Improving baggage screening effectiveness
Improving baggage screening using collective intelligence and machine learning
At many U.S. border entry points, baggage screening features modern computed tomography (CT) equipment for primary baggage screening. However, the effectiveness of this process often relies on the judgment of individual transportation security officers (TSOs), who must then determine whether suspicious bags should be cleared or be manually inspected. In many cases, these decisions could lead to excessive numbers of false positives or clearing dangerous bags that pose substantial risks.
This project studies conventional baggage screening with the goal of creating a functioning prototype of an intelligent group screening tool to increase accuracy and reduce human error. This tool can be used to guide the evolution of new baggage screening solutions for major U.S. airports and other border entry points.
Through research and analysis, the project team will develop sophisticated tools that integrate human intelligence and machine learning to dramatically improve screening accuracy. As part of this process, the team will generate synthetic image data sets to replicate the cognitive challenges of baggage screening tasks. Armed with that information, a process for human agents will be developed that improves visual screening classification efficiency at lower costs using scalable models. Finally, optimized algorithms will aggregate data inputs to mimic real life scenarios for the purposes of evaluation and refinement of the design.
As part of an improved screening process, this machine learning-enhanced, group screening tool learns from - and adapts to - the strengths and weaknesses of individual TSO agents. This improves the accuracy of threat screening and classification decisions. This solution may also serve as a guide in the ongoing development of enhanced baggage screening solutions for major U.S. airports and other border entry points.
Research Leadership Team
Principal Investigator: Adolfo Escobedo, Arizona State University
Co-PI: Olac Fuentes, University of Texas at El Paso