Research On High Resolution 3D Human Model Reconstruction System

The printing quality of 3d portrait in 3d printing photographic house, depend on the 3d scan of the three dimensional human body model identification. However, the traditional 3D body scanner is expensive, complicated manipulation and other reasons, 3D portraits of high printing costs, timeconsuming and less accurate printing. In view of these shortcomings, this paper presents two simple and efficient 3D model reconstruction system. It is based on the depth of the scanner system and the reconstruction of three-dimensional model of the human body three-dimensional model reconstruction system which based on a sequence of images.The first method is proposed based on the deep scanner reconstruct the high degree of recognition three-dimensional human body model. The method combines with the advantages of three groups of different types of the depth scanner, collaboration, respectively to obtain high precision characters facial features and hair detail point cloud data, upper body and the body surface contour point cloud data. Then, the captured three sets of point cloud data will be aligned, replaced, registered by introducing feature points and iterated closest point. The registered of non-topological relations of point cloud data for surface reconstruction can get high precision of 3D human body model. The method of scanning time is shorter,at a relatively low cost to build the 3d portrait with high identification model.3D human model image reconstruction system based on image acquisition, the first step is to use a designed hardware system to obtain a clear sequence of the target object image. The second step for image pre-processing, to remove the jitter, noise, light and other interference image acq-uisition, and in order to reduce the additional time it takes to rebuild, so this research deletes the background target object.The third step is to use SIFT feature point extraction,matching algorithms and multi-view 3D model reconstruction method to reconstruct the audience’s three-dimensional sparse point cloud from pretreatment image and use ICP algorithm reduces the error of reconstruction model. The fourth step in this study use dense generation algorithm PMVS to make the sparse point cloud which generate in step 3 denser. The fifth step this study process the generate dense point cloud by point cloud noise reduction pretreatment and surface reconstruction algorithm processing, eventually get the three-dimensional mannequin in a high degree of recognition. Sequence image reconstruction method of three-dimensional human body model proposed in this study, only need to use a digital camera, you can build a high degree of recognition for 3D point cloud 3D studio use, promotion of high value.