting the recognized digit, that uses neural
networks, deep learning and machine learning. The applications and demand of handwritten digit recognition
systems such as zip code recognition, car number plate recognition, robotics, banks, mobile applications and
numerous more, are soaring every day. It can be done through numerous approaches, but convolutional neural
network is considered one of the best methods. The special neural network uses multilayer architecture for
identification and classification. Although the accuracy factor can be increased, based on image preprocessing, in
this paper we discuss how the accuracy of the system can be increased for better handwritten digit recognition,
using convolutional neural networks, image preprocessing; binarization, resizing, rotation. The accuracy rate
obtained is 99.33%.
Keywords : Digit Recognition, Image Processing, CNN.
I. INTRODUCTION
Digit recognition is an application of a computer
system which makes it able to recognize the
handwritten inputs like digits from various kinds of
sources like images, papers, emails, letters and more.
This ability of a computer is very important as it can
reduce manpower in areas like postal address
interpretation, bank check processing, signature
verification, etc. This topic has been for a lot of years
a big topic of research. A lot of classification
techniques using Machine Learning have been
developed and used for this like K-Nearest Neighbors,
SVM Classifier, Random Forest Classifier, etc. but
these methods although have enough accuracy, are not
enough for the real-world applications. One example
of this is, if you send a letter with postal code as
“452311” and the system detects and recognizes it as
“452811” then it will not be able to deliver the letter
due to incorrect postal code [1]. Thus, accuracy in
these
applications is very critical but these techniques do not
provide the required accuracy due to very little
knowledge about the topology of a task. Here comes
the use of Deep Learning. In the past decade, deep
learning has become the hot tool for Image Processing,
object detection, handwritten digit and character
recognition etc. A lot of machine learning tools have
been developed like scikit-learn, SciPy-image and
pybrains, Keras, Theano, TensorFlow by Google,
TFLearn etc. for Deep Learning. These tools make the
applications robust and therefore more accurate, also
the Artificial Neural Networks can almost mimic the
human brain and are a key ingredient in image
processing field. We applied CNN model, using Keras
library in python, for handwritten digit recognition,
with improves image preprocessing. Implementing a
better and accurate approach to perceive and foresee
manually written digits from 0 to 9 from images [2].
Previously many problems have been faced in this
research area one of them which