Title: LULC classification using deep convolution neural networks for change detection analysis

Authors: H.N. Mahendra; V. Pushpalatha; S. Mallikarjunaswamy; D. Mahesh Kumar; H.S. Ganesha; Rama Subramoniam Sudalayandi; Trupthi Rao

Addresses: JSS Academy of Technical Education, Bengaluru, Karnataka – 560060, India; Affiliated to: Visvesvaraya Technological University, Belagavi, India ' JSS Academy of Technical Education, Bengaluru, Karnataka – 560060, India; Affiliated to: Visvesvaraya Technological University, Belagavi, India ' JSS Academy of Technical Education, Bengaluru, Karnataka – 560060, India; Affiliated to: Visvesvaraya Technological University, Belagavi, India ' JSS Academy of Technical Education, Bengaluru, Karnataka – 560060, India; Affiliated to: Visvesvaraya Technological University, Belagavi, India ' Department of Electronics and Communication Engineering, School of Engineering, JSS University, Noida, Uttar Pradesh – 201301, India ' Regional Remote Sensing Centre (South), National Remote Sensing Centre, Indian Space Research Organization, Bengaluru, Karnataka, India ' Global Academy of Technology, Bengaluru, Karnataka -560098, India; Affiliated to: Visvesvaraya Technological University, Belagavi, India

Abstract: Land use and land cover (LULC) classification is a fundamental task for monitoring environmental changes, planning sustainable land management, and change detection studies. Traditional classification methods often rely on manual interpretation, which can be time-consuming and limited in accuracy. In this study, we proposed a novel approach based on deep convolutional neural networks (DCNNs) for LULC classification. The proposed methodology involves the collection of multi-temporal satellite imagery datasets covering the study area of Mysuru taluk, Karnataka State, India. Preprocessing techniques are applied to improve the quality of the input data, including normalisation, filtering, and geometric correction. Subsequently, a DCNN architecture is designed and trained using labelled datasets to classify land cover types accurately for the satellite data of 2012 and 2022. The classification accuracies achieved, 92.70% for 2012 and 94.79% for 2022, highlight the capability of DCNNs in LULC classification. Further, classified maps are used to perform the change detection analysis using the post-classification comparison technique. The results of this research contribute to a better understanding of the land use dynamics of the study area and provide valuable insights for land management and policymaking.

Keywords: deep learning; land use and land cover; LULC; deep convolutional neural networks; DCNNs; geographic information systems; GIS; multispectral data.

DOI: 10.1504/IJESD.2026.152026

International Journal of Environment and Sustainable Development, 2026 Vol.25 No.1/2, pp.98 - 121

Received: 23 May 2024
Accepted: 24 Jan 2025
Published online: 04 Mar 2026 *

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