This article explores visual data processing methods for underwater navigation and environmental reconstruction, based on modern approaches in computer vision and robotics. A system implemented in the ROS (Robot Operating System) environment is proposed, enabling simultaneous localization and mapping (SLAM) in underwater conditions. The system evaluation methodology includes virtual experiments using the Gazebo simulator, which replicate realistic underwater operational scenarios. The study demonstrates the feasibility of integrating stereo cameras, confirms the effectiveness of image processing methods in underwater environments, and presents quantitative metrics for navigation accuracy and object reconstruction. The results validate the proposed solutions as promising for real-world applications in autonomous underwater exploration.

Keywords: underwater navigation, environmental reconstruction, computer vision, visual odometry, SLAM, ROS, Gazebo, simulation, visual data, robotics