Dipartimento di Ingegneria informatica, automatica e gestionale

In the last two decades, visual data acquisition in underwater environments has dramatically increased due to the need to face a wide range of challenges that still require further research, including site monitoring, seabed anomaly detection, object detection and classification, and many others. Most of these activities require frequent data acquisition and processing over time, even at different altitudes, view angles, and perspectives. Recent improvements of small-scale Autonomous Underwater Vehicles (AUVs), in terms of navigation time, automatic control, and onboard processing, are making these submersible vehicles particularly suitable for activities as those reported above. Moreover, thanks to their cableless navigation, limited size, and agility, small-scale AUVs (hereinafter simply AUVs) can reach sites otherwise inaccessible with other kinds of underwater vehicles (e.g., medium and large AUVs). The payload capacity of current AUVs allows us to equip them with different vision sensors, including Red Green Blue (RGB) camera and Side Scan Sonar (SSS). In this context, an open issue remains the efficient transmission of visual data from AUV through an underwater acoustic network to allow a remote workstation an online and/or real-time data processing. In this paper, a data compression module for the SUNSET platform is presented. The module is composed of a set of novel algorithms that enables compression of RGB and SSS information with and without data loss. The module also implements some novel features, including progressive compression and Region Of Interest (ROI) selection; the first used to gradually transmit the image data (e.g., sites in which the acoustic transmission is a hard task), the second used to transmit, with higher quality than the rest of the image, the items contained in a specific area. Exhaustive experiments on RGB and SSS datasets prove the effectiveness of the presented module.