Performance Tradeoff Considerations in a Graphics Processing Unit (GPU) Implementation of a Low-Detectable Aircraft Sensor System

摘要

The United States Naval Research Laboratory (NRL) is developing a Large Area Scanning and Surveillance Optical System (LASSOS) for identifying and tracking low detectable manned and unmanned aircraft. The system employs altitude-azimuth swept optical sensors to scan the surrounding airspace and give timely warning of pre-attack targeting operations. Due to their size and standoff distances, the smallest of these aircraft present very small sensor footprints, requiring high-resolution, high-data scans which must be processed in real time. Given packaging size and weight constraints and given the image feature-extraction nature of the sensor data processing problem, NRL is investigating Graphics Processing Unit (GPU) technology for the high-computational-load front end of the processing chain.