基于局部特征匹配的旋转对称提取

Rotational symmetry extraction based on local feature match

摘要提出了一种基于局部特征匹配的旋转对称提取算法，为基于特征的逆向工程提供了一种新的对称约束的提取方法。该算法首先利用特征线将模型分割成对应各个特征类型的体素的集合；然后计算各个体素的形状分布曲线，通过形状分布曲线匹配的方式获取几何形状相似的体素，并利用最小包围盒从中提取大小相似的体素以构成对称体素的集合；最后利用迭代最近点算法计算对称体素的旋转对称信息并利用均值漂移算法优化提取的对称。实验结果表明，体素分割结果对应构成模型的各个特征，具有明显的工程语义，对称提取结果的误差较小并具有较强的鲁棒性。

An algorithm for extraction of rotational symmetry based on local feature match is presented to provide a new symmetric restriction extraction method for the feature-based reverse engineering. The algorithm can be described below ： Firstly, a model is segmented into the set of voxels which are corresponding to each feature based on its fea- ture lines. Then, for each voxel, a shape distribution histogram is calculated with the D2 shape distribution func- tion. By comparing each pair of shape distribution histogram using well-know curve matching techniques,the voxels with the similar geometric shape are acquired and a set of symmetric voxels are constructed by the ones with a simi- lar minimum bounding box. Finally, the symmetry is extracted by the algorithm of iterative closest point and opti- mized by the mean shift approach. The experimental result shows that the voxels of the segmentation corresponding to the features which are composed of the model have obvious engineering semantics and the extracted symmetry in- formation is accurate and robust.