In training vision-based autonomous system algorithms, simulation plays a pivotal role in enhancing and expediting system recognition and development. However, creating a realistic virtual underwater environment is particularly difficult. To address the need for advanced control system modeling, we developed an integrated simulation framework that combines a 3D computer graphics engine, Unreal Engine, with Simulink. This integration facilitates the design and visualization of complex control systems. The framework was evaluated using an autonomous system tasked with tracking underwater cables, employing YOLO-based object detection for visual guidance. The results demonstrate the effectiveness of the proposed simulation environment in accurately replicating the behavior of vision-based autonomous systems operating in underwater conditions.