Scholarlink

The ongoing Industry 4.0 revolution is reshaping the construction sector, emphasising the growing importance of digital technologies for knowledge management. This transition is supported by Agricultural Knowledge and Innovation Systems), promoting sustainable development and environmental protection. Through 3D modelling, geomatics enhances spatial data analysis with improved speed and precision. However, the high cost of terrestrial laser scanners necessitates more economical alternatives. This study contributes to defining an acquisition process across different platforms by presenting an innovative approach for 3D documentation of a barn and manure pit. The method combines images from a DJI Mavic Enterprise 3 drone with a GNSS-RTK antenna. Additionally, the study compares the accuracy and final product of Neural Radiance Fields (NeRF) and Gaussian Splatting (GS) methods, using drone acquisition as the validation reference. The results show that NeRF and GS are similar in accuracy and produce a realistic final product but limited to the barn and manure pit retaining wall coverage. The volume calculation depends on the point density generated on the ground based on the number of shots and different angles. This paper has implications for the inventory of rural buildings and associated structures, offering an effective method for generating 3D models in the agricultural field. This technology can quickly document and evaluate the available space within agricultural structures, providing new insights into possible alternative uses and efficiencies.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Assessment of 3D Models of Rural Buildings Using UAV Images: A Comparison of NeRF, GS and MVS-SFM Methods

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Stefano Chiappini,
  • Ernesto Marcheggiani,
  • Roberto Pierdicca,
  • Md Abdul Choudhury

摘要

The ongoing Industry 4.0 revolution is reshaping the construction sector, emphasising the growing importance of digital technologies for knowledge management. This transition is supported by Agricultural Knowledge and Innovation Systems), promoting sustainable development and environmental protection. Through 3D modelling, geomatics enhances spatial data analysis with improved speed and precision. However, the high cost of terrestrial laser scanners necessitates more economical alternatives. This study contributes to defining an acquisition process across different platforms by presenting an innovative approach for 3D documentation of a barn and manure pit. The method combines images from a DJI Mavic Enterprise 3 drone with a GNSS-RTK antenna. Additionally, the study compares the accuracy and final product of Neural Radiance Fields (NeRF) and Gaussian Splatting (GS) methods, using drone acquisition as the validation reference. The results show that NeRF and GS are similar in accuracy and produce a realistic final product but limited to the barn and manure pit retaining wall coverage. The volume calculation depends on the point density generated on the ground based on the number of shots and different angles. This paper has implications for the inventory of rural buildings and associated structures, offering an effective method for generating 3D models in the agricultural field. This technology can quickly document and evaluate the available space within agricultural structures, providing new insights into possible alternative uses and efficiencies.