基于多视图虚拟建模的三维人体尺寸测量

Three-Dimensional Human Dimension Measurement Based on Multi-View Virtual Modeling

摘要: 开发一种操作便携且精度高的非接触式自动测量虚拟建模的三维人体尺寸技术。本文虚拟建模的三维人体数据图像通过便携手机进行非接触式摄像采集，视频通过FFmpeg解析成多个单帧图像，帧间光流能量算子通过稠密逆搜索算法计算，基于SFM虚拟建模三维空间位置点云，分割后虚拟建模的着装轮廓稠密点云通过Graphonomy得到，虚拟建模的三维人体关键特征点通过正面、侧面点云投影轮廓线的曲率迅速定位到。将计算得到的虚拟建模人体特征尺寸与手工测量尺寸对比，论证了该方法的合理性。得出了程序测量值和手工测量值之间的误差较小以及程序测量数据较稳定的结果。

Abstract: Objective: Develop a portable and high-precision non-contact automatic measurement virtual mod-eling three-dimensional human body size technology. In this paper, the virtual 3D human body data image is collected by non-contact camera through portable mobile phone, and the video is analyzed into multiple single-frame images through FFmpeg. The optical flow energy operator between frames is calculated by dense inverse search algorithm, and the 3D spatial position point cloud is modeled based on SFM. After segmentation, the dense point cloud of virtual modeling is obtained through Graphonomy, and the key feature points of virtual modeling of 3D human body are quickly located through the curvature of the projection contour lines of front and side point clouds. By comparing the calculated virtual modeling human body feature size with the manual measurement size, the rationality of this method is demonstrated. The results show that the error between the programmed measurement value and the manual measurement value is small and the pro-grammed measurement data is stable.