Title: Machine learning-based land usage identification using Haralick texture features of aerial images with Kekre's LUV colour space

Authors: Sudeep D. Thepade; Shalakha Vijaykumar Bang; Rik Das; Zahid Akhtar

Addresses: Department of Computer Engineering, Pimpri Chinchwad College of Engineering, Savitribai Phule Pune University, Pune, India ' Department of Computer Engineering, Pimpri Chinchwad College of Engineering, Savitribai Phule Pune University, Pune, India ' Siemens Advanta, Siemens Technology and Services Private Limited, Bengaluru, India ' Department of Network and Computer Security, State University of New York Polytechnic Institute, Utica, NY, USA

Abstract: Study of gathering some useful insights from our planet Earth - its natural, man-made, physical, and biological structures is quite engrossing. Earth observation despite being intuitive, also helps in mitigating the adverse impacts of human civilisation on our mother Earth. Multiple techniques that help in observing the Earth's surface include Earth surveying techniques, remote-sensing technology, etc. The properties which are measured using remote-sensing technology stimulate the study of land usage identification which refers to the purpose the land is used for. The rapid increase in population, immense growth in infrastructure and technology has led to massive urbanisation posing a great number of challenges. The knowledge of land use identification will help in developing strategies to tackle issues related to the depletion of forest areas, urban encroachment, monitoring of natural disasters, etc. This paper attempts to give a more robust approach towards land usage identification that extracts Haralick texture features from input aerial images of the Earth by considering their representation in two different colour spaces namely RGB and Kekre-LUV. Comparing the results obtained by using different machine learning classification algorithms, it is found that an ensemble of simple logistic and random forest classifiers outputs maximum classification accuracy.

Keywords: grey level co-occurrence matrix; GLCM; random forest; simple logistic regression; land usage identification; remote sensing.

DOI: 10.1504/IJCSE.2022.126255

International Journal of Computational Science and Engineering, 2022 Vol.25 No.5, pp.562 - 572

Received: 31 Mar 2021
Accepted: 16 Sep 2021
Published online: 18 Oct 2022 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article