Using Gpu For High Fidelity Low-Cost Sensor Simulation

Today\u27s training system requires Forward-Looking Infra-red (FLIR), Night Vision Goggles (NVG) and special sensor effects, including gain, blur, noise, haloing, atmospherics, plumes and time of day. Traditionally, sensor simulation applications either require high performance graphics workstation or special hardware in order to have high fidelity and interactive frame-rate. Such requirements significantly raise the cost for the training system. Recent appearance of graphics processing units (GPUs) provides a new implementation method for different kinds of sensor simulation. This paper discusses framework design for sensor simulation applications leveraging the highly programmability of the GPUs\u27 streamline. By making small changes to the High-Level Shader Language (HLSL) without change to the rendering engine, users can greatly customize their own visual effect thus providing adaptability and ease of system integration to the training system