多源融合的实景三维一体化建模研究
Research on Reality-based 3D Integrated Modeling with Multi-Source Integration
刘萍 1高晓兵 2苏胜 1陈晨 1宋志强3
作者信息
- 1. 西安工业大学设计学院,西安 710021
- 2. 西安工业大学机电工程学院,西安 710021
- 3. 西安翻译学院艺术与设计学院,西安 710105
- 折叠
摘要
针对传统单一数据源难以满足高精度建模需求的问题,本研究提出一种基于多源融合的实景三维一体化建模方法,旨在通过整合无人机航拍、倾斜摄影和地面RTK相机等数据源,实现高精度的实景三维模型构建.通过空中三角测量优化相机参数,利用KAZE算子进行特征点检测与匹配,实现多视角立体重建和泊松表面重建,生成高精度三维模型.结合GPU/CPU加速和3D Tiles技术优化纹理映射.在西安工业大学的测试显示,模型平面坐标误差0.09 m,高程坐标误差0.367 9 m,高度误差0.105 6 m,长度误差0.05 m,位置不确定度0.003～0.051 9 m,验证了方法的高精度和可靠性,为城市数字化研究提供参考.
Abstract
Traditional single data sources cannot meet the demands of high-precision modeling.To solve this problem,this paper proposes an integrated three-dimensional modeling method based on multi-source fusion for high-precision three-dimensional actual scene reconstruction by integrating data sources such as drone aerial photography,oblique photography,and ground-based RTK cameras.Multi-perspective stereo reconstruction and Poisson surface reconstruction are realized by optimizing the camera parameters through aerial triangulation,and detecting and matching feature points by the KAZE algorithm,and a high-precision 3D model is generated.The texture mapping is optimized by combining GPU/CPU acceleration and 3D Tiles technology.A test conducted at Xi'an Technological University shows that the model has a planar coordinate error of 0.09 m,a vertical coordinate error of 0.367 9 m,a height error of 0.105 6 m,a length error of 0.05 m,and a positional uncertainty ranging from 0.003 m to 0.051 9 m,which verifies the high precision and reliability of the method.This study provides robust geographic information support for urban digital operation and planning.
关键词
多源融合/实景三维/一体化建模/倾斜摄影Key words
multi-source integration/reality-based 3D/integrated modeling/oblique photography引用本文复制引用
出版年
2024
NETL
NSTL
万方数据