Unconstrained Object Understanding using Deep Learning for Computer Vision in X-ray Security Screening

About the Project

X-ray security screening is widely used to maintain aviation/transport security, and its significance poses a particular interest in future automated screening systems that can be deployed for transportation and border security screening alike. Within this context deep learning based computer vision techniques have already been used to enable automated prohibited and threat item detection spanning X-ray scanned baggage, freight and postal items. However, the complexity, variety and unconstrained nature of X-ray security imagery in terms of the variation of objects present, object orientation, inter-object occlusion and complex concealment by adversaries continue to present a range of challenges for effective and robust automated screening approaches. This project aims to address these challenges by leveraging a range of recent advances in aligned domains spanning both computer vision and deep machine learning research.

There are a wide range of potential research directions, that could form the basis for a specific PhD project in this area:

• material and appearance based anomaly detection
• multi-view object detection and classification
• adaptation of AI foundational models, including visual-language models (LLM/VLM)
• occluded and disassembled object detection
• transparent object segmentation and extraction
• category discovery and out of distribution detection
• complex, multi-part object detection