基于数字几何信息增益的服装3D人体特征识别

3D Human Body Features Recognition Based on Geometric Information Gain

提出了一种基于数字几何信息增益的服装3D人体特征识别算法.首先,给出3D扫描人体的数字几何信息增益的定义,计算3D扫描人体模型上各点的数字几何增益信息并将统计结果由大到小进行标记,由人体比例学将3D扫描人体模型区域分割为颈部、肩部、腋下、胸部、腰部、臀部,并选取数字几何信息增益值变化显著的点作为特征点.然后对特征点进行3次B样条拟合得到3D扫描人体轮廓线,并测定指导服装设计的人体尺寸数据.最后将人体尺寸测量结果与标记测试数据进行比较,计算误差值.仿真结果表明本文算法对不同的3D人体扫描模型具有准确的识别效果.

A new method to recognize 3D human body features based on geometric information gain was proposed in this paper. The definition of digital geometric information gain of the three-dimensional（ 3D） objects was given first,then the local geometry of each point on the 3D scan human body was measured,which is followed by sorting the computed digital geometric information gain values in descending order. After that,the whole 3D scan human body was segmented into such regions as neck,shoulder,axillary,chest,waist and hipbased on the human body proportion relations,andthe points with salient values of digital geometric information gain were chosen as optional feature points,which were applied to generate the 3D human body feature line by cubic B-spline fitting. The 3D human body feature lines were measured to obtain the human body size data for the guidance of fashion design. Finally,the resulted human body size data were compared with the standard measurement data to examine the performance of the proposed method in this paper. The results showed that the simulation have good recognition to various 3D scan human body.