针对人体体型分类标准选择问题,提出了通过人体纵截面特征曲线进行分类的方法.选择了124名青年女性,使用三维扫描仪扫描获得三维人体数据.并提取人体纵截面轮廓曲线数据,通过椭圆傅里叶拟合对原始曲线进行降噪处理.使用转向角函数将截面轮廓曲线转换成角函数曲线,并根据角函数曲线特征提取了5个特征指标,利用主成分分析法提取3个主成分.采用K?means聚类算法进行体型聚类,用伪F统计量确定最佳分类数目,将体型分为4类.最后通过对比分析4类体型特点,描述了各类体型在背部、背中心线、腹部、臀部和侧面厚度的差异.%A classification method of the human body based on the longitudinal section cure was proposed for classification basis of somatotype. A 3?D human body measurement instrument was used to measure 124 young female samples. The point cloud data of the longitudinal section cure could be got, and analyzed and processed by some way. The elliptical Fourier fitting was used for original curve denoise processing. To get turning function cure of longitudinal section curve a function which was called turning function was used, then five characteristic indexes were extracted according to characteristics of turning function cure. A method called the principal component analysis was used to extract three principal components. As for shape clustering the K?means cluster analysis was used, and the Pseudo F?statistics was used to determine the optimal class number. Then longitudinal section curve could be divided into four classes. Finally, the difference on back, centre back seam line, abdomen, hip and side thickness of all kinds of shapes was described.
展开▼
