基于四目立体视觉的三维全自动建模
Automatic three-dimensional modeling based on four-view stereo vision
黄荣娟 1仲思东 2屠礼芬1
作者信息
- 1. 武汉大学 电子信息学院，湖北 武汉430072
- 2. 武汉大学 电子信息学院，湖北 武汉430072; 武汉大学 测绘遥感信息工程国家重点实验室，湖北 武汉430079
- 折叠
摘要
为保护并传承敦煌莫高窟世界文化遗产，研究基于四目立体视觉系统的三维全自动建模技术。由于提取的点云数量较大，导致自动建模过程中，点云的三角构网步骤无法顺利进行，因此着重研究点云的精简算法。将包围盒算法与点云的曲率特征相结合，提出一种点云精简算法，即改进的包围盒法。实验结果表明，该方法对点云的采样精度高于原始包围盒法，点云精简率可达93．6％，点云数据精简后三维建模效率大大提高，生成的三维模型精度较高。
Abstract
Fully automatic 3D modeling technology based on four-view stereo vision system was studied for the protection and in-heritance of Dunhuang Mogao Grottoes,the world’s cultural heritage.Due to the large number of feature points,point cloud triangulation network configuration steps could not proceed smoothly,so some point cloud reduction algorithms were studied on purpose.The curvature characteristic of point cloud was combined with the bounding box methods,a point cloud reduction algo-rithm was proposed,which is the improved bounding box method.Experimental results show that sampling precision of the im-proved bounding box method to point cloud is higher than the original method,the point cloud reduction rate reaches 93. 6%. After point cloud reduction,the efficiency of three-dimensional model is greatly improved,and the precision of the produced three-dimensional model is high.
关键词
莫高窟/立体视觉/三维建模/点云精简/包围盒法Key words
Mogao Grottoes/stereo vision/three-dimensional modeling/point cloud reduction/bounding box method引用本文复制引用
基金项目
国家973重点基础研究发展计划基金(2012CB725301)
国家测绘地理信息公益性行业科研专项基金(201412015)
出版年
2015
NETL
NSTL
维普
万方数据