Tsinghua University

Yebin Liu

Qionghai Dai

Di Wu

Lu Fang

Wei Cheng

Lan Xu

Mengqi Ji

Genzhi Ye

Provided are a three-dimensional reconstruction method, apparatus and system of a dynamic scene, a server and a medium. The method includes: acquiring multiple continuous depth image sequences of the dynamic scene, where the multiple continuous depth image sequences are captured by an array of drones equipped with depth cameras; fusing the multiple continuous depth image sequences to establish a three-dimensional reconstruction model of the dynamic scene; obtaining target observation points of the array of drones through calculation according to the three-dimensional reconstruction model and current poses of the array of drones; and instructing the array of drones to move to the target observation points to capture, and updating the three-dimensional reconstruction model according to multiple continuous depth image sequences captured by the array of drones at the target observation points.

Dynamic scene three-dimensional reconstruction method, apparatus and system, server, and medium

A method and a system for three-dimensionally reconstructing a non-rigid body based on a multi-depth-map are provided. The method comprises: obtaining a plurality of depth maps by shooting the non-rigid body in different postures and from different angles; transforming each depth map to one group of three-dimensional point clouds and obtaining a plurality of matching point pairs among a plurality of groups of three-dimensional point clouds; conducting a position transformation for each matching point and obtaining a transformation parameter corresponding to the each matching point after the position transformation; mosaicing all transformation parameters to obtain a mosaicing result and constructing an energy function according to the mosaicing result; and solving the energy function to obtain a solution result and reconstructing a three-dimensional model of the non-rigid body according to the solution result.

Method and system for three-dimensionally reconstructing non-rigid body based on multi-depth-map

A method for vision field computing may comprise the following steps of: forming a sampling system for a multi-view dynamic scene; controlling cameras in the sampling system for the multi-view dynamic scene to perform spatial interleaved sampling, temporal interleaved exposure sampling and exposure-variant sampling; performing spatial intersection to the sampling information in the view subspace of the dynamic scene and temporal intersection to the sampling information in the time subspace of the dynamic scene to reconstruct a dynamic scene geometry model; performing silhouette back projection based on the dynamic scene geometry model to obtain silhouette motion constraints for the view angles of the cameras; performing temporal decoupling for motion de-blurring with the silhouette motion constraints; and reconstructing a dynamic scene 3D model with a resolution larger than nominal resolution of each camera by a 3D reconstructing algorithm.

Method for vision field computing

Growing community of inventors

Beijing, China

Yebin Liu