3D Reconstruction Technology Based On Kinect Point Cloud Data And Sequence Images

With the development of science and technology,3D reconstruction technology has been used more and more in various fields,such as virtual reality,3D print and so on.Microsoft's Kinect device is being used more and more in 3D reconstruction for its low price and relatively good point cloud quality.In this paper,3D reconstruction based on Kinect's RGB-D data is studied.The main research contents are as follows:First of all,this paper recalibrates internal parameters of RGB camera and depth sensor,and recalibrates the relative pose of RGB camera and depth sensor.In the past development and research of Kinect,developers are using the rough camera parameters given by the manufacture,and there is no calibration method of high precision to calculate the relative pose between RGB camera and depth sensor.This not only leads to a decline in the quality of the point cloud,but also become an obstacle to RGB-D data fusion.In this paper,the principle of Kinect imaging is analyzed,and we use checkboard calibration method,calculate the internal and the external parameters of these two cameras.Through this step,the color information has been fully utilized.Secondly,to deal with the problem of Kinect's point cloud's poor quality,this paper puts forward some improvement schemes for the optimization of the point cloud data.For the optimization of Kinect data,we divided into two steps,optimization of depth map level and optimization of point cloud level.After these steps,we reduce the impact of noise on point cloud data to some extent,and delete much error points,get a higher quality point cloud.Then,this paper presents a joint registration algorithm based on depth data and RGB information.This algorithm improves the traditional point cloud registration algorithm based on RGB-D data,firstly,the SIFT feature points of two view points are extracted,then the feature points are matched to find the corresponding feature points,then the two-dimensional feature points are converted into 3D point clouds,finally,the method of singular value decomposition is used,calculate the relative transformation matrix of two point clouds.This matrix is used as the initial value of ICP registration to achieve a higher registration accuracy.Finally,based on the improved algorithm,a fast 3D reconstruction system based on multi-Kinect device is proposed,which can realize automatic operation,and has a good registration precision and model reconstruction effect.Experimental results show that,after recalibration,the quality of point cloud is much higher than before.After using the point cloud registration algorithm presented in this paper,greatly improve the success rate and accuracy of point cloud registration.