The flight simulation industry desires less expensive and more realistic content for training pilots. Advances in machine learning techniques have brought us closer to creating digital twins of the whole Earth by extracting 3D features from carefully selected photographic imagery. Unfortunately, these databases are too big to distribute, are limited to the texel resolution of the source imagery, and require lots of effort to process and update. Instead, using whole Earth geo-specific metadata, Machine Learning techniques can be used to generate photo-realistic imagery correlated with 3D typical content created at run-time for the whole Earth at any texel resolution. Database modelers’ scope of work is greatly reduced whereby they only provide key landmark 3D models where required. Updating Fraternal Twin metadata is much easier than retrieving and processing new satellite imagery.

This paper explains what a Fraternal Twin is and how it is better suited to serve the simulation industry than identical Digital Twins for whole Earth simulation. This paper examines the machine learning processes and meta-data needed to create a photo-realistic Fraternal Twin of the whole Earth that pilots can use for training more cost-effectively. The result is a super high-resolution photo-realistic geo-typical seasonally accurate representation of the whole Earth that can be easily distributed and give pilots-in-training  realistic visual content that they can train with.

Keywords
EMERGING TECHNOLOGIES;ENVIRONMENTS;FLIGHT TRAINING;GEOSPATIAL DATA ;MACHINE LEARNING;MODELING;REAL-TIME;SYNTHETIC ENVIRONMENT;TERRAIN;URBAN ENVIRONMENT;VECTOR DATA

Additional Keywords