基于移动最小二乘法法矢估计的建筑物点云特征提取

Feature Extraction of Building Point Cloud Based on Moving-least Squares Vector Estimation

特征提取对建筑物精细建模的品质和精度起着重要作用。为清晰准确地提取建筑物的特征信息,本文针对采用传统的法矢估计方法受噪声影响大、存在误判的问题,提出了一种基于移动最小二乘法矢估计的建筑物点云特征提取方法。该方法首先采用移动最小二乘法进行法矢估计,然后将K邻域法矢夹角的均值作为点的显著性指标进行特征点判别,最后对提取出的特征点集进行下采样,进一步消除冗余信息。试验结果表明,采用移动最小二乘法进行点云法矢估计,其结果更加准确和稳健,从而有效提升了建筑物点云特征提取的精确性和可靠性,对特征点集的下采样能够删除大量冗余特征点,使提取的特征线更加简洁、清晰、完整。

The feature extraction quality and accuracy of the fine modeling building plays an important role in the feature information extraction of buildings clearly and accurately.The traditional vector estimation method is affected by noise,and exists some misjudgment problem,this paper proposed a feature extraction method of the point cloud mobile minimum two multiplication of buildings based on vector estimation. This method firstly uses the moving least squares method for estimation of normal vector,then the mean K neighborhood method as a significant index vector included angle point feature identification,finally the extracted feature points set by resampling,to eliminate redundant information.The experimental results show that the estimation of point cloud normal is more accurate and robust by moving least square method,thus effectively improve the accuracy and reliability of the building point cloud feature extraction,the feature point set resampling can delete the redundant feature point,the feature line extraction is more concise,clear and complete.