关键词: 表面重建, KD树, 曲面变分, 特征曲率

Abstract:

With the improvement of the technology of data acquirement,the cloudpoint data
is used more and more widely in 3D reconstruction. The huge data size becomes the
bottleneck of reconstruction efficiency. The feature of models is blurred because of the
calculation accuracy of the curvature used in the existing simplification methods. A
quantitative definition of the surface feature is proposed based on qualitative analysis.
The approximate surface near a sampled point is obtained by the local surface fitting
method. Then the feature of the surface near the sampled point is described by the average
of the normal curvature in 360 degree instead of the average curvature. A KD tree
partitioning method is adopted to segment the cloud points according to the surface
feature,the size of space area and the size of sampling nodes. Experiments show that this
method preserves the geometry feature of the surface better. This result demonstrates the
efficiency of the method.

Key words: surface reconstruction;KD tree;surface variance;feature curvature