An Artificial Intelligence (AI) ¢ Based 3 Dimension (3D) Building Model Generation Capability for Chemical Biological Radiological and Nuclear (CBRN) Situational Awareness (SA) and Hazard Modeling

Accurate visual representations of environments subjected to chemical and biological (CB) hazards represent consequential information to the Warfighter, allowing them to gain critical situational awareness and training before an operation or response. Furthermore, it is vital that the simulated dispersion of the CB threat must be physically consistent with the environment in which it occurs. While the technologies necessary to create realistic three-dimensional (3D) models and accurate simulations of the CB hazards independently exist, a gap exists in our ability to link these technologies, resulting in 3D models and CB simulations that are often created using different data and assumptions, leading to physically inconsistent domain representations. This issue is amplified in scenarios that occur in urban environments and inside buildings, where the details required to replicate the area of operation (AO) accurately may be absent. In the Small Business Innovative Research (SBIR) Phase II effort proposed here, we harden, refine, and expand the capabilities of the SCRIBE (Structurally Consistent Rapid Building Estimation) system. This initiative aims to build upon the success of the SCRIBE Phase I proof-of-concept, which demonstrated the capability to ingest and process multi-modal building imagery input data, infer missing characteristics of the structure, and create physically consistent 3D models and chemical and biological transport and dispersion (CB T&D) simulations of the building. We will advance the technology readiness level of SCRIBE by using state-of-the-art machine learning (ML)/artificial intelligence (AI) and bespoke algorithms to process, fuse, infer missing building features, and build physically consistent 3D models and CB T&D simulation inputs using a range of multi-model data inputs such as satellite imagery, LIDAR, infrared imagery, CAD files, blueprint, and open-source maps. We will design SCRIBE so that scientific and industry standards support the models used to infer architectural details in the 3D models and airflow characteristics in the CB T&D simulations. Through this SBIR Phase II effort, we will conduct rigorous validation and testing of our approaches and develop a comprehensive and intuitive user interface (UI). The system will enable operators to seamlessly input a diverse selection of data types into the system, automatically producing realistic and physically consistent modeling and simulation products for use in extended reality (XR) training applications. By the end of the Phase II effort, we will present a refined prototype that promises a realistic, physically consistent, and fully automated system for automatic building reconstruction, 3D model creation, and CB T&D simulation input generation. The SCRIBE system will advance the methods in which CB operations are modeled, troops are trained, and strategic decisions are made.