摘要：AbstractMost data collected with autonomous underwater vehicles (AUVs) can only be reviewed and analyzed post-mission. This limitation is due to the low-bandwidth acoustic links used for communication between AUV and surface, which are typically devoted to monitoring a few mission-critical parameters. In this work, we propose a plug-and-play solution that would enable live streaming of side-scan sonar data on such low-bandwidth modems with no additional hardware requirements. Using principles from image super resolution, we first downsample and compress sonar images on the sensor(AUV)-side and then estimate the full high-quality sonar image on the receiver(surface)-side. The bandwidth requirements for this livestreaming are below 1.87 kbit/s while still providing reconstructions difficult to distinguish from the originals. Our complete end-to-end system has been implemented using EvoLogics underwater modems and successfully tested in our laboratory and harbor basin.
关键词：KeywordsDeep LearningSuper-ResolutionAcoustic Data TransmissionSide-Scan Sonar ImagingAutonomous Underwater Vehicle (AUV)
