计算机工程

• •    

基于径向曲线与支持向量回归的颅骨修复

  

  • 发布日期:2021-01-08

Skull Restoration Based on Radial Curve and Support Vector Regression

  • Published:2021-01-08

摘要: 颅骨修复技术是对有缺损的颅骨补全对应的缺损部分,进而实现颅骨形状的完整性。针对高维颅 骨数据,本文采用径向曲线来表示颅骨几何特征,结合最小二乘支持向量回归的方法构建颅骨修复模型。 首先,提取完整的三维颅骨模型的径向曲线,将其分为已有径向曲线和缺失径向曲线两部分作为训练样本; 然后采用最小二乘支持向量回归(LSSVR)统计模型复原出待修复颅骨的缺失径向曲线,进而合并生成待修复 颅骨的完整径向曲线;最后,通过迭代最近点(ICP)算法将合并的颅骨径向曲线与颅骨统计模型进行匹配 生成完整的三维颅骨模型。在 190 套颅骨数据上进行修复结果对比,该方法的平均误差达到 6.834×10 -3 ,比 主成分分析( PCA)方法提高了 2.90 倍,实验结果证明该方法具有更好的修复效果。

Abstract: The skull restoration technology is to restore the missing part of the defected skull to complete the skull shape. For high-dimensional skull data, radial curves are used to represent the skull geometry features in this paper, a skull restoration model is constructed by combining with the method of least square support vector regression. First, the radial curves of the complete three-dimensional skull model are extracted as training data, which is divided into two parts: the existed radial curves and the defective radial curves; Then the least squares support vector regression (LSSVR) statistical model is used to restore the defective radial curve of the skull to be repaired. Next, they are combined into a set of radial curves of the skull; Finally, the iterative closest point (ICP) algorithm is used to match the generated skull radial curve with the skull statistical model to generate a complete three-dimensional skull model. Compared the restored results on 190 sets of skull data, the average error of our method was 6.834×10 -3 , which was 2.90 times higher than that of Principal Component Analysis (PCA) method. The experimental results proved that this method had a better repair effect.