Artificial Vision System for Object Classification Using AI for Autonomous Underwater Vehicles

This article addresses the resolution of a specific problem in the field of underwater robotics. The focus is on improving the perception and analytical capabilities of an autonomous underwater vehicle. Thus, the overall objective is to advance the creation and development of a prototype artificial vision system for object detection and classification using stereo cameras in an underwater environment and compatible with ROS. To achieve this, a suitable container has been designed and developed for the cameras and data capture hardware. Additionally, a stereo vision system has been implemented to detect and classify objects more accurately and in more detail in underwater environments. Finally, its operation is validated under real conditions through image acquisition and subsequent analysis.