基于SFM与高斯渲染优化的建筑物三维重建方法  

作　　者：周健 罗玉恒 

机构地区：[1]甘肃林业职业技术大学,甘肃天水741020 [2]甘肃农业大学,兰州730070

出　　处：《科技创新与应用》2025年第11期152-155,共4页Technology Innovation and Application

基　　金：2025年度甘肃省级科技计划自然科学基金(25JRRE030);2023年甘肃省科技厅自然科学基金(23JRRE0731);2023年甘肃省高校教师创新基金(2023A248);2022年度甘肃林业职业技术学院院级基金(GSLY2022-A04)。

摘　　要：为满足高精度建模任务对三维重建技术的需求,提出一种改进的三维重建方法,优化相机姿态估计、点云处理及网格建模与纹理映射等关键环节。点云生成阶段引入基于连续高斯分布的渲染模型,实现三维点的权重优化与密度控制,在边缘与纹理弱区域具备良好重建效果。网格建模阶段结合法向量引导的全局重建算法与多视角融合纹理映射策略,提升模型几何连贯性与视觉一致性。实验结果表明,该方法在多个复杂场景下均优于传统方法,适用于建筑环境下的高质量三维建模任务。To address the increasing demand for high-precision 3D modeling,this study proposes an enhanced reconstruction framework.The method focuses on optimizing camera pose estimation,point cloud processing,mesh generation,and texture mapping.During point cloud generation,a rendering model based on continuous Gaussian distribution is introduced to adjust point-wise weights and control spatial density.This approach improves reconstruction performance,particularly in edge regions and low-texture areas.During mesh modeling,a global reconstruction algorithm using surface orientation is combined with a multi-view texture fusion method.This integration enhances both the geometric continuity and visual coherence of the 3D models.Experimental results demonstrate that the proposed method achieves higher accuracy and visual quality than conventional techniques in complex scenarios.It is especially suitable for high-quality 3D modeling in architectural environments.

关 键 词：三维重建 高斯渲染 点云优化 泊松表面重建 纹理融合 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]