Rearrangement Of Coordinate Selection For Triangle Features Improvement In Digit Recognition

Triangle geometry feature demonstrated as useful properties in classifying the image. This feature has been implemented in numerous recognition field such as biometric area, security area, medical area, geological area, inspection area and digit recognition area. This study is focusing on improvin...

Full description

Saved in:
Bibliographic Details
Main Authors: Miswan, Siti Aishah, Azmi, Mohd Sanusi, Arbain, Nur Atikah, Tahir, Azrina, Radzid, Amirul Ramzani
Format: Article
Language:English
Published: Penerbit Universiti, UTeM 2018
Subjects:
Online Access:http://eprints.utem.edu.my/id/eprint/22861/2/Siti%20Aishah%20-%204437-11613-1-SM.pdf
http://eprints.utem.edu.my/id/eprint/22861/
http://journal.utem.edu.my/index.php/jtec/article/viewFile/4437/3292
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Triangle geometry feature demonstrated as useful properties in classifying the image. This feature has been implemented in numerous recognition field such as biometric area, security area, medical area, geological area, inspection area and digit recognition area. This study is focusing on improving triangle features in digit recognition. Commonly, triangle features are explored by determining three points of triangle shape which represent as A, B and C to extract useful features in digit recognition. There is possibilities triangle shape cannot be formed when chosen coordinate are in line. Thus, a prior study has proposed an improvement on triangle selection point technique by determining the position of coordinate A, B and C use gradient value to identify the triangle shape can be modelled or vice versa. The suggested improvement is based on the dominant distribution which only covers certain areas of an image. Hence, a method named Triangle Point using Three Block (Tp3B) was proposed in this study. The proposed method proposes the arrangement of selection coordinate point based on three different blocks which where all coordinates points of an image were covered. Experiments have developed over image digit dataset of IFCHDB, HODA, MNIST and BANGLA which contains testing and train data of each. Features classification accuracy tested using supervised machine learning (SML) which is Support Vector Machine (SVM). Experimental results show, the proposed technique gives a promising result for dataset HODA and MNIST.