Author ORCID Identifier

0000-0002-4766-7297

Document Type

Conference Paper

Rights

Available under a Creative Commons Attribution Non-Commercial Share Alike 4.0 International Licence

Disciplines

1.2 COMPUTER AND INFORMATION SCIENCE, Geosciences, (multidisciplinary)

Publication Details

ICGDA 2022: 2022 5th International Conference on Geoinformatics and Data Analysis Paris, France, January 21-23, 2022

Abstract

Detecting changes to built environment objects such as buildings/roads/etc. in aerial/satellite (spatial) imagery is necessary to keep online maps and various value-added LBS applications up-to-date. However, recognising such changes automatically is not a trivial task, and there are many different approaches to this problem in the literature. This paper proposes an automated end-to-end workflow to address this problem by combining OpenStreetMap (OSM) vectors of building footprints with a machine learning Generative Adversarial Network (GAN) model - where two neural networks compete to become more accurate at predicting changes to building objects in spatial imagery. Notably, our proposed OSM-GAN architecture achieved over 88% accuracy predicting/detecting building object changes in high-resolution spatial imagery of Dublin city centre.

DOI

https://doi.org/10.21427/07C4-BY87

Funder

COAT


Share

COinS