We present a novel real-time system to render realistic visible and infrared (IR) surveillance and reconnaissance training scenarios. This simulation accurately generates dynamic temporal energy maps during runtime to produce imagery for a variety of sensors and effects to create a realistic visualization experience for training. The temporal energy maps are mappings that represent the amount of spectral energy accrued over time for a given pixel on any surface form. Our system utilizes automatic material classification and a spectral material database to facilitate realistic synthetic environments under different sensor spectral band requirements. This work describes our approach to generate convincing material mapped terrain and objects, and we present runtime algorithms that store temporal spectral energy in a physically-based approach to accurately simulate temporal thermal behavior throughout the day.  

This approach is capable of accurately rendering large high-resolution geospatial imagery. We describe our work to demonstrate the feasibility of creating real-time temporal radiative heat transfer in a commodity game engine that provides essential detail to phenomenological sensor signatures and effects. The resulting prototype delivers robust training capabilities with a level of detail that allows for better training and mission planning than traditional techniques. Additionally, synthetic sensor data can be generated from this approach and can be used to train machine learning algorithms that require large datasets where imagery is difficult to acquire.