CBL - Campus del Baix Llobregat

Projecte llegit

Títol: Foreign Object Debris Detection on Runways by Computer Vision from Drones

Estudiants que han llegit aquest projecte:

Director/a: PASTOR LLORENS, ENRIC

Departament: DAC

Títol: Foreign Object Debris Detection on Runways by Computer Vision from Drones

Data inici oferta: 02-10-2022     Data finalització oferta: 02-05-2023


Estudis d'assignació del projecte:
    GR ENG SIS TELECOMUN
    GR ENG SIST AEROESP
    GR ENG TELEMÀTICA
Tipus: Individual
 
Lloc de realització: EETAC
 
Paraules clau:
FOD, AI Algorithms, Drones
 
Descripció del contingut i pla d'activitats:
Detection of FOD objects over operative runways through the
acquisition of visual images taken from a dedicated drone. The
project will determine the best AI algorithms that could be
employed and will train the algorithms by creating a dedicated
image dataset. Trajectory planning will be automated according to
the onboard camera's runway parameters and specifications.
 
Overview (resum en anglès):
The Foreign Object Debris (FOD) is an important topic to discuss in aeronautics as
it not only involves safety but also can produce huge economic losses in airports. It
is considered FOD any object that should not be in the place it is and as a
consequence of this can be a risk to aircrafts or workers.
The total cost of FOD to the aerospace industry is calculated to be approximately 4
billion dollars per year. Although new technologies are being implemented, it is a
field that still has enhancements to be made.
This project is intended to improve the detection of FOD in airports with AI
technology applied through a drone with object recognition algorithms that allow to
meet the frequencies of inspection and search of operational areas at a low cost of
implementation and operation compared to other current detection technologies.
The project develops the necessary system architecture to carry out FOD
inspections at airports. It comprises studying the requirements, characteristics of the
devices involved in the process, and adapting the configuration and parameters of
the equipment (drone and camera) to ensure feasibility across all categories of
runways. A method for data acquisition, management, and transmission is proposed
to determine the position and obtain real-time images of the detected objects. Topics
such as the Ground Base Station (GBS), drone tracking, charging station, flight
plans, and routes are also addressed.
To complete the designed system architecture, an AI model is trained using a FOD
database, which will be responsible for obtaining detections during inspections. A
new FOD database has also been developed to expand the currently available
database. The project's obtained structure is tested through drone flights that
simulate real-world applications. The results obtained reveal a functional model with
potential for real-world application, but with two current limitations: inspection time
and the open nature of FOD. Proposals and new approaches are presented to
address these restrictions in future studies

© CBLTIC Campus del Baix Llobregat - UPC