Real-Time Interferometric SAS Processing with Ultra-Low Power Consumption

摘要

In order to fully realize the potential of unmanned systems, vehicles must carry sensors to observe the underwater environment, and sensor data must be processed in real-time for systems to intelligently interpret the environment and adapt their behavior accordingly. Interferometric synthetic aperture sonar (InSAS) is an ideal tool for seabed survey operations because of its superior area coverage rate and resolution compared to conventional sidescan and multibeam sensors. Due to the intensive nature of forming two SAS images per side of the vehicle plus bathymetric processing, InSAS data have traditionally been processed offline in a non-real-time environment after completion of the mission. Kraken Robotic Systems have been performing real-time InSAS processing of seabed imagery and bathymetry for the past five years using a variety of graphics processing units (GPUs). Here, we demonstrate that recent advances in the computational efficiency of InSAS algorithms and the development of ultra-low power GPUs for embedded systems have combined to create the situation where it is now possible to perform the complete InSAS processing chain at full resolution in real-time using extremely low energy consumption.