Enhancing ROV/AUV Imagery Using Real-Time Underwater Optical Image Processing

J. Wade
ZMicro, Inc., California, United States

Keywords: AI, MCM, ROV, UUV, video

Undersea visibility and the acquisition of clear and detailed underwater imagery has always been a challenge for defense applications such as mine countermeasures, hull and equipment inspections, diver safety, surveillance, and search and rescue. Autonomous underwater vehicles (AUVs) and remotely operated underwater vehicles (ROV) play a rapidly expanding role in underwater missions as they become ever more capable and autonomous. However, poor visibility remains a limiting factor in exploiting their full potential. This paper introduces a new computing platform that dramatically improves underwater image clarity using real-time optical image enhancement algorithms. The platform also integrates deep learning capabilities that can enable more sophisticated autonomous behavior, including Anti-Submarine Warfare (ASW) and ISR. Perhaps one of the easiest and most straightforward methods to improve underwater image quality is to apply image processing algorithms that, in essence, undo some of the undesirable effects of the physics of underwater vision. These algorithms run on our JetPack, which is a compact, rugged stand-alone system that can be plugged into an existing ROV/AUV system architecture to enable image enhancement and deep learning capabilities. The JetPack includes the NVIDIA Jetson TX2 System-on-Module (SOM), which is a fast, power-efficient embedded AI computing device.