Identification of Deforestation Areas in the Amazon Rainforest Using Change Detection Models

TL;DR

This study reframes Amazon deforestation detection as a change-detection problem using PRODES data and Landsat imagery, building a unified dataset to benchmark CNN and Transformer-based detectors. It systematically evaluates four architectures (UNet++, MultiResUNet, TransUNet, SwinUNETR-V2), applies targeted pre/post-processing (histogram equalization, magenta-pattern replacement), and explores three ensemble strategies, achieving an overall F1-score of 80.41% with FCN-based fusion. The findings show Transformer-based models benefit from global context but require careful data handling, while ensemble fusion yields the strongest performance, offering a practical, scalable approach for monitoring deforestation. The work emphasizes reproducibility and provides insights for applying similar pipelines to other biomes, with future directions including cloud occlusion handling and alternative loss functions.

Abstract

The preservation of the Amazon Rainforest is one of the global priorities in combating climate change, protecting biodiversity, and safeguarding indigenous cultures. The Satellite-based Monitoring Project of Deforestation in the Brazilian Legal Amazon (PRODES), a project of the National Institute for Space Research (INPE), stands out as a fundamental initiative in this effort, annually monitoring deforested areas not only in the Amazon but also in other Brazilian biomes. Recently, machine learning models have been developed using PRODES data to support this effort through the comparative analysis of multitemporal satellite images, treating deforestation detection as a change detection problem. However, existing approaches present significant limitations: models evaluated in the literature still show unsatisfactory effectiveness, many do not incorporate modern architectures, such as those based on self-attention mechanisms, and there is a lack of methodological standardization that allows direct comparisons between different studies. In this work, we address these gaps by evaluating various change detection models in a unified dataset, including fully convolutional models and networks incorporating self-attention mechanisms based on Transformers. We investigate the impact of different pre- and post-processing techniques, such as filtering deforested areas predicted by the models based on the size of connected components, texture replacement, and image enhancements; we demonstrate that such approaches can significantly improve individual model effectiveness. Additionally, we test different strategies for combining the evaluated models to achieve results superior to those obtained individually, reaching an F1-score of 80.41%, a value comparable to other recent works in the literature.

Figures (14)

• Figure 1: Map of the coverage area of the Brazilian Amazon biome. Data obtained from prodes.
• Figure 2: Diagram of (a) single-stream and (b) double-stream networks for change detection. Adapted from cd_review_2020.
• Figure 3: Methodological workflow for data processing, from PRODES and Landsat sources to the final training and test sets for change detection models.
• Figure 4: (a) Map showing locations of selected Landsat scenes for training and testing. (b-d) 2019 imagery of Landsat scenes. Adapted from pozzobon20.
• Figure 5: Examples of samples from the dataset adapted for the problem of identifying deforested areas through change detection. The images were adjusted to be visualized in an appropriate color space.