MANIFOLD ANALYTICS, INC. SBIR Phase II Award, February 2021

Manifold Analytics and business partner, Lockheed Martin, are proposing an automated microservice pipeline to automate the process of data discovery, source preparation, and conversion of geospatial data into CDB layers for USSOCOM. This toolset is designed to integrate into directly into the existing microservice platform used by the Common Operating Picture (COP) and will leverage two decades of Manifold’s geospatial analysis and machine vision experience and Lockheed Martin’s geospatial processing engine, Rosetta (a commercially fielded, TRL-9 product). Manifold will provide a series of microservices, coined: the Microservice CDB Automated Production Pipeline (MCAPP) to automate conversion of satellite imagery, tactical imagery, elevation models, point clouds, vector data, material ID layers, and existing 3D models to CDB. MCAPP will also provide “upstream” workflows to prepare source data for CDB production by classifying and registering inputs, creating a bare-earth DTM layer, and automating the extraction of 3D building models. Georegistration and rectification approaches are provided to tightly align each raster, mesh, point cloud, and vector data source with other data sources as they become available. Manifold has provided several DTM generation approaches for the COP as part of existing contracts with Lockheed Martin, but MCAPP will enhance these techniques to leverage data from other sources to improve and guide the generation of bare-earth elevation. Manifold will also provide a parametric 3D building extraction method to automate the conversion of simple building structures to 3D model entities that can be exported to CDB. Material classification is the foundation for each of these approaches. MCAPP leverages a heterogeneous neural network analysis of spectral bands, spatial texture data, elevation profiles, and vector data to determine material IDs for each pixel of CDB data. These material IDs provide cues for the enhanced DTM generation and building extraction workflows, as well as a novel georegistration approach that uses material classification results to “bridge” registration between difficult sources (such as elevation and spectral data). Each of these workflows is designed to run atomically and asynchronously to allow portions of the CDB generation pipeline to occur in parts as data becomes available. Manifold will supply a web client interface for generating production workflows that will allow users to visually build processing jobs as node-based graphs so that they can be designed/tailored to fit the COP environment as it evolves.