Future military engagements are increasingly likely to occur in dense urban environments, accelerating the need for accurate representations of buildings, both exterior and interior. A variety of formats are required to support training, operational, and intelligence communities. Current data collection methods and automated data processes that generate 3D synthetic models are not mature enough to handle the complexities of dense urban environments.  

State-of-the-art drone technology enables the automated collection of ground-truth exteriors and solves some of the problems with building exterior generation, using photogrammetric techniques and laser scans from air or ground-based sensors. These techniques, however, do not adequately apply to interior collections.

Modeling a building interior presents a complex problem to the modeling and simulation community.  The traditional approach to modeling building interiors, by hand, using CAD or other 3D modeling tools is costly, both financially and in turnaround time. This especially true for multi-story buildings or buildings with unusual geometry. Without an interior paired with the exterior model, there are limitations to the situations and scenarios where a building model asset can be used.  Every collection method presents unique advantages and disadvantages impacting its value to the automated processes employed to create an interior model.

Our research suggests that a broad spectrum of techniques and algorithms be employed to create interior models, rather than to rely on a single automated collection and processing mechanism. In this paper, we discuss how a combination of commercial and open-source software, organized in a pipeline, incrementally improve, and optimize interior geometry. We also discuss lessons learned from a) data collection analysis; b) various data formats; and c) machine learning experiments that provide contextual clues. Lastly, we discuss how automation is leveraged to correct data collection and generation errors, and to derive complex metadata for interiors.  

Keywords
AUTOMATION, BEST PRACTICES, CONTENT GENERATION, M&S, SIMULATIONS, SYNTHETIC ENVIRONMENT, TERRAIN, URBAN ENVIRONMENT

Additional Keywords