逆向工程中基于小波和曲率图像的离散数据配准

Discrete data registration based on wavelets and curvature image in reverse engineering

提出了基于小波和曲率图像的数据配准算法,用于逆向工程中不同视角测量数据的配准。对原始数据不需要进行预处理,该算法可直接操纵原始数据,并自动检测不同视角数据集的公共重叠部分,进行快速配准。首先采用局部Bezier曲面拟合方法计算测量数据的离散曲率,然后采用主元分析方法对数据进行采样,构造曲率图像。利用小波分析对曲率图像进行匹配后,可求得原数据集的公共重叠数据。在此基础上利用单位四元数法可求得空间旋转和平移矩阵,实现原始数据配准。运行实例证明了此算法的有效性。

A new data registration algorithm based on wavelets and curvature image is proposed to align measured data from different angle of views in reverse engineering. Without pre-processing the original data, this method can operate the original data directly and detect the common overlap part among data sets from different views automatically, carrying out fast registration. First is to calculate the discrete curvatures of measured data by applying local Bezier curve surface fitting method, and then is to use Principal Component Analysis （PCA） to sample the data and build the curvature imagesAfter registering the curvature images by using a wavelets analysis, the common overlap part among different data sets can be calculated. Based on this, the space rotation and translational matrixes can be calculated by unit quaternion to realize the registration of original data. The running instance has proved that this algorithm is effective.