摘要
针对传统算法在处理激光扫描点云与倾斜摄影点云融合过程中出现的计算性能低下、匹配精确度不足等问题,按照“化整为零、分而治之”的思想,设计了一种基于CUDA架构的多源点云融合处理算法。该算法利用CUDA架构特点,先对多源点云重叠区域数据进行空间划分,再将每个空间分割单元内的点云集输入GPU处理管道进行配准处理,最后对配准后的点云数据进行整合处理输出结果点云。实验结果表明:文中提出的算法与传统算法相比能够大幅提升计算性能,并且在配准精确度方面也取得了较好的效果,为智慧工地建设提供高精度三维建模支持。
Aiming at the problems of low computational performance and insufficient matching accuracy in the fusion process of laser scanning point clouds and oblique photography point clouds by traditional algorithms,this paper designs a multi-source point cloud fusion processing algorithm based on the CUDA architecture in accordance with the idea of“divide and conquer”.This algorithm takes advantage of the characteristics of the CUDA architecture,first spatially divides the overlapping area data of multi-source point clouds,then inputs the point cloud set in each spatial division unit into the GPU processing pipeline for registration processing,and finally integrates the registered point cloud data to output the result point cloud.Experimental results show that the algorithm proposed in this paper can significantly improve the computational performance compared with traditional algorithms,and also achieves good results in registration accuracy,providing high-precision 3D modeling support for smart construction site construction.
作者
杨蕊
杜广林
Yang Rui;Du Guanglin(Zhejiang Changzheng Vocational&Technical College,Hangzhou Zhejiang 310013,China;China Communications Information Technology Group Co.,Ltd.,Beijing 101399,China)
出处
《山西建筑》
2026年第2期170-173,185,共5页
Shanxi Architecture
基金
2024年浙江省教育厅一般科研项目(项目编号:Y202455942)
2024年浙江省高等教育学会研究课题(项目编号:KT2024282)。
关键词
多源点云融合
CUDA架构
空间划分
配准精确度
multi-source point cloud fusion
CUDA architecture
space division
registration accuracy
