The real-time detection of concealed threats is critical for protecting public transportation, sports arenas, and other open, difficult-to-secure environments. New sensing technologies can help security personnel screen customers and bags quickly without affecting the flow of traffic. However, accurately detecting threats in the complex environment of crowds carrying everyday items remains a challenge.
The HIVE algorithmic framework enables new approaches to protecting people and infrastructure in areas where traditional security checkpoints are not feasible. HIVE (Hierarchical Inference for Volumetric Estimation) is a custom deep convolutional neural network architecture that interprets volumetric video generated by a standoff, active-RF imagers. The architecture performs multi-resolution detection, classification, and segmentation of objects in the scene at various scales in order to produce automated threat detections, alerts, and visualization products without requiring the person to stop, pose, or remove their belongings.
|HIVE: A Novel Algorithmic Framework for Standoff Concealed Threat Detection