Reconstruction On 3D Human Body Based On Kinect

3D human modeling technology is one of the important research contents in the field of computer graphics and image,computer vision and virtual reality technology.3D human modeling is usually scanned by a three-dimensional scanner to create a unique threedimensional human body model,and further apply the human body model to the virtual field.In this paper,we use the characteristics of Kinect depth camera,such as easy operation and low cost,three-dimensional human reconstruction method was proposed in this paper by uses the Kinect device to rebuild the 3D human model first,and then integrate the SCAPE 3D human body model database,and bring out the 3D human model reconstruction by finding similar body 3D human models,at last registration and mesh deformation.The experimental results show that the method is simple and easy to operate,does not require specific scenes and any human calibration,has a shorter reconstruction time,the reconstructed 3d human model is very close to the real data of the scan human body in height and shoulder width and so on,and can reconstruct high accuracy 3D human models,can satisfy the application in many fields.The main parts of this paper are as follows:(1)Scanning the upper body and the lower body of the body for a week with Kinect equipment,and reconstruction of 3D human body model by point cloud splicing.Because of the surface data of the scanned model with the low precision,noise,redundancy and so on,need to preprocess the acquired data first.(2)Combination of body somatotype,classification of models in SCAPE 3D body mesh database based on clustering of A,X,and V types is proposed in this paper,within the allowable error range of the body shape index,finding a similar standard model to the scanned human body model in the SCAPE 3D body mesh database is proposed;(3)ICP registration of the standard model and the scan model to find the corresponding point pair of the corresponding feature,experimental proof,by finding models of similar body types,the errors in model registration and finding corresponding point pairs are greatly reduced,this will lay a good foundation for the reconstruction of 3D human model.(4)Finally,a method for iterative deformation of standard model using grid deformation migration algorithm is proposed.The method is to give a corresponding pair of points between a source grid and a target grid,establish a mapping relationship between models,in order to achieve the constraint deformation of the source mesh to the target mesh.The approach is general and does not require the source and target to share the same number of vertices or triangles,or to have identical connectivity,and the correspondence between the source and the target solves an optimization problem,which consistently apply the transformations to the target shape.Experimental proof,this method can reconstruct a three-dimensional human model with high accuracy and user characteristics.