Page 62 - IJEEE-2022-Vol18-ISSUE-1

Received: 25 November 2021 Revised: 04 January 2022 Accepted: 12 January 2022
DOI: 10.37917/ijeee.18.1.7
Vol. 18| Issue 1| June 2022
? Open Access

Iraqi Journal for Electrical and Electronic Engineering

Original Article

Semantic Segmentation of Aerial Images Using U-Net
Architecture

Sarah Kamel Hussein1*, Khawla Hussein Ali2
Department of Computer Science, College of Education for Pure Sciences, University of Basrah, Basrah, Iraq

Correspondence
* Sarah Kamel Hussein
College of Education for Pure Sciences,
Education College for Pure Sciences,
University of Basrah, Basrah, Iraq
Email: cepsm510003@avicenna.uobasrah.edu.iq

khawla.ali@uobasrah.edu.iq

Abstract
Arial images are very high resolution. The automation for map generation and semantic segmentation of aerial images are

challenging problems in semantic segmentation. The semantic segmentation process does not give us precise details of the remote
sensing images due to the low resolution of the aerial images. Hence, we propose an algorithm U-Net Architecture to solve this
problem. It is classified into two paths. The compression path (also called: the encoder) is the first path and is used to capture the
image's context. The encoder is just a convolutional and maximal pooling layer stack. The symmetric expanding path (also called:
the decoder) is the second path, which is used to enable exact localization by transposed convolutions. This task is commonly
referred to as dense prediction, which is completely connected to each other and also with the former neurons which gives rise to
dense layers. Thus it is an end-to-end fully convolutional network (FCN), i.e. it only contains convolutional layers and does not
contain any dense layer because of which it can accept images of any size. The performance of the model will be evaluated by
improving the image using the proposed method U-NET and obtaining an improved image by measuring the accuracy compared
with the value of accuracy with previous methods.
KEYWORDS: U-Net, Deep Learning, Image Processing, Semantic Segmentation, Convolutional Neural Network (CNN),
Feature Extraction.

I. INTRODUCTION Convolutional Neural Network, or Recurrent Neural
Network [2].
Deep learning is one of the most controversial
technologies at this time, as its energy and ability to simulate Deep learning is a relatively recent term; As we did not
the human mind is very strange and frightening, deep really care about it until we had a lot of data during the
learning is a technology invented by humans in order to try modern technological revolution, and it is necessary in order
to imitate the way the human mind works, deep learning tries to make difficult decisions through big data, without any
to simulate the mind human being in all his abilities, of human intervention or limiting the data or properties, deep
which; Seeing, understanding speech, composing it, hearing, learning as we will see is the one who takes care of all these
and other powerful abilities that our human mind possesses Things are just like the human mind. Deep learning is a new
and is not rivaled by anything else. Not only did it stop at this science to a large extent, but it has roots in human
point, but scientists have studied the human brain and how it knowledge, and it has stages through which it has developed
works in order to design algorithms and programs capable of since the forties of the last century until the present day,
simulating it, and for this reason, we find that these these stages are the stage of cybernetics: which extended
algorithms are inspired by human medical and neurological from the forties of the twentieth century until the nineties,
studies and try as much as possible to imitate them, but in The stage of Connectionism, which was in the nineties and
computer ways not biological [1]. eighties of the last century, The stage of Deep Learning: It is
Deep learning is a section or part of machine learning that is what we now know as deep learning, and it started from 2006
part of the larger science called Artificial Intelligence. and extended until now [3][4].
Neurons or Neural Networks have been replaced by a
computer to become a perceptron or Artificial neural Semantic image segmentation is a form of dense
network, of which we now have many types such as: segmentation task in computer vision in which the model
outputs a dense feature map of the input RGB image with the
same dimensions (height and width) as the input image. The

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and
reproduction in any medium, provided the original work is properly cited.
© 2021 The Authors. Iraqi Journal for Electrical and Electronic Engineering by College of Engineering, University of Basrah.

https://doi.org/10.37917/ijeee.18.1.7 https://www.ijeee.edu.iq 58