Publication

Abstract :

Oceans sensors play a vital role in measuring parameters such as salinity, turbidity, and marine pollution caused by plastics and petroleum. The data collected from these sensors enables us to mitigate the adverse climate impacts and protect the economics, tourism, and fishing industries that rely on the ocean. Traditionally, Autonomous Underwater Vehicles (AUVs) and Remote Operated Vehicles (ROVs) are deployed to record videos over extended periods in an effort to locate active and inactive sensor nodes or buoys in the ocean. The identification of these sensor nodes has become increasingly crucial in recent times. However, the post-processing of these vast datasets requires human intervention, leading to time-consuming analysis. In our work, we have aimed to streamline this process by employing an Edge AI device (NVIDIA Jetson AGX Xavier) to process recorded videos from a pool environment. We designed a square-shaped target customized to mimic ocean sensors or buoys and utilized the custom YOLOv5 model to detect the target frame by frame in a video. Furthermore, our research can be seamlessly integrated with advanced automation technology, incorporating robotic components and waterproof casings to enhance underwater object tracking applications.