According to the Brasil Escola website, deforestation in the Amazon is currently a cause for concern for researchers, environmentalists, traditional populations that depend on forest resources and various sectors of national and international civil society. This is because, unlike previous periods, there has again been an increase in the rates of removal of vegetation cover, which began in 2017 and has maintained an upward movement and growth trend, as shown by the most recent data from Inpe.
Also according to Brasil Escola, the increase in the deforested area, between 2017 and 2021, was 87.6%, jumping from 6,947 km² to 13,038 km². The intervals in which the greatest acceleration in the deforestation rate occurred are 2018-2019, in which Inpe identified an increase of 34.4%, and 2020-2021, when the deforested area increased by 20.5%. In the same period of time, there was an increase in fires and burnings in the Amazon, a phenomenon that has been caused with the intention of opening new areas in the midst of vegetation. As a result, recent years have been marked by the worsening of deforestation in the Amazon and the relaxation of environmental regulations, with a reduction in inspection of threatened areas and fewer fines for those who practice illegal deforestation.
According to the authors of the article “Paving the Way for Automatic Mapping of Rural Roads in the Amazon Rainforest”, deforestation of the Amazon rainforest severely impacts the environment in many ways, including the reduction of biodiversity, climate change and several other chains. “A key indicator of deforestation is the sudden appearance of unofficial rural roads, often exploited to transport raw materials extracted from the forest. To detect roads early and prevent deforestation, remote sensing images have been widely used. Precisely, some researchers have focused on tackling this task using low-resolution images, mainly due to their public availability and long time series,” they explained.
However, according to the scientists, extracting roads using low-resolution images presents several challenges, many of which are not addressed in existing work, including high similarity between classes, structural complexes, among other issues,” they explained.
Motivated by this, the authors proposed an approach to perform low-resolution satellite road extraction with images based on contextual and pixel-level decision fusion. “We conducted a systematic evaluation of the proposed method using a new set of data presented in this work. The experiments show that the proposed method outperforms state-of-the-art algorithms in terms of intersection over union and F1 score,” they concluded.
The study was carried out by researchers Lucas Costa de Faria and Matheus Brito, from the Department of Computer Science (DCC) at UFMG; Keiller Nogueira, from the Department of Computer Science and Mathematics at the University of Stirling, in the United Kingdom, and Jefersson A. dos Santos, from the Department of Computer Science at the University of Sheffield, in the United Kingdom.