实时便携大范围多人带拓扑变化的动态场景三维重建方法研究

Real-time dynamic scene reconstruction is a hot research topic in the field of computer vision. The traditional scene reconstruction algorithm is mainly for offline, rigid, single-person, fixed, small-scale, and topologically unchanged simple application scenarios. It is difficult to deal with real-time, non-rigid, multi-person, portable, large-scale, complex with topology changes. Application scenario. This topic will first build a camera hardware platform for portable followers; on this basis, model a large number of non-rigid motion fields with topological changes and design robust parameter estimation strategies; finally, the design is applicable to The real-time nonlinear optimization strategy of GPU parallel acceleration realizes real-time and convenient multi-dimensional three-dimensional reconstruction of dynamic scenes with multi-layer topology changes. This project will provide new ideas and methods for real-time and convenient dynamic scene 3D reconstruction, and provide reference measurement solutions for holographic communication, orbital aerospace shape measurement applications.

实时动态场景重建是计算机视觉领域的热点研究问题。传统的场景重建算法主要是面向离线、刚性、单人、固定、小范围、拓扑结构不变的简单应用场景，难以应对实时、非刚性、多人、便携、大范围、带拓扑结构变化的复杂应用场景。本课题将首先搭建便携可跟随人员拍摄的相机硬件平台；在此基础上，对大范围多人带拓扑结构变化的非刚性运动场进行建模，并设计鲁棒参数估计策略；最后，设计适用于GPU并行加速的实时非线性优化策略，最终实现实时便捷大范围多人带拓扑结构变化动态场景三维重建。本项目研究将为实时便捷动态场景三维重建提供新的思路和方法，并为全息通信、在轨航天员形貌测量等应用提供参考测量方案。