Integration of Sentinel-2 Semantic Segmentation Results with Spectral Indices for Forest Change Detection

Integration of Sentinel-2 Semantic Segmentation Results with Spectral Indices for Forest Change Detection
  IJETT-book-cover           
  
© 2025 by IJETT Journal
Volume-73 Issue-6
Year of Publication : 2025
Author : Shilpa P. Pimpalkar, Sai Madhavi D
DOI : 10.14445/22315381/IJETT-V73I6P119

How to Cite?
Shilpa P. Pimpalkar, Sai Madhavi D, "Integration of Sentinel-2 Semantic Segmentation Results with Spectral Indices for Forest Change Detection," International Journal of Engineering Trends and Technology, vol. 73, no. 6, pp.225-237, 2025. Crossref, https://doi.org/10.14445/22315381/IJETT-V73I6P119

Abstract
Deforestation presents significant challenges to biodiversity, high conservation-value species, global warming, and climate change. Recent developments in satellite image processing using deep learning and computer vision techniques have greatly improved methods for monitoring forest cover changes. This paper introduces a novel hybrid approach named AResU-Net, an attention-based residual-U-Net for forest change detection. The model generates a binary mask for the given input image leveraging Red (R), Green (G), and Blue (B) bands, along with the Near-Infrared (NIR) band from Sentinel-2 imagery. Spectral indices NDSI and NDVI identify snow/ice masks from the input image. Snow masks are removed from predicted masks to eliminate the possibility of incorrect deforestation detection due to seasonal variations, especially in winter. The effectiveness of the proposed model is validated by experimental results that demonstrate substantial improvements in key performance metrics: “Accuracy” - 0.964782, “Precision” - 0.946866, “Recall” - 0.968095, “F1 Score” - 0.957363, and “Mean Intersection over Union (mIOU)" - 0.929982. The evaluation metrics, complemented by visual analyses, indicate a strong correlation, confirming the model's effectiveness in accurately detecting forest changes. The performance assessment was conducted using a diverse array of validation images, including randomly selected .tif images of varying sizes sourced from Google Earth Engine in regions of Nepal, with an integration of the snow index. The AResU-Net model represents a significant advancement in automated image segmentation methodologies, contributing to environmental conservation efforts by effectively monitoring deforestation trends over time.

Keywords
Afforestation, Deforestation, Forest change, Residual, Sentinel.

