Research On 3D Reconstruction And Interactive Technology Of Root Carving Based On Kinect

Root carving is an art treasure that pinpoints emotions and has a unique three-dimensional art form.At present,people’s demand for creating root carving works is increasing.In order to provide more amateurs with a simple,convenient and low-cost creative environment,as well as realize the online three-dimensional display of root carving works,digital three-dimensional reconstruction of root carving and virtual pruning technical research is particularly important.In order to meet the technical requirements of digital 3D reconstruction and virtual branch trimming of root carvings,this article uses Kinect sensors to collect multi-angle and multi-frame point cloud data of root carving objects,And then through point cloud data preprocessing,point cloud registration,3D reconstruction and virtual trimming,the interactive technology research has finally designed a three-dimensional reconstruction and virtual pruning system for root carving that is convenient,fast and inexpensive.The main work completed in this article is as follows:(1)In terms of three-dimensional data collection,based on the analysis of the working principle of Kinect,the calibration method of converting depth images into three-dimensional spatial information,the alignment mapping method of color images and depth images,and other technologies,this paper uses Kinect v2 to complete the eight angle multi-frame point cloud data collection.(2)In the research stage of point cloud data preprocessing,this paper carried out a series of preprocessing operations including background removal based on conditional filtering,deletion of outliers based on statistical analysis methods,voxelized downsampling and so on.Among them 1)In view of the interference of noise points in the point cloud data to the follow-up work,on the analysis of the traditional filtering method,bilateral filtering is used to process the depth image,retain the details of the image and eliminate the noise points.2)In order to eliminate the hole problem in the depth image acquisition,the color image is used as the guide image and it is processed using joint bilateral filtering.3)In order to solve the shortcomings of high time complexity of joint bilateral filtering,this paper adopts a method based on fast Gaussian transform to improve joint bilateral filtering,which reduces the calculation time of the filtering algorithm.4)For the problem that the root carving detail information is not completely displayed in the depth image,taking sample 1 as the experimental object,this paper uses the method of multi-frame fusion to increase the number of effective point clouds and perform preliminary smoothing;Compared with the average number of 10-frame effective point clouds,the number of point clouds has increased by 19.5%.The above research has laid a good foundation for subsequent registration and other studies.(3)In the point cloud registration and 3D reconstruction research,1)In the initial registration stage,for the registration problem of point clouds with different perspectives unified into the same coordinate system,the ISS algorithm is used to obtain the main feature points of the cloud from two angle points,and use FPFH to describe these feature points,and then use the sampling consensus algorithm to roughly register the two different angle point clouds,providing a good initial position for accurate registration.2)In the precise registration stage,in view of the shortcomings of the traditional ICP algorithm that takes a long time to operate and is easy to fall into the local optimal solution,this article introduces the Kd-Tree search method and the method of normal threshold angle to improve the ICP algorithm and improve the registration effectiveness.Finally,the improved ICP algorithm was used to complete the registration of multiple point clouds,and a complete root carving model was obtained,which verified the effectiveness of the improved algorithm.3)In the surface reconstruction stage,through theoretical analysis and experimental verification,the Poisson reconstruction algorithm used in this paper has faster operation speed,more realistic reconstruction effect,and no additional texture mapping than the PCL-based greedy projection triangulation algorithm.The above research provides a good root carving model for the subsequent virtual pruning.(4)In the research of virtual branch pruning interaction technology,in view of the characteristics of the root carving model and the requirements of picking accuracy,this paper proposes an improved ray picking method,which uses a buffer to build a virtual screen,get a part of the root carving selected by the user,and obtain its Index ID,calculate the intersection point between the part and the projection ray,and get the XYZ coordinates of the pickup position.Then,according to the selected location,delete the data set of the part to be cut,modify the model data and render the result.This completes the fast and precise pruning operation of the root carving model.The method in this paper has the characteristics of relatively accurate position information of picking parts and convenient and interactive pruning operations.(5)In terms of system implementation,this article uses the C++language combined with OpenCV,OpenGL,PCL,and VTK open source libraries to design and develop a Kinect-based 3D model and virtual pruning system on the Visual Studio 2013 platform.The operating results of the system show that the system can accurately complete the three-dimensional data collection and preprocessing,reconstruct the three-dimensional model of the root carving and complete the operation of virtual branch trimming,which has good practical value.