© 2018, IJCERT All Rights Reserved https://doi.org/10.22362/ijcert/2018/v5/i2/v5i207 50
International Journal of Computer Engineering in Research Trends
Multidisciplinary, Open Access, Peer-Reviewed and fully refereed
Research Paper Volume-5, Issue-2 ,2018 Regular Edition E-ISSN: 2349-7084
Evaluation of Industrial Based Object Detection
Method Using Artificial Neural Network
F. S. Ishaq
1*
, I. A. Alhaji
2
, Halis Altun
3
, Y. Atomsa
4
, M. L. Jibrin
5
, S. A. Sani
6
1*
Dept. of Mathematics and Computer Science, Faculty of Science, Federal University, Kashere, Gombe, Nigeria
2
Dept. of Physics, Faculty of Science, Federal University, Kashere, Gombe, Nigeria
3
Dept. of Computer Engineering, Faculty of Engineering, Karatay University, Konya, Turkey
4
Dept. of Mathematics and Computer Science, Faculty of Science, Federal University, Kashere, Gombe, Nigeria
5
Dept. of Mathematics and Computer Science, Faculty of Science, Federal University, Kashere, Gombe, Nigeria
6
ICT Directorate, Federal University, Kashere, Gombe, Nigeria
F. S. Ishaq _ faisalsishaq@gmail.com, I. A. Alhaji _ alhaji259@gmail.com, Y. Atomsa _ au.nlaro@gmail.com, M. L.
Jibrin_mljtech@gmail.com, S. A. Sani _ aminushafiu@yahoo.com
*Corresponding Author: F. S. Ishaq1_ faisalsishaq@gmail.com
Available online at: http://www.ijcert.org
Received: 18/February/2017, Revised: 19/February/2017, Accepted: 23/February/2017, Published: 27/February/2017
Abstract:- The essence of the study is to analyse an algorithm which will provide a robust and
computationally light method, which might be suitable to implement in the real-time industrial application
such as object detection and recognition. For industrial applications, the primary step in automatic detection
and classification of an object is to find the object automatically from an image using features related to its
shape. This chore is a very complex one. Therefore, to hit the target Histogram of oriented gradient (HOG)
algorithm is selected to extract the image features. Average Magnitude Difference Function AMDF is
employed to correct the alignment defect. Finally, Artificial Neural Network (ANN) was employed to detect
the type of object in the image efficiently. None the less, a database was generated. The database consists
of images of real industrial products which are of different shapes and sizes, captured under different
lightning conditions. The outcome of the experiment conducted on the database recorded 98.10% success.
Keywords: 1-D mask, HOG algorithm, AMDF algorithm, k-nearest Neighbours algorithm, Cross-correlation
Functions algorithms and MLP algorithm
1. Introduction
For industrial applications, the primary step in
automatic detection and classification of an object is to
find the object automatically from an image using
features related to its shape. Therefore shape detection
plays an important role and has an intensive usage in
various applications [1] such as in robotics; object
classification is needed to recognize a certain object in
a cluttered scene [2]. In the industries detecting objects
With computers is difficult. This is as a result of some
Complications such as Initially, products from
industries vary in many forms such as type, size and
Shape. Hence, using object detection system in the
industries is only possible if the algorithm is robust
enough to eliminate these variations. Subsequently, the
industrial operation involves moving objects from one
place to the other; this might result in objects being
placed in different poses. The methods used needs to
distinguish the objects with different orientations
accurately.
Thirdly, the illumination in the environment varies due
to, shadows of other machine or humans. Therefore the