摘要
论文提出基于无人机倾斜摄影的大规模城市建筑单体化的方法,使用先进的运动结构和多视点立体算法,首先从无人机获取的影像生成密集点云。基于建筑物网格的统计分析,将点云分为不同类别(即建筑物、地面、树木和其他)。使用马尔可夫随机场优化提取每个建筑物的屋顶结构;利用基于中心点检测的轮廓细化算法对面片轮廓进行细化;从细化后的轮廓中提取多边形网格模型。在各种场景上的实验以及与现有重建单体化方法的比较证明了该方法的有效性和可靠性。
With advanced kinematic structures and multi-view stereo algorithms,this paper proposes a method of reconstructing large-scale individual urban buildings based on aerial images.First dense point clouds are generated from aerial images.Based on statistical analysis of the building grid,point clouds are divided into different categories(i.e.buildings,ground,trees,and others).The roof structure of each building is extracted using Markov random field optimization method.The contour refinement algorithm based on center point detection is used to refine the face contour,from which the polygonal mesh model is extracted.Experiments on various scenarios and comparison with existing monomer reconstruction methods demonstrate the effectiveness and reliability ofthe proposed method.
作者
毛智俊
MAO Zhijun(Shanghai Municipal Institute of Surveying and Mapping,Shanghai 200063,China)
出处
《江西测绘》
2024年第2期15-18,22,共5页
JIANGXI CEHUI
关键词
建筑单体化
航空影像
点云
马尔可夫随机场
影像分割
Individual Building
Aerial Image
Point Cloud
Markov Random Field
Image Segmentation
