Object-Based Supervised Change Detection Using Extracted Features from VHR Imagery

Document Type : Original Article

Authors

School of Surveying and Geospatial Engineering, College of Engineering, University of Tehran, Tehran, Iran

10.22059/eoge.2025.384576.1162

Abstract

This study addresses the challenges of change detection in Very High Resolution (VHR) satellite imagery, which are characterized by complex feature interactions, noise sensitivity, and intricate land use changes. We propose a novel approach that integrates object-based, feature-based, and dual learning-based methods to enhance the detection of changes on the Earth's surface using bi-temporal images. Our methodology begins with the separate segmentation of VHR images to manage complexities and account for pixel relationships. We then enrich spectral bands with additional textural, mathematical, and geometrical features derived from paired images to conduct a comprehensive conceptual analysis. Finally, we implement two categories of supervised classifiers—individual and ensemble—to generate a binary change map that classifies regions as either changed or unchanged. The results from experiments on two VHR datasets demonstrate that our approach outperforms traditional methods, achieving the highest F1-score and Intersection over Union (IoU) of 99.02% and 96.78%, respectively. The novelty of this research lies in the integration of object-based, feature-based, and learning-based approaches, leading to a comprehensive feature extraction process. Additionally, our method creates detailed change detection maps with numerous homogeneous areas, contributing significantly to advancements in the field.

Keywords

Main Subjects