Korean Journal of Computational Design and Engineering

The shipbuilding and marine industry have recently been working to achieve digital transformation(DX) by utilizing advanced technologies such as augmented reality(AR), virtual reality (VR), and AI in ship construction and operation. In the ship production design phase, when augmented reality is used for design and interference reviews, it enables quick identification of design errors or equipment that causes interference through 3D models. Additionally, it allows for the visualization of work information for components that need to be installed, which can enhance work efficiency. However, in augmented reality, virtual objects do not sufficiently interact with real objects, as the entire model, including parts that should be occluded by real objects, is augmented. It fails to provide sufficient information about virtual objects and acts as a factor that reduces the user's sense of immersion and realism. This study examined image processing solutions using generative adversarial networks(GAN) and OpenCV for occlusion handling, where real and virtual objects overlap. Through a comparison of the results from the testing approaches, the study presented a methodology suitable for the current shipbuilding and marine industry and suggested future development directions.

키워드 Augmented Reality, Generative Adversarial Networks, Occlusion