Research On 3D Reconstruction Algoriyhm Of Object Based On Multi-view Color-depth Data

Three-dimensional reconstruction technology refers to the establishment of three-dimensional objects' mathematical model in computer,so that a series of operations and analysis of three-dimensional objects can be done in computer environment.Three-dimensional reconstruction technology is a multi-disciplinary technology,and has many important applications in human-computer interaction,virtual reality,robot navigation and many other fields.Traditional three-dimensional scanning devices are expensive and cumbersome to operate,so the research of a low-cost and easy-operate three-dimensional reconstruction method is very meaningful.In view of this,this paper regards the low-cost Kinect sensor as the input device of three-dimensional reconstruction,and studies the data acquisition,target extraction,point cloud denoising,text mapping,meshing and other related technologies involved in three-dimensional reconstruction process,then a simple and efficient three-dimensional reconstruction system based on Kinect sensor is realized.The main contents of this paper are divided into the following aspects:Firstly,OpenNI is used to drive the Kinect sensor to obtain the depth information and color information of the object to be scanned,then the camera of the Kinect sensor is calibrated by the checkerboard calibration algorithm through combing with MATLAB calibration toolbox and convert the obtained depth and color information into three-dimensional point cloud by coordinate transformation.Secondly,due to the large amount of noise in the depth image acquired by Kinect sensor,in order to improve the quality of model reconstruction,the depth image is denoised by the bilateral filtering algorithm,and the edge information is preserved at the same time.In order to reconstruct precise three-dimensional model,the object to be reconstructed needs to be segmented and extracted.In this paper,the improved region marker algorithm is used to segment the depth image.The segmented depth image is repaired,so that the holes of small area have been effectively filled.Finally,a three-dimensional reconstruction system through Kinect sensor is built,which can obtain the depth images and RGB images from multiple perspectives through Kinect sensor,obtain the point cloud of object to be reconstructed after the segmentation,repair and denoising of the depth images,and reconstruct the object model,what's more,when the camera's posture changes,the system will automatically adjust the posture and make the new frame integrated into the existing model,then ultimately the complete three-dimensional mesh model with texture information can be reconstructed.In addition,the system uses GPU high-performance programming to speed up the operation,and carries out some optimization processing,which reduces computing time drastically and improves the speed of model reconstruction.This paper uses low-cost Kinect sensor and ordinary computer to complete three-dimensional reconstruction of objects,this method is easy-operate.This method greatly reduces the cost and difficulty of three-dimensional reconstruction compared with the traditional three-dimensional scanning devices.The three-dimensional model generated in the experimental results has high accuracy and can meet the application requirements of inaccurate calculation.