基于单目RGB/RGBD相机的身体运动和面部运动同步捕获方法研究

项目名称： 基于单目RGB/RGBD相机的身体运动和面部运动同步捕获方法研究

项目编号： No.61772499

项目类型： 面上项目

立项/批准年度： 2018

项目学科： 计算机科学学科

项目作者： 夏时洪

作者单位： 中国科学院计算技术研究所

项目金额： 17万元

中文摘要： 在虚拟现实技术和产业发展过程中，虚拟现实硬件发展迅速且已逐渐普及，虚拟现实内容短板日益显现，高质量的虚拟人动画是虚拟现实内容的重要组成部分，如何快速制作高度真实感的虚拟人全身动画就成了当务之急。人们往往通过全身的动作来表达自己内心的想法和情绪。比如，人们在生气的时候不仅脸上会有生气的表情，更有与之相对应的肢体动作。因此，一个具有高度真实感的虚拟人全身动画也必须具有相互协调的肢体和人脸动画这两个重要组成部分。目前，全身运动捕获设备成本非常高，急需发展低成本易于普及的设备来捕获生成全身运动内容。为此，本课题拟研究一种适宜于普及的、轻量级的全身人体运动捕获技术与系统，特别地，基于单目彩色相机和深度相机的身体运动和面部运动的同步捕获方法与系统，以快速制作高度真实感的虚拟人全身动画，为虚拟现实内容制作提供有力的方法和工具支撑，促进我国虚拟现实技术的深入发展和虚拟现实产业的蓬勃发展。

中文关键词： 人体运动；人脸表情捕获；运动捕捉

英文摘要： During the development of virtual reality technology and industry, the hardware for virtual reality has met a fast popularization, which exposes the dificiency of the virtual reality content. High quility virtual human animation is an important part of the virtual reality content. Thus, how to fast build the vivid virtual full-body animation has been of the first priority. Human tends to express their emotion and feelings though full-body motion(movement). For example, an anrgy person not only has the facial expression, but also the correspoding body motion. That is why a highly lifelike virtual full-body animation is composed of two important parts, namely, the facial animation and the corresponding body motion(movements). At the current stage, due to the high cost of full-body motion capture equipment, it is in dire need of devloping a low-cost and easily-popularized system to capture the full-body movement. Therefore, this project aims to build a lightweight algorithm and system for full-body human motion capture, especially, the system of single RGB camera and the depth camera based synchronously capture for body motion and facial motion. With that, we can effeciently build the lifelike full-body animation and provide a robust algorithm and tool for generating virtual reality content, and finally promote the further development of our country's virtual reality technology and industry.

英文关键词： human motion;face performance capture;motion capture