一种新的散乱点云尖锐特征提取方法

An Extraction Algorithm for Sharp Feature Points from Point Clouds

针对散乱点云的尖锐特征识别与提取问题,提出一种基于平均曲率运动的散乱点云尖锐特征提取算法.该算法利用采样点的加权邻域重心近似表示离散Laplacian算子;利用采样点邻域的主成分分析估算散乱点云法向量,通过张量投票的方法平滑估算得到的点云法向场,进一步提高了该算法识别细微尖锐特征的能力;将采样点和其对应加权邻域重心之间的距离投影到法向方向,消除了因为采样密度不均匀以及边界点所引起的尖锐特征点误判.该算法直接对散乱点云进行操作,不需要维护采样点之间的连接关系和任何全局的拓扑信息,简捷且易于实现,对点云中的噪声和局外点保持鲁棒.该算法应用于合成点云和实际扫描点云的实验结果表明了方法的有效性.

A novel algorithm for extraction of sharp feature points from scattered point cloud is proposed.The proposed method bases on the concept of mean curvature motion.The weighted local barycenter is used as an approximation of discrete Laplacian operator.If the distance between a sample point and its weighted local barycenter is greater than a given threshold,then the sample point is labeled as a candidate sharp feature point.Then,the normal directions of these candidate points are estimated using the local PCA method.In order to extract finer sharp feature points,the estimated normal field is further smoothed via tensor voting frame.The distance between a point and its weighted local barycenter is projected along the direction of normal,so that the fault sharp feature points due to uneven sample ratio and point cloud boundary can be eliminated successfully.Experiments on both real scanner point clouds and synthesized point clouds show that the proposed method is easy to implement,efficient for both space and time overhead,and is robust to noise,outlier and uneven sample ratio inherent in point clouds.