Date of Graduation

Summer 2024

Degree

Master of Natural and Applied Science in Computer Science

Department

Computer Science

Committee Chair

Lloyd Smith

Abstract

This work proposes an artificial intelligence model based on U-Net architecture to map road networks in the Brazilian Amazon. Over the years, the Amazon region has been heavily exploited, leading to increased deforestation rates, contributing to CO2 emissions, amplifying global warming, and causing a disturbance in local fauna and flora. The expansion into the forest by illegal miners, loggers, and land grabbers can be tracked down by the construction of roads, which we can refer to as the arteries of deforestation. Previous works on the matter proposed algorithms that use high-resolution imagery to map roads precisely. However, this work approach goes a step further and proposes the usage of medium-resolution imagery from Landsat satellite to map the roads of the Amazon, spatially and temporally. By taking advantage of years of data acquired by the Landsat satellite family, I was able to create a historical map of roads in the Amazon from 1985 to 2020, which provides information on the process of expansion of roads into the Amazon region. The model was trained on Landsat 8 imagery using more than 3 million kilometers of roads as its dataset, reaching 0.6577, 0.6237l, and 0.6401 for precision, recall, and f1-score metrics. These results are important for predictive models of high-risk areas of deforestation, allowing for more precise estimation of roadless forests, having been made publicly available.

Keywords

artificial Intelligence, remote sensing, road network, deep learning, cloud-based platforms

Subject Categories

Artificial Intelligence and Robotics | Data Science | Environmental Indicators and Impact Assessment | Geographic Information Sciences | Remote Sensing | Spatial Science

Copyright

© Jonas Paiva Botelho Jr

Download

Open Access

Share

COinS