Real-time 3D Reconstruction Based On Multiple Kinect

With the rise and development of digital media and virtual reality,3D reconstruction has been widely used in many fields.This paper designs and implements a set of 3D reconstruction system with good friendliness and high practicality through multi-Kinects.At present,the existing three dimensional information acquisition equipment is expensive and complicated,which is not suitable for application.In view of the shortcomings of existing technology,a 3D reconstruction method based on multi-Kinect is proposed in this paper.The main research work process is as follows:firstly,the color images and depth images obtained by Kinect are registered and converted into 3D point clouds with color information,then the background subtraction method is used to remove the image background,and the target image information is extracted,secondly,the point cloud data of the object is denoised and filtered,next,place the three Kinects into 120 degree angles for each other,the Kinects are calibrated by Zhang's calibration algorithm,therefore,the conversion between three coordinate systems is completed,three point clouds of different angles are spliced by combining AABB bounding box and ICP algorithm,and the algorithm is improved by KD-tree acceleration,then the 3D model is reconstructed.Finally,using regional growth algorithm to complete surface reconstruction of 3D model.The experimental results show that the proposed method in this paper is low cost and easy to operate,which can be used to reconstruct the object with high accuracy in a short time.