Object Identification and Segmentation using Student Developed Machine Learning (ML) Algorithm for Unmanned Aerial System (UAS) Thermal Imaging

Read full text Summarise with VIRAT
Authors: Adeel Khalid
Cited by: 0
References: 10

Summary

No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.

Summarise with VIRAT Free to try with your Science Catalogs account.

Abstract

In this project, students help design a Machine Learning (ML) Algorithm to help segment and annotate thermal images of various objects.Several thermal images are captured using Unmanned Aerial System (UAS) mounted Forward Looking Infrared (FLIR) camera.These images contain pixel level radiometric data.A team of students ranging from high school to graduate level are tasked with training a YOLO (You Only Look Once) based Machine Learning Algorithm.They start with a library of existing objects that includes the ability to identify people, cars, houses and other objects.They train and refine the model with additional data captured from various camera pitch angles and aircraft altitudes.They compare the fidelity and the confidence level of the images in order to further increase the accuracy of annotation.The applications include but are not limited to identifying hot spots in new and old buildings, solar panel inspection, identification of other UASs in restricted airspaces, pavement and runway inspection.Results obtained from pre-trained and custom Artificial Intelligence (AI) models are discussed.Student involvement, recruitment, training, research process, findings and conclusions are discussed.

Fields of Study & Subjects

Advanced Neural Network Applications Infrared Target Detection Methodologies Infrastructure Maintenance and Monitoring

Publication details

Journal / source
ASEE Southeast Section Conference
Publication date
31 March 2026
Language
EN