Large-Scale Mussel Farm Reconstruction with GPS Auxiliary

The expanding mussel farming industry, especially in regions like New Zealand (NZ), necessitates advanced monitoring and management solutions for sustainable and efficient operations. Currently, manual labor and on-site operations limit real-time capabilities and overall efficiency. To address these challenges, large-scale 3D reconstruction offers a viable solution, enabling reproducible representations of mussel farm scenes based on captured video for off-site evaluation and precise decision-making. Conventional 3D reconstruction methods often require intricate manual camera calibration, posing challenges for real-world applications, particularly in marine scenarios. In contrast, this paper presents an innovative and practical approach for large-scale mussel farm reconstruction by integrating GPS auxiliary data without heavy camera calibration. The primary objective is to develop a comprehensive large-scale 3D visualization of mussel farms using video footage and simultaneous GPS information captured by a vessel-mounted camera. Our methodology involves a two-step process: First, making adaptive image sampling from captured video and applying Structure from Motion (SfM) for camera pose estimation and a dense 3D reconstruction, accurately reproducing local 3D scenes of partial mussel farms. Then, GPS data will be integrated for georeferencing and localization. This alignment empowers farm managers with geospatial accuracy for decision-making, resource allocation, and strategic planning. Experimental results validate the effectiveness of the proposed approach on reconstruction accuracy and completeness, and GPS geo-registration precision, highlighting its potential for advancing mussel farm monitoring.