摘要
采用多尺度分析技术实现三维轮廓曲线匹配.三维轮廓曲线通过不同尺度的Gaussian函数滤波和等距重采样,将曲率和挠率的乘积为局部极大值的点作为轮廓曲线的特征点,利用特征点将轮廓分段,对轮廓曲线进行Fourier变换得到Fourier描述符;选择Fourier描述符的低频分量构成三维轮廓曲线的特征矢量,通过比较特征矢量决定2条轮廓是否相似;在2条轮廓相似的基础上,实现三维物体轮廓曲线的匹配.结果表明本文提出的算法具有快速、准确、效果好等特点.
Multiscale analyzing technique was used to realize 3D contour curve matching.3D contour curves were filtered by Gaussian function at different scales and were resampled with equal interval.Points where the product of curvatures and torsions were local maxima were regarded as feature points.3D contour curves were segmented by feature points.Fourier descriptors were gotten by Fourier transform of contour curves.Low frequency components of Fourier descriptors were selected to construct eigenvectors of 3D contour curves.Similarity between two contour curves was determined by comparing eigenvectors of two 3D contour curves.Two 3D contour curve matching was realized when the eigenvectors of two contours were similar.Experimental results showed that this approach had high speed,accuracy and better effect in 3D contours matching.
出处
《福建农林大学学报(自然科学版)》
CSCD
北大核心
2010年第6期668-672,共5页
Journal of Fujian Agriculture and Forestry University:Natural Science Edition
基金
福建省自然科学基金资助项目(S0750003)
福建省教育厅资助项目(JA10119)