An effective per-processing pipeline of NeRF 3D reconstruction for benthic target

The nearshore oceans, home to rich benthic ecosystems, remain an area of significant research interest. While 2D visual representations have been the mainstay in this area, the intricate, multi-dimensional nature of the seafloor ecosystems underscores the need for 3D modeling to capture their full essence. This research introduces a methodology tailored for static image processing and 3D modeling using Neural Radiance Fields (NeRF), specifically the optimized instantNGP variant. A streamlined pipeline has been developed, focusing on mitigating the visual challenges posed by light interference and seawater in underwater imaging. This pre-processing approach effectively prepares images for the NeRF-based 3D reconstruction without an excessive computational burden. Visual enhancements successfully corrected color imbalances in underwater images, addressing the common blue-green tint caused by light conditions. Furthermore, by dynamically detecting and eliminating seawater borders, the pipeline ensures that the reconstruction models remain concentrated on the seafloor ecosystem. This process does not necessitate extensive datasets or immense computational resources, marking it as an efficient solution for near coast underwater images. Offering a cost-effective and efficient alternative to traditional methods, this research provides marine ecologists with a robust tool for RGB-based 3D modeling of nearshore environments. However, its application might benefit from integration with neural networks for better adaptability across various marine scenarios.