3D reconstruction based on digital speckle correlation in virtual sport

摘要

Virtual sport is a natural, harmonious and immersive system of human-computer interaction. It implements by sensing 3D information of the body movement in the scene. The traditional speckle correlation based 3D reconstruction algorithm cannot meet the actual requirement because of the low searching speed. In this paper, we study the method of the digital speckle correlation measurement deeply, and propose an effective searching method using the relevant information of neighborhood. The propose method can reduce the computational complexity of point correspondence greatly, and improve the speed of depth measurement. Besides, we present a new interpolation method based on distance-driven weights according to the error feature. It can improve the accuracy of depth measurement to a great extent. The tests performed on real captured images confirm the validity of our approach.