Surface Reconstruction Of 3D Data And Its Applications To Human Animation And Anthropometry

Laser 3D digitization technique includes obtaining the surface point data of 3D real object with laser methods, implementing data pretreatment, creating and interavtively editing 3D surface models in computer and other correlative research fields. The technique integrates optics, mechanics, electronics and algorithms and commands increasing applications, such as medical