Abstract
In the 20th century, aviation has proven itself as one of the most important
techno-social revolutions of modern human history. Now,
Unmanned Aerial Systems (UAS) are on the verge of leading the
next techno-social revolution, remodelling several areas and moving
a global market that does not stop growing. The establishment
of UAV Traffic Management (UTM) systems is necessary to make
this revolution happen, and one of the main obstacles is the risks
involved in UAS flight, especially in urban environments. A methodology
to estimate risks on the overflown environment was carried
out by the usage of convolutional neural networks on the imagery
of such environments, allowing the definition of safer routes and
their management on the fly. Development and experimentation
processes were carried out, with promising results including a convolutional
neural network for pixel-wise domain that was capable
of estimate risks from satellite imagery and return overflight risks
heat maps.