利用曲率多特征融合改进人体三维脊椎模型特征点的标注方法  

Improvement of feature point annotation for three-dimensional spine model based on curvature multiple feature fusion

作　　者：惠宇[1] 武君胜[2] 鱼滨[3] 张琛[3] 武震[2] HUI Yu WU Jun-sheng YU Bin ZHANG Chen WU Zhen(College of Computer Science, Northwestern Polytechnical University, Xi＇an 710072, China College of Software and Microelectronics , Northwestern Polytechnical University, Xi＇an 710072, China College of Computer Science, Xidian University, Xi＇an 710072, China)

机构地区：[1]西北工业大学计算机学院,陕西西安710072 [2]西北工业大学软件与微电子学院,陕西西安710072 [3]西安电子科技大学计算机学院,陕西西安710072

出　　处：《光学精密工程》2016年第11期2872-2879,共8页Optics and Precision Engineering

基　　金：国家自然科学基金资助项目(No.61172147)

摘　　要：由于脊椎生理结构的精准坐标描述和准确匹配尚未达到医学精度的要求,本文对如何精确描述脊柱腰骶段特征点的物理坐标进行研究。介绍了人体脊椎采样特征点的定义和传统标记方法。针对手动标注特征点精确度不够,易产生较大误差等问题提出了一种基于曲率多特征融合的自适应标注特征点的方法。该方法首先找出某个特征点的高斯曲率和平均曲率流的定义值,得到该特征点的法曲率相对极大值,并计算在指定极小半径r范围内的所有模型点的法曲率相对极大值。由于极大值曲率越大,三维模型表面在该点处的弯曲程度越大,该点就越能表现三维模型的大致轮廓,故以r范围内极大值最大的点作为特征点的曲率描述来替换手动拾取的点,从而准确反映该点的特征变化情况。最后,对改进结果进行偏差验证分析。结果表明:改进方法的准确度比现有手动标注特征点方法的准确度提高了约37%,验证了本文方法的有效性。Because the precise coordinate description and accurate matching of physiology of the lumber spine have not yet reached the accuracy requirement of the medicine,this paper researches how to describe the physical coordinates of lumber spine feature points precisely.The definition and traditional annotation method of human spine sample features were introduce.Then,an improved method to adaptively annotate the feature points was proposed based on curvature multiple featurefusing to solve the problem of weaker precision for the manual annotation of feature points.In this method,the definiting values of Gaussian curvature and average curvature of a certain feature point were found out firstly,and a relative maximum value of curvature of the feature point was obtained.Then,the relative maximum values of all the model points within a specified minimum radius r were obtained.Since the larger the curvature of maximum value was,the larger the crook degree of the 3-dimensioanl model surface on the point was,and the point can show the outline of the 3-dimensional model.Therefore,the maximum value was taken as the curvature description of the feature point to replace the manual pick up point,thus the feature variation of the point could be described precisely.Finally,the verification results were analyzed,and the results indicate that the improved method increases the accuracy by 37%as compared with that of traditional annotation methods.These results demonstrate that the new method is feasibility and effectiveness.

关 键 词：脊椎 三维模型 特征点标注 曲率多特征融合 极大值曲率 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]