基于图像的人体颈肩部三维模型构建

为实现个性化服装在线设计和虚拟试衣,提出一种基于人体正侧面照片的尺寸测量及建模方法。首先结合202名青年女性的三维点云数据确定与人体颈肩部形态相关的8个特征截面层,以各截面层中心点为基准点每隔10°测量角度半径并分析其与截面厚度、宽度之间的关系,建立各截面层的曲线形态规则。然后基于人体正侧面照片进行图像分割以提取人体轮廓,通过人体高度比例关系识别颈肩部特征点所在高度区域,结合颈肩部形态规律提取各特征截面层曲线形态规则所需参数。最后根据曲线形态规则和人体测量参数,通过NURBS曲面建模实现基于人体正侧面照片的颈肩部三维模型构建。结果表明:基于人体照片提取的人体基本参数值的误差百分比均小于5%,且使用该方法构建的三维模型与真实值之间的误差绝对值均在1 cm或1°以内。

3-D modeling of neck-shoulder part based on human photos

In order to facilitate personalized clothing online design and virtual fitting,a method of size extraction and 3-D modeling based on the front and side photographs of human body was proposed.Eight characteristic cross-sectional layers related to the neck-shoulder shape were determined based on 202 young women′s 3-D point cloud data.Taking the center point of each cross-section layer as the reference point,the angle radius was measured every 10°and the relationship among the angle radius,the thickness and width of the curves were analyzed to establish the curve shape rules of each sectional layer.Then,the image segmentation technology was used to extract the human contour based on the front and side photos,the height area where the neck-shoulder feature points were located was identified through the height proportion of the human body,and the parameters required for the curve shape rules of each characteristic section were extracted with the neck-shoulder shape rules.Finally,according to the curve shape rules and anthropometric parameters,NURBS surface modeling was used to realize the 3-D model construction of neck and shoulder based on human body photos.The results show that the error percentage of the basic parameters extracted from human body photos is less than 5%,and the absolute error between the 3-D model constructed by this method and the real value is within 1 cm or 1°,which proves the feasibility of this method,and provides technical support for the automatic generation of garment pattern and 3-D virtual fitting.