Work

A Sentinel-1 SAR imagery dataset for airstrips detection and segmentation in the Brazilian Amazon Rainforest

Year: 2026

Type: data-paper

Abstract: The Brazilian Amazon Rainforest holds a large ecological and economic importance and is considered one of the most biodiverse regions on the planet. The region faces numerous challenges from illegal human activities that threaten its sustainability and well-being, which are often supported by the construction of unauthorized airstrips. Additionally, due to its persistent cloud cover, which often hinders monitoring with optical satellites, Synthetic Aperture Radar (SAR) imagery provides a crucial alternative for the region surveillance. Thus, this dataset was developed to support the training and evaluation of machine learning techniques, including deep learning models for detecting and segmenting airstrips in the Brazilian Amazon Rainforest using SAR imagery. The dataset comprises images from the Sentinel-1 satellite, acquired primarily between 2021 and 2024, covering 1040 locations of known airstrips sourced from the MapBiomas project (published in 2023, based on 2021 reference data). For the change detection task, historical "before" images were selected from the period between 2014 and 2021 to capture the pre-construction state. The data is structured to support three distinct machine learning tasks: object detection (e.g., YOLOv8), semantic segmentation (e.g., U-Net), and change detection. For each task, specific images and annotations are provided. Additionally, geospatial files (Shapefile, GeoPackage) are included to facilitate the integration and visualization of the dataset in a GIS environment. The data is valuable for researchers in remote sensing, computer vision, environmental monitoring, security and defense, enabling the development of automated systems to monitor irregular activities in remote forest regions. The dataset is available at a Mendeley Data repository: https://data.mendeley.com/datasets/x7rn78ymtn/1.

Source: Data in Brief

DOI: 10.1016/j.dib.2026.112472

Authors: Leandro da Silva Gomes, Gustavo Henrique de Queiroz Stabile, Tahisa Neitzel Kuck, Felipe Augusto Pereira de Figueiredo, Elcio Hideiti Shiguemori, Dimas Irion Alves

Institutions: Instituto Tecnológico de Aeronáutica, National Institute of Telecommunications

Language: English

Topic: Remote-Sensing Image Classification

Subfield: Media Technology

Field: Engineering

Domain: Physical Sciences

SDGs: Life on land, Industry, innovation and infrastructure

Open Access status: gold

Estimated APC paid: $1,220