In-Chang Yeo, Myung-Il Roh, Hye-Won Lee, "An Optimization Method of the Surround-View Camera System for Automatic Berthing of Ships", Proceedings of G-NAOE(Global Conference on Naval Architecture and Ocean Engineering) 2022, Changwon, Korea, 2022.11.06-10
조회 수 276
|Generally, the port is very crowded with towing ships and ships that enter and depart. The crowded environment could cause difficulties in the berthing process which requires precise control. In addition, especially in the case of large ships, the visibility is very limited, and even a small collision of a ship can lead to a significant accident. Therefore, a surround-view camera system can be helpful to recognize the surrounding environment of a ship during the berthing process. A surround-view camera system refers to a system that produces a new viewpoint as if looking down around an object from the sky using images acquired by the multiple cameras attached around the object. Generally, fisheye cameras with a wide viewing angle are used for a surround-view camera system.
In the case of ships, the range of the surround-view camera system is much wider than that of cars, where the surround-view camera system is generally applied. Therefore, the number of cameras required to create a surround-view image is more. Furthermore, as the ships have various shapes and sizes, the design of the number, location, and installation angle of the cameras are very complicated that depend on the experience of the expert.
Therefore, this study proposes an optimization method of the number, location, and installation angle of fisheye cameras to improve the quality of surround-view image with the minimum number of fisheye cameras. The number, position, and installation angle of fisheye cameras were set as design variables to formulate an optimization problem for this purpose. Furthermore, a new evaluation method was introduced for the quantitative evaluation of various surround-view images created by the surround-view camera system. This evaluation method compares the created surround-view image with the actual bird’s eye view image. The objective function was to minimize the difference between the surround-view image and the actual bird’s eye view image using the minimum number of fisheye cameras. The virtual environment, including the ship, fisheye cameras, and the area around the port, was made to create fisheye images of various situations for optimization. The virtual environment was implemented in the context of the design variables using Unity 3D. We developed an integrated system that automatically creates a surround-view image based on the fisheye images from the virtual environment and evaluates the quality of the surround-view image. We applied the proposed method to various examples. As a result, it was confirmed that the proposed method could be used for establishing the optimal surround-view camera system.