Font Size: a A A

Research On RGB-D Scene Plane Reconstruction Algorithm Based On Image Segmentation

Posted on:2019-11-07Degree:MasterType:Thesis
Country:ChinaCandidate:C GuFull Text:PDF
GTID:2428330566489177Subject:Pattern Recognition and Intelligent Systems
Abstract/Summary:PDF Full Text Request
3D reconstruction technology has been widely used in many fields such as virtual reality,augmented reality,medical field,robotics,reverse engineering and human-computer interaction.In this paper,the 3d reconstruction method of single frame rgb-d data collected by Kinect is studied and an RGB-D three-dimensional plane scene reconstruction scheme based on image segmentation is proposed.The main research contents of this paper are as follows:First of all,this paper proposes an RGB color image segmentation method based on image edge information.The algorithm first extracts anchor points to form the initial edge map.Edge projections for the initial edge operation,and ultimately form a chain of single pixel edge image.Breakpoint detection is performed on the basis of the predicted edge map,and the connection and edge growth are performed according to the Manhattan distance,the distance threshold,and the distribution of pixel points around the breakpoint between the breakpoints.Finally,an edge is formed to complete the image segmentation.Through experiments,we can see that this algorithm has greater advantages than other segmentation algorithms in image segmentation details,segmentation efficiency.Secondly,it realizes the alignment of color and depth image,and RGB color image segmentation area mark and depth image segmentation.Alignment of RGB-D data is essentially the translation and rotation of space coordinates.In this paper,two-pass two-sided scanning method is used to mark the result of color image segmentation,and the result of labeling is mapped to the already-aligned depth image.The corresponding region of the depth image is extracted,and the larger areas are generated to generate point cloud datas.Finally,the plane extraction is realized based on the combination of color image and depth image.In this paper,the random sampling coincidence and least squares algorithm are used to fit the three-dimensional point cloud to find the larger planes part of the region.Through the greedy projection algorithm,the plane point cloud is triangulated,and the inverse mapping method of bilinear interpolation is used to map the results of plane triangulation and finally complete the reconstruction.In this paper,the indoor scene ofrgb-d database is reconstructed and the results of reconstruction are obtained.
Keywords/Search Tags:3D reconstruction, RGB color image segmentation, Color-depth data alignment, Plane fitting
PDF Full Text Request
Related items