Smart Aviation: Intelligent Analytics for Flight Safety and Airport Operations

Smart Aviation: Intelligent Analytics for Flight Safety and Airport Operations This project develops deep learning models for analyzing airport operational voice data to improve flight safety and airport operations. We address challenges such as domain-specific terminology, non-standard communication patterns, limited labeled data, and real-time inference by designing domain-adapted learning strategies that learn robust representations from air traffic communication, supported by specialized pre-training, self-supervised learning, data augmentation, and efficient model design.

Funded: Nebraska University (NU) Collaboration Initiative 

Collaborator: UNO Aviation Institute

Selected Publication:

• Dual-Pipeline Machine Learning Framework for Automated Interpretation of Pilot Communications [https://www.mdpi.com/2226-4310/13/1/32];

Foreign Object Debris Detection with Deep Learning

Foreign Object Debris (FOD) refers to any material on airport surfaces that can threaten aircraft safety. This project develops an integrated detection system that leverages deep learning and computer vision with unmanned aerial platforms to enable efficient and scalable FOD monitoring, addressing cost and operational challenges in small and mid-sized airports. 

Funded: 

• NASA Nebraska Space Grant

• Nebraska University (NU) Collaboration Initiative 

Collaborator: UNO Aviation Institute

Selected Publications: 

• [1]  Self-supervised Foreign Object Debris Detection (https://arxiv.org/abs/2210.16901);      Dataset

• [2] FOD-A: A Dataset for Foreign Object Debris in Airports (https://arxiv.org/abs/2110.03072);      Dataset

Multimodal Representation Learning for Classroom Analysis 

This project develops multimodal representation learning methods to model instructional processes in classroom video data. By integrating visual and auditory signals, the approach learns representations that capture teaching behaviors and interaction patterns, enabling more accurate and scalable analysis of classroom instruction. 

Collaborators: UNL Nebraska Academy for Methodology, Analytics and Psychometrics; UNL Teaching, Learning & Teacher Education  

Selected Publications: 

• Artificial Intelligence (AI) to Enhance Computer Science Instruction [IEEE Xplore Link]