증강현실 모델 위치 정합 활용을 위한 기계학습 기반의 선박 블록 윤곽선 검출 연구

초록

As interest in autonomous ships is growing, research on augmented reality-based remote sup- port systems that can be used onboard ships is being conducted. The markerless method is more suitable than the marker method because corrosion and damage can easily occur in the ship. However, in the case of the markerless method, a process of selecting a desired feature point is required, but it is not easy in a complicated ship. Therefore, in this study, the outline detection system was studied to utilize the outline as a feature point for augmenting the augmented real- ity model. Although many existing studies have preceded it, it is not suitable for detecting the contour of a ship block with many internal and external components. Therefore, in order to overcome the limitations of the previous method, in this study, to detect only the outline of a ship block, a model was built through CNN, GAN, and Segmentation algorithm, which are deep learning techniques widely used in image processing, and block outline detection was per- formed. It is also expected that these results will help automate and optimize the stockyard management system.

키워드