Machine Learning Based Integration of Alarm Resolution Sensors

The purposes of the work is to determine feasibility of confidently characterizing shape, size and color, product brand name, volume, weight, density, utility (i.e., books, electronics, liquids, food, toiletries, etc.) of alarming objects and composition of their container (i.e., plastic, glass, metal, etc.) from collection and analysis of physical imagery of such objects, product barcode and gravimetrics sensor. Additionally, determine the feasibility of confidently identifying external fingerprints on alarming objects from collection and analysis of UV images determining solid / liquid content of alarming objects from collection and analysis of gravimetric data. Supporting TSA’s Stream of Commerce (SoC) Data Collection Program ANC has developed autonomous, mobile sensors to collect Alarm Resolution (AR) data leveraging an AI-vision engine. This is the main element of the Joint Alarm Resolution Controller (JARC). Since 2018 ANC has collected over hundreds of thousands of X-ray images, data and physical images of thousands of alarming items. ANC will use its AI engine and data to analyze these physical images and augment it with outputs from other sensors (gravimetrics, UV, ultrasound, etc.) used on a smaller set of lab-collected items and containers. The AI engine will be extended to include all sensor outputs to provide a fused output and ultimately be presented to TSOs recommending an optimal AR TSE and procedure based on the alarming item characteristics. ANC’s commercialization strategy is to develop a product for TSA and other Federal customers with similar missions (CBP, USSS) and international aviation security customers in need of optimizing alarm resolution.