摘要
提出了一种新的、有效地用于逆向工程中的多视曲面拼合算法,即点纹法。系统阐述了点纹法的思路,主要包括:在候选点云的基础上生成纹路圈,在某幅视图中取点p,并做出切平面,把这些纹路圈投影到切平面上形成二维点纹,其它视图也如此产生相对应的点纹,并进行基于二维轮廓的多视曲面拼合,阐述了改进的ICP算法和基于径向基函数(RBF)神经网络的数据融合算法。点纹法与其它多视曲面拼合算法相比,能够大大提高曲面匹配的精度,在模式识别和逆向工程中有现实的使用效果。
A new and efficient multi-view surface registration algorithm in reverse engineering, point fingerprint algorithm was purposed. The algorithm is to generate a geodesic circle based on the candidate point, to form a set of two dimensional contours that are the projections of geodesic circles on the tangent plane, then to match multi-view surface on the 2-D contours. This algorithm uses revised ICP algorithm and the data fusion algorithm based on the radial primary function neuron network. Point fingerprint algorithm can improve the matching accuracy compared with other multi-view surface registration algorithm. Point fingerprint was successfully applied to pose estimation of real range data and match multi-view in reverse engineering.
出处
《机床与液压》
北大核心
2008年第1期124-125,128,共3页
Machine Tool & Hydraulics
关键词
点纹法
纹路圈
多视曲面拼合
数据融合
RBF网络
改进ICP算法
Point fingerprint
Geodesic circle
Multi-view surface registration
Data fusion
RBF network
Revised ICP algorithm