医疗三维几何模型的对称面定位研究
发布时间：2018-09-13 08:32
【摘要】：医疗三维几何模型的对称平面定位适用于从生物识别到口腔正畸科和整形外科的医学面部重建等各种应用场景。对称面定位的关键通常是三维几何模型表面的配准技术,其中使用最广泛且最具代表性的是迭代最近点算法,基于迭代最近点配准技术的对称面定位避免了传统点构建法生成对称平面容易产生随机偏差的问题,该研究对于临床医学领域具有十分重要的意义。本文首先对基本的迭代最近点算法做了详细的介绍,主要包括算法的基本原理、算法步骤和坐标变换参数的求解等,接着研究了迭代最近点算法的时间复杂度和特性等问题,并通过实验对算法的理论进行了验证,分析得到算法的待改进点。本文提出了一种基于快速近似最近邻搜索和多重曲率约束的迭代最近点算法,该算法从标准算法的不足之处入手,将主成分分析应用于初始配准,在算法的点对匹配问题上,一方面用改进的最近邻搜索方法进行加速,另一方面则通过建立多重曲率特征的约束对匹配的精度进行优化,并通过实验证实了改进算法的确具备更好的配准效果。本文将上述改进的迭代最近点配准技术应用于医学颅骨模型的对称面定位功能,通过点云配准技术对三维模型进行镜像、配准和对称平面定位等操作,并从量化的角度分析了提取平面的对称效果,实验证明定位的对称面比较精确,对于临床医学领域具有较好的实用价值。
[Abstract]:The symmetrical plane localization of 3D medical geometry model is suitable for various applications from biometric recognition to orthodontic and orthodontic medical facial reconstruction. The key of symmetrical surface location is usually the registration technique of 3D geometric model surface. The most widely used and most representative is iterative nearest point algorithm. The symmetric plane location based on iterative nearest point registration technique avoids the problem that the symmetric plane generated by the traditional point construction method is prone to generate random deviations. This research is of great significance in the field of clinical medicine. In this paper, the basic iterative nearest point algorithm is introduced in detail, including the basic principle of the algorithm, the steps of the algorithm and the solution of coordinate transformation parameters, and then the time complexity and characteristics of the iterative nearest point algorithm are studied. The theory of the algorithm is verified by experiments, and the points for improvement are obtained. In this paper, an iterative nearest point algorithm based on fast approximate nearest neighbor search and multiple curvature constraints is proposed. Starting with the shortcomings of the standard algorithm, principal component analysis is applied to the initial registration. On the one hand, the improved nearest neighbor search method is used to accelerate the algorithm; on the other hand, the matching accuracy is optimized by establishing the constraints of multiple curvature features, and the experimental results show that the improved algorithm does have better registration effect. In this paper, the improved iterative nearest point registration technique is applied to the symmetrical plane location function of the medical skull model. The 3D model is mirrored, registered, and symmetrical plane location is performed by the point cloud registration technique. From the point of view of quantification, the symmetry effect of extracting plane is analyzed. The experimental results show that the symmetrical plane is more accurate and has good practical value in the field of clinical medicine.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP18
本文编号：2240618
[Abstract]:The symmetrical plane localization of 3D medical geometry model is suitable for various applications from biometric recognition to orthodontic and orthodontic medical facial reconstruction. The key of symmetrical surface location is usually the registration technique of 3D geometric model surface. The most widely used and most representative is iterative nearest point algorithm. The symmetric plane location based on iterative nearest point registration technique avoids the problem that the symmetric plane generated by the traditional point construction method is prone to generate random deviations. This research is of great significance in the field of clinical medicine. In this paper, the basic iterative nearest point algorithm is introduced in detail, including the basic principle of the algorithm, the steps of the algorithm and the solution of coordinate transformation parameters, and then the time complexity and characteristics of the iterative nearest point algorithm are studied. The theory of the algorithm is verified by experiments, and the points for improvement are obtained. In this paper, an iterative nearest point algorithm based on fast approximate nearest neighbor search and multiple curvature constraints is proposed. Starting with the shortcomings of the standard algorithm, principal component analysis is applied to the initial registration. On the one hand, the improved nearest neighbor search method is used to accelerate the algorithm; on the other hand, the matching accuracy is optimized by establishing the constraints of multiple curvature features, and the experimental results show that the improved algorithm does have better registration effect. In this paper, the improved iterative nearest point registration technique is applied to the symmetrical plane location function of the medical skull model. The 3D model is mirrored, registered, and symmetrical plane location is performed by the point cloud registration technique. From the point of view of quantification, the symmetry effect of extracting plane is analyzed. The experimental results show that the symmetrical plane is more accurate and has good practical value in the field of clinical medicine.
【学位授予单位】：浙江大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP18
【相似文献】
相关期刊论文 前2条
1 ;谨防“画病”[J];新闻通讯;1987年06期
2 张宝贵;刘露;宋明;蒋田仔;;一种精确提取三维脑图像对称面的方法[J];中国医学影像技术;2011年08期
相关硕士学位论文 前2条
1 程小富;医疗三维几何模型的对称面定位研究[D];浙江大学;2017年
2 罗莉;具有一个弹性对称面的各向异性介质椭圆夹杂反平面问题[D];同济大学;2008年
，本文编号：2240618
本文链接：https://www.wllwen.com/kejilunwen/zidonghuakongzhilunwen/2240618.html
