Deep Learning Application for Urban Change Detection from Aerial Images

Tautvydas Fyleris, Andrius Kriščiūnas, Valentas Gružauskas, Dalia Čalnerytė

Abstract

Urban growth estimation is an essential part of urban planning in order to ensure sustainable regional development. For such purpose, analysis of remote sensing data can be used. The difficulty in analysing a time series of remote sensing data lies in ensuring that the accuracy stays stable in different periods. In this publication, aerial images were analysed for three periods, which lasted for 9 years. The main issues arose due to the different quality of images, which lead to bias between periods. Consequently, this results in difficulties in interpreting whether the urban growth actually happened, or it was identified due to the incorrect segmentation of images. To overcome this issue, datasets were generated to train the convolutional neural network (CNN) and transfer learning technique has been applied. Finally, the results obtained with the created CNN of different periods enable to implement different approaches to detect, analyse and interpret urban changes for the policymakers and investors on different levels as a map, grid, or contour map.

Download


Paper Citation


in Harvard Style

Fyleris T., Kriščiūnas A., Gružauskas V. and Čalnerytė D. (2021). Deep Learning Application for Urban Change Detection from Aerial Images. In Proceedings of the 7th International Conference on Geographical Information Systems Theory, Applications and Management - Volume 1: GISTAM, ISBN 978-989-758-503-6, pages 15-24. DOI: 10.5220/0010415700150024


in Bibtex Style

@conference{gistam21,
author={Tautvydas Fyleris and Andrius Kriščiūnas and Valentas Gružauskas and Dalia Čalnerytė},
title={Deep Learning Application for Urban Change Detection from Aerial Images},
booktitle={Proceedings of the 7th International Conference on Geographical Information Systems Theory, Applications and Management - Volume 1: GISTAM,},
year={2021},
pages={15-24},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0010415700150024},
isbn={978-989-758-503-6},
}


in EndNote Style

TY - CONF

JO - Proceedings of the 7th International Conference on Geographical Information Systems Theory, Applications and Management - Volume 1: GISTAM,
TI - Deep Learning Application for Urban Change Detection from Aerial Images
SN - 978-989-758-503-6
AU - Fyleris T.
AU - Kriščiūnas A.
AU - Gružauskas V.
AU - ÄŒalnerytÄ— D.
PY - 2021
SP - 15
EP - 24
DO - 10.5220/0010415700150024