基于SMPL的视频人物动作迁移研究  

Research on motion transfer of video characters based on SMPL

作　　者：叶柯 麻福旦 陈相余 YE Ke;MA Fudan;CHEN Xiangyu(Hangzhou Dianzi University,Hangzhou 310018,China)

机构地区：[1]杭州电子科技大学,杭州310018

出　　处：《计算机应用文摘》2023年第16期91-93,共3页Chinese Journal of Computer Application

基　　金：国家级大学生创新创业训练项目(202210336020)。

摘　　要：作为3D视觉领域最热门的方向之一,三维视觉重建在近几年正飞速发展,并在未来的数字人、元宇宙等方面展现出巨大的潜力。作为计算机视觉的重要技术之一,基于视觉的三维重建技术通过深度数据获取、预处理、点云配准与融合、生成物体表面等,将真实场景转换为符合计算机逻辑表达的数学模型。本文利用ROMP,smplify-x,OpenPose等技术,基于SMPL模型,探究了一种仅需要正背面两张照片即可快速构建粗略人物模型的方案,实现了对人物模型的可泛化、可驱动和可快速上色的动作迁移。As one of the most popular directions in the field of 3D vision,3D vision reconstruction is developing rapidly in recent years,and shows great potential in the future digital human,Metaverse and other aspects.As one of the important technologies of computer vision,vision based Iterative reconstruction technology transforms the real scene into a mathematical model that conforms to Logic in computer science expression through the process of depth data acquisition,pre-processing,point cloud registration and fusion,and object surface generation.This paper uses ROMP,smplify-X,OpenPose and other technologies,based on the SMPL model,to explore a scheme that can quickly build a rough Model figure with only two photos on the front and back,and realize the action migration of the Model figure that can be generalized,driven and quickly colored.

关 键 词：三维视觉 三维重建 SMPL模型 动作迁移 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]