Method for 3D gaussian splashing fast reconstruction for power transmission tower

The 3D Gaussian splashing rapid reconstruction method solves the problems of insufficient integrity and high complexity in the reconstruction of power transmission towers, and realizes efficient and low-cost 3D reconstruction, which is suitable for rapid modeling and real-time application of power grid equipment.

CN122244309APending Publication Date: 2026-06-19HEBEI UNIV OF ENG
View PDF 0 Cites 0 Cited by

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
HEBEI UNIV OF ENG
Filing Date
2026-03-16
Publication Date
2026-06-19

AI Technical Summary

Technical Problem

Existing technologies for rebuilding transmission towers suffer from problems such as insufficient reconstruction integrity, easy generation of voids on weakly textured metal surfaces, and high complexity in data acquisition and post-processing, failing to meet the rapid modeling needs of power grid emergency repairs or frequent inspections.

Method used

A fast 3D Gaussian splash reconstruction method is adopted, including video data acquisition, feature extraction and sparse reconstruction, dataset construction, model training and point cloud extraction. It combines the COLMAP database and differentiable rasterization technology, block parallel processing and Gaussian pruning to simplify the point cloud post-processing process.

Benefits of technology

It achieves high-fidelity and rapid 3D reconstruction of power transmission towers, reduces hardware and computing costs, improves reconstruction accuracy and efficiency, is suitable for real-time deployment at the edge, and provides lightweight obstacle models.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122244309A_ABST
    Figure CN122244309A_ABST
Patent Text Reader

Abstract

This invention discloses a rapid 3D Gaussian splash reconstruction method for power transmission towers, belonging to the field of computer vision and 3D reconstruction technology. It comprises six sequential steps: video data acquisition and preprocessing, feature extraction and sparse reconstruction, Gaussian splash dataset construction, Gaussian splash model training, model merging and point cloud extraction, and point cloud post-processing. This invention achieves four major innovative advantages: innovation at the acquisition layer, reconstruction layer, optimization layer, and application layer. This invention is applicable to the rapid reconstruction of 3D models of power transmission towers.
Need to check novelty before this filing date? Find Prior Art