Advanced Machine Learning for Explosives Detection

Transportation Safety Officers (TSOs) currently use a multi-step decision tree to screen and resolve an alarm of a suspicious object in crowded and high-throughput environments, such as aviation and security checkpoints. In high-throughput environments, this can lead to information overload and mental fatigue. Homeland Security desires a more capable integrated Alarm Resolution (AR) sensor suite, with smart algorithms, to enhance their ability to accurately and efficiently detect hazardous materials.Streamline Automation (SA) and Summit TRC (STRC) Team (SA-STRC) proposes to develop an innovative Alarm Integrated Resolution System (AIRS) to:1) Characterize object utility, shape, size, color, and content phase 2) Detect presence of fingerprints 3) Accurately discriminate object content (normal verses simulated explosive)Many different organizations are moving to increased security measures to improve personnel safety and access control when entering their buildings. The SA/STRC team plans to market and employ this technology to other business sectors and federal/state government entities by responding to Broad Area Announcements (BAAs) and other Requests for Proposal (RFPs) from these organizations.The anticipated results of the Alarm Integrated Resolution System (AIRS) approach are: 1) the sensor suite shall differentiate the normal versus anomalous (inert explosive simulant) contents, enclosed within various material barriers;2) the confirmation that X-Ray computed tomography can determine if the content of an object is a gas, liquid, powder, or solid; 3) the confirmation that the digital camera can identify if there are fingerprints on the object; and 4) successfully develop the Machine Learning data input library format.