微型复杂曲面零件散乱点云特征点提取

Feature Point Extraction of Scattered Point Cloud for Complex Micro-Surface Parts

基于特征线约束的曲面不仅比自由曲面表达的模型有更高的精度,而且能够捕捉到模型其他的几何关系。微型复杂曲面零件多由曲面构成,曲面曲率变化大,曲率估算难度大,是逆向工程的难点。以皮带头复杂曲面为例,对微型复杂曲面微分几何量的估算,特征点的提取进行了研究。通过改进的空间分块策略建立散乱点云的拓扑关系,采用多种曲面模型分别拟合采样点邻域的方法估算曲率,构造协方差矩阵计算特征值和点云法矢,对曲率和协方差矩阵的特征值设置交叉阈值提取特征点。将利用交叉阈值提取到的特征点与利用单一阈值提取的特征点进行对比,结果显示:交叉阈值提取特征点的方法能够较为准确、完整的提取特征点。

The surface based on feature line constraint not only has higher precision than free-form surface, but also can capture the other geometric relations of the model. Micro parts with complex surfaces are mostly composed of free-form surfaces. The curvature of the surfaces varies greatly and it is difficult to estimate the curvature of the surfaces, which is the difficult of reverse engineering. Taking the complex surface of a buckle as an example, the estimation of differential geometry of complex micro-surface and the extraction of feature points are studied. The topological relation of the scattered point cloud is established by the improved spatial blocking strategy, and the curvature is estimated by fitting the sampling point neighborhood with various surface models, and the eigenvalues and the normal vectors are calculated by constructing the covariance matrix. The cross threshold which combine eigenvalue and curvature is set to extract the feature points. The feature points extracted by cross threshold are compared with the feature points extracted by single threshold. The results show that the method of cross-threshold extraction of feature points can be more accurate and complete.