SIGNATURE WORK
CONFERENCE & EXHIBITION 2024

This paper presents a novel approach to 3D reconstruction of tunnel environments using Neural Radiance Fields (NeRF), addressing the challenge of GPS navigation in areas where signals are unreliable, such as tunnels. It emphasizes the potential of NeRF in creating detailed, three-dimensional models from a collection of 2D images, facilitating better navigation and safety measures within tunnels. By comparing different NeRF models and incorporating convolutional neural networks, the study demonstrates improvements in reconstruction accuracy and efficiency. Despite the reliance on external datasets and the limitations associated with dynamic environments, the research highlights significant advancements in 3D modeling technologies, suggesting a promising future for real-world applications in tunnel navigation and beyond.