3D Human Model Reconstruction Based On Single Depth Camera

With the rapid development of computer graphics and computer vision in the past few decades,the tracking and reconstruction of 3D scenes including human bodies have become increasingly important.How to quickly and conveniently obtain accurate three-dimensional models of moving objects in dynamic scenes is an important part in virtual reality,computer animation and games.This paper designs a human body depth data acquisition system,and uses the dynamic human body depth sequence obtained by Kinect to reconstruct the three-dimensional human body model.The following work is done in the point cloud corresponding point estimation,point cloud non-rigid registration,etc.:Based on the idea of structuer from motion(SFM),this paper designs a human body 3D reconstruction system based on a single depth camera based on group reconstruction,and researches on several key technologies in the process of human body 3D reconstruction.In order to eliminate the accumulation of errors in the non-rigid registration process of the ring motion sequence,the minimum arc distance and dynamic segmentation method are used to divide the entire sequence into several segments;the local non-rigid beam is adjusted to align each segment.Then,the global optimization method is used to combine each segment,and the drift problem is handled through closed-loop constraints.Local non-rigid beam adjustment to accurately register each local segment and reduce error accumulation.Aiming at the erroneous matching that is unavoidable in the process of point cloud overlap acquisition and registration,this paper uses the minimum distance of the arc and the normal vector in combination with the search for matching points to filter to obtain the best matching points and improve the registration accuracy.In this paper,kinect is used to finally generate a deformed three-dimensional model sequence of dynamic objects.Experiments show that this method can improve the 3D scanning speed while maintaining accuracy,and the final experiment verifies the feasibility of this method.