【作者】 周术诚；

【导师】 周明全；

【作者基本信息】 西北大学 ， 计算机软件与理论， 2007， 博士

【摘要】 基于形状匹配的拼接技术广泛应用于考古、三维重建、工业设计等领域。破碎物体的复原是将形状匹配的相邻碎片拼接以使破碎物体恢复原貌。形状匹配和拼接、破碎物体复原技术的研究涉及到计算机图形学、图形图像融合、模式识别和计算机可视化等内容。本文研究了破碎物体形状特征的提取和表示、形状匹配、破碎物体复原中的碎片拼接和拼接误差的修补等技术，提出了一系列的形状匹配和拼接方法，满足破碎物体复原的需求。研究的主要内容和创新如下：1．研究了破碎物体的数据采集、网格模型的简化、数字化物体轮廓曲线的提取、曲线特征表示及选择方法，提出了一种基于小波变换自适应阈值的图像分割技术。2．研究了二维Fourier轮廓描述符的特点，提出了基于多尺度Fourier描述符的二维曲线匹配方法。轮廓曲线通过不同尺度的Gaussian函数滤波后，以Fourier变换系数的实部和虚部构成曲线特征向量，通过特征向量的比较以实现二维曲线匹配和碎片拼接，该方法具有准确、简单、快速等优点。3．研究三维曲线的B样条表示方法，通过建立局部直角坐标系计算轮廓曲线上连接点的方向角，提出应用三维曲线上连接点方向角构造特征向量的计算方法及基于方向角的三维曲线匹配算法以实现碎片拼接。4．研究多尺度小波轮廓描述符的计算，提出了一种多尺度下轮廓曲线特征提取及基于多尺度分析的三维物体碎片拼接方法。曲线经多尺度小波变换后，计算轮廓曲线上各点的曲率和挠率构成曲线的特征矢量，经过相似性比较实现轮廓曲线匹配和拼接。研究了小波轮廓描述符匹配起点的解决方法，提出一种判断曲线匹配段的方法。实验表明该算法的准确性、有效性、鲁棒性。5．在分析曲面表示和曲面特征的基础上，研究了物体曲面分割技术，提出了分别用曲面上点的有向脚标和无向脚标构成曲面形状的有向特征向量和无向特征向量，利用粗匹配和细匹配技术实现曲面匹配计算，完成破碎物体断裂面的拼接技术。研究了破碎物体拼接方法、拼接误差处理和孔洞的修补问题。6．在理论研究和分析的基础上，将形状拼接技术应用于计算机辅助破碎物体复原系统的开发。介绍了系统功能、数据结构、技术特点和实现方法。更多还原

【Abstract】 Shape macthing is widely used in many disciplines, such as archaeology, 3D reconstruction, and so on. Neighboring fragments with shape matching is reassembled to make broken objects restoration. Shape matching and Mosaicing of broken objects relate to computer graphics, graphics and images fusion, computer visualization technology. Shape features of fragments extracting and representing, shape matching, mosaicing fragmens, methods of repairing errors techniques are researched. Some techniques of shape matching and fragments mosaicing are presented. The main research works and contributions are concluded as follows.1. Fragments digitalization, mesh simplification, extracting mesh outlines, representation of contour curves and feature are researched. Segmenting Method Based on wavelet transform adaptive threshold is presented.2. Two dimension Fourier contour descriptors at multiscales are researched. A novel 2D curves matching method based on multiscale Fourier descriptors is presented. Contour curves are filtered by Gaussian filter at different scale. Eigenvectors are consisted of real and imaginary part of Fourier transformation coefficients to realize contour matching and objects mosaicing through comparing feature vectors. This approach has advantage of simple, accuracy and fastness.3. Curve represented by B spline is researched. Local coordinate system is built to calculate direction angle. Calculating method of Direction angles used to constructe eigenvectors and 3D curve matching algorithm based on direction angle are presented to complete fragments mosaicing.4. Calculating method of multiscale wavelet contour descriptors is researched. Methods of Extracting contour curves features at multiscale and technique of mosaicing 3D fragmented objects based on multiscale analysis are presented. Spatial contour curves are transformed by wavelet at an appropriate scale. The curvatures and torsions of smoothed contour by wavelet transform at a scale are calculated. Eigenvectors is consisted of Curvatures and torsions. Contour curves matching and fragments mosaicing through similarity measuring are completed. The problem of matching starting point based on wavelet contour descriptors is researched. Decision method of matching curve segment is presented. Time complexity of the algorithm is estimated. Experimental results show that method is accuracy, effectiveness and robustness.5. Surface representation, characteristics of surface are analyzed. Objects surface segmentation technique are researched. Directed eigenvector and undirected eigenvector consisting of directed footprints and undirected footprints of points on surface respectively are presented. Surface matching method by coarse matching and fine matching is presented to complete fractured surface mosaicing. Mosaicing methods, mosaicing errors processing and holes repairing technique are researched.6. Shape mosaicing techniques is applied in development of Computer-aided broken objects recovering system based on theory analysis. Functions analysis, data structure, technique characteristic and development process are introduced.更多还原