GE Research is Principal Investigator of an IARPA-sponsored research activity entitled Creation of Operationally Realistic 3D Environment (CORE3D), collaborating with several world-class universities toward a common goal.
The goal of the CORE3D program is to develop technology that automatically generates accurate 3D object models with real physical properties from multiple data sources. Data sources include commercial satellite panchromatic and multi-spectral imagery for global coverage, as well as airborne imagery and Geographic Information System (GIS) vector data, where available, for improved resolution and fidelity.
Advanced image processing algorithms are used to construct 3D surface models from 2D satellite and aerial imagery. From this starting point, AI is deployed to automate the detection of man-made objects and produce contextually consistent 3D models. Specifically, multi-spectral data analysis and scene context can enable identification of materials found in the scene.
GE Research brings many decades of multi-modal information fusion, remote sensing, image segmentation, classification and 3D image reconstruction expertise to the table, backed by decades of experience driving new technology development for our business units in the aerospace, medical and security domains.
The CORE3D program supports the need for timely access to geo-spatially accurate 3D object data on a global scale, to support critical missions around the world. Although accurate and reliable manual processes exist to capture the 3D geometry and surface properties of objects on the Earth, they are time-consuming and require human expertise.
Automated methods currently lag behind manual approaches in creating 3D models that contain mission-relevant information and understanding of scenes. The vision for the CORE3D program is to adopt AI as a means to automate and accelerate the 3D data generation process, enabling rapid operational planning anywhere in the world, even locations where up-to-date models may not exist.
More broadly, the techniques being developed under this program can be applied to a variety of inspection technologies, allowing AI to reduce manual intervention by either automatically complete the inspection and/or involve humans only in the more complicated decisions.
Phase 1 of the project kicked-off with 4 teams. There was an evaluation after 15 months, and GE was one of two teams down-selected by DARPA for Phase 1B.
Phase 1 Review
GE wins contract for Phase 1B