Geospecific Night Imagery for Real-time Training Simulators

The current modeling of night-time visual and sensor scenes for real-time training simulators is expensive, inconsistent, and lacking in realism. During the SBIR Phase I effort we have developed innovative geospatial data, tools, processes, and simulation techniques to automatically derive geospecific cultural lighting models and real-time representations. Our technical approach is driven by the assumption that today’s remotely sensed night imagery lacks the dynamic range, intensity precision, spatial resolution, and broad availability required to support advanced night simulations, particularly in the sensor domain. Instead, the approach focuses on using night imagery, along with many other data types, to discern and/or predict the presence of specific types of cultural lighting at specific locations. Several core concepts were demonstrated and will be continued in Phase II: the extraction of cultural lighting types and positions from geospecific night imagery; the prediction of cultural lighting types and positions from commonly available geospatial data other than night imagery; the fusion of the predicted cultural lighting types and positions; the real-time rendering of high density, high dynamic range cultural lighting representations.