Advanced Machine Learning for Explosives Detection

Streamline Automation and Summit Technology Research Corporation will develop and demonstrate an automated Alarm Integrated Resolution System (AIRS) to characterize Suspicious Objects (SOs), which TSOs have flagged as potentially containing hazardous materials. These objects will be placed in the AIRS for comparison to a well-characterized data library, developed with Machine Learning (ML). The AIRS will automate the Alarm Resolution (AR) decision tree for the TSO by recommending the next, optimum follow-on evaluation/testing/analysis steps. These post-AIRS steps could include Raman and/or Mass Spectroscopy, chemical analyses, etc., to accurately identify the hazardous material. The AIRS Module will contain digital cameras, a fingerprint detection spray, and a digital X-ray system, which will provide basic characterization of the SO. ML algorithms will be developed, including an object-recognition CNN - a class of artificial neural networks - proven to be extremely effective in analysis of visual imagery. The use of CNN technology will increase the probability of correctly categorizing objects according to their utilities/functions and the probability of recommending the optimum AR tools. A Software Graphical User Interface will be developed to provide the TSO an easily-understood summary of the AIRS analysis and greatly reduce the time and fatigue for TSO personnel.