References
[1] Kostiantyn Isaienkov et al., “Deep Learning for Regular Change Detection in Ukrainian Forest Ecosystem with Sentinel-2,” IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing, vol. 14, pp. 364-376, 2021.
[CrossRef] [Google Scholar] [Publisher Link]
[2] Kassim Kalinaki, Owais Ahmed Malik, and Daphne Teck Ching Lai, “FCD-Attresu-Net: An Improved Forest Change Detection in Sentinel-2 Satellite Images Using Attention Residual U-Net,” International Journal of Applied Earth Observation and Geoinformation, vol. 122, 2023.
[CrossRef] [Google Scholar] [Publisher Link]
[3] Flavie Pelletier et al., “Inter- and Intra-Year Forest Change Detection and Monitoring of Aboveground Biomass Dynamics Using Sentinel-2 and Landsat,” Remote Sensing of Environment, vol. 301, 2024.
[CrossRef] [Google Scholar] [Publisher Link]
[4] Ana María Pacheco-Pascagaza et al., “Near Real-Time Change Detection System Using Sentinel-2 and Machine Learning: A Test for Mexican and Colombian Forests,” Remote Sensing, vol. 14, no. 3, pp. 1-21, 2022.
[CrossRef] [Google Scholar] [Publisher Link]
[5] Ali Jamali et al., “TransU-Net++: Rethinking Attention Gated Trans U-Net for Deforestation Mapping,” International Journal of Applied Earth Observation and Geoinformation, vol. 120, 2023.
[CrossRef] [Google Scholar] [Publisher Link]
[6] David John, and Ce Zhang, “An Attention-Based U-Net for Detecting Deforestation Within Satellite Sensor Imagery,” International Journal of Applied Earth Observation and Geoinformation, vol. 107, 2022.
[CrossRef] [Google Scholar] [Publisher Link]
[7] Dongoo Lee, and Yeonju Choi, “A Learning Strategy for Amazon Deforestation Estimations Using Multi-Modal Satellite Imagery,” Remote Sensing, vol. 15, no. 21, pp. 1-20, 2023.
[CrossRef] [Google Scholar] [Publisher Link]
[8] John Burns Kilbride et al., “A Near Real-Time Mapping of Tropical Forest Disturbance Using SAR and Semantic Segmentation in Google Earth Engine,” Remote Sensing, vol. 15, no. 21, pp. 1-20, 2023.
[CrossRef] [Google Scholar] [Publisher Link]
[9] Maria Kaselimi et al., “A Vision Transformer Model for Convolution-Free Multilabel Classification of Satellite Imagery in Deforestation Monitoring,” IEEE Transactions on Neural Networks and Learning Systems, vol. 34, no. 7, pp. 3299-3307, 2023.
[CrossRef] [Google Scholar] [Publisher Link]
[10] See Ven Lim et al., “Attention-Based Semantic Segmentation Networks for Forest Applications,” Forests, vol. 14, no. 12, pp. 1-22, 2023.
[CrossRef] [Google Scholar] [Publisher Link]
[11] P. Tovar, M.O. Adarme, and R.Q. Feitosa, “Deforestation Detection in the Amazon Rainforest with Spatial and Channel Attention Mechanisms,” The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, vol. 43-B3-2021, pp. 851-858, 2021.
[CrossRef] [Google Scholar] [Publisher Link]
[12] K. Karaman, V. Sainte Fare Garnot, and J.D. Wegner, “Deforestation Detection in the Amazon with Sentinel-1 Sar Image Time Series,” ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences, vol. 10-1/W1-2023, pp. 835-842, 2023.
[CrossRef] [Google Scholar] [Publisher Link]
[13] Imran Md Jelas et al., “Deforestation Detection Using Deep Learning-Based Semantic Segmentation Techniques: A Systematic Review,” Frontiers in Forests and Global Change, vol. 7, pp. 1-24, 2024.
[CrossRef] [Google Scholar] [Publisher Link]
[14] Daliana Lobo Torres et al., “Deforestation Detection with Fully Convolutional Networks in the Amazon Forest from Landsat-8 and Sentinel-2 Images,” Remote Sensing, vol. 13, no. 24, pp. 1-20, 2021.
[CrossRef] [Google Scholar] [Publisher Link]
[15] P.J. Soto et al., “Deforestation Detection with Weak Supervised Convolutional Neural Networks in Tropical Biomes,” The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, vol. 43-B3-2022, pp. 713-719, 2022.
[CrossRef] [Google Scholar] [Publisher Link]
[16] Alexandru-Toma Andrei, and Ovidiu Grigore, “Low-Cost Optimized U-Net Model with GMM Automatic Labeling Used in Forest Semantic Segmentation,” Sensors, vol. 23, no. 21, pp. 1-20, 2023.
[CrossRef] [Google Scholar] [Publisher Link]
[17] Bruno Menini Matosak et al., “Mapping Deforestation in Cerrado Based on Hybrid Deep Learning Architecture and Medium Spatial Resolution Satellite Time Series,” Remote Sensing, vol. 14, no. 1, pp. 1-22, 2022.
[CrossRef] [Google Scholar] [Publisher Link]
[18] Zhipan Wang et al., “Repddnet: A Fast and Accurate Deforestation Detection Model with High-Resolution Remote Sensing Image,” International Journal of Digital Earth, vol. 16, no. 1, pp. 2013-2033, 2023.
[CrossRef] [Google Scholar] [Publisher Link]
[19] Abdul Haseeb Azizi et al., “Machine Learning-Based Estimation of Fractional Snow Cover in the Hindukush Mountains Using MODIS And Landsat Data,” Journal of Hydrology, vol. 638, 2024.
[CrossRef] [Google Scholar] [Publisher Link]
[20] Charlotte Poussin, Pascal Peduzzi, and Gregory Giuliani, “Snow Observation from Space: An Approach to Improving Snow Cover Detection Using Four Decades of Landsat and Sentinel-2 Imageries Across Switzerland,” Science of Remote Sensing, vol. 11, 2025.
[CrossRef] [Google Scholar] [Publisher Link]
[21] Yunlong Wang, and Jianshun Wang, “Monitoring Snow Cover in Typical Forested Areas Using a Multi-Spectral Feature Fusion Approach,” Atmosphere, vol. 15, no. 4, pp. 1-14, 2024.
[CrossRef] [Google Scholar] [Publisher Link]
[22] Gongxue Wang et al., “Characterization of NDSI Variation: Implications for Snow Cover Mapping,” IEEE Transactions on Geoscience and Remote Sensing, vol. 60, pp. 1-18, 2022.
[CrossRef] [Google Scholar] [Publisher Link]
[23] Rui He et al., “Effective Improvement of the Accuracy of Snow Cover Discrimination Using a Random Forests Algorithm Considering Multiple Factors: A Case Study of the Three-Rivers Headwater Region, Tibet Plateau,” Remote Sensing, vol. 15, no. 19, pp. 1-21, 2023.
[CrossRef] [Google Scholar] [Publisher Link]
[24] Renan Bides de Andrade, Guilherme Lucio Abelha Mota, and Gilson Alexandre Ostwald Pedro da Costa, “Deforestation Detection in the Amazon Using DeepLabv3+ Semantic Segmentation Model Variants,” Remote Sensing, vol. 14, no. 19, pp. 1-19, 2022.
[CrossRef] [Google Scholar] [Publisher Link]