AI-based Real-time Automatic 3D Reconstruction and Modeling from Multiple Images for SA, Transport, and Dispersion Modeling

CBD seeks a solution that can improve the speed and automation of the processing workflow to generate inputs for use with CONTAM and other aligned tools such as extended reality (XR) platforms. The analytical focus is on assessing the transport and dispersion (T&D) of airborne contaminants and associating these with other risks and predictions for what is generally termed “Multizone Airflow and Contaminant Transport” (MACT) analysis. This requires the proper building configuration inputs (e.g., 3D building content, interior layout, etc.) to generate these airflow models and analytics. CBD desires an improved solution for 3D model generation (e.g., buildings, facilities, interiors, exteriors, etc.), implementing advanced methods such as computer vision (CV) and photogrammetry, data fusion, Artificial Intelligence (AI), and Machine Learning (ML). The resource data of interest for this include, but are not limited to, 2D satellite electro-optical (EO) imagery, aerial imagery (e.g., from manned or unmanned systems), photos, blueprints, and more. Tangentially, significant advances in AI, ML, CV, and data fusion have also proven successful in the extraction of Full Motion Video (FMV) frames and content, hand-held image content, and utilization of other non-traditional resources (e.g., RADAR, x-ray, and thermal imaging [“TIR”]), that supplement and enhance 3D content. IAVO has worked in the 3D modeling/photogrammetry disciplines for more than two decades, and we have contributed to the steady pace of research, development, and engineering for these technologies. Twenty short years ago, the idea of creating a solution as sought by CBD would have been a technology stretch, but today, tools and methods have matured to levels whereby the envisioned 005 objectives are readily achievable. Indeed, the emphasis is more on processing speed, workflow efficiencies, and secondary enhancements instead of the core modeling per se. Aligning with the thoughts above, we propose to expand our ongoing work and experience in 2D/3D modeling-processing domains and satisfy the SBIR objectives per the following tasks: - Modernize our 3D modeling methods by leveraging specific AI/ML - Establish interior model hypothesis from the external data to include leveraging, when and if available, any/all of the following: (1) interior imagery/data; (2) additional input sources; (3) blueprints; and (4) exterior imagery/data - Fuse interior and exterior building data (i.e., adhere to the “OpenBuilding” modeling paradigms) - Utilize AI/ML to identify building elements - Define model delivery methods to support CONTAM import and other simulation engines We believe these SBIR work plan themes fit well with our proven expertise and breadth of ongoing and overlapping DoD initiatives and will result in a successful SBIR project. We expand upon these further in presenting our project objectives, goals, and tasking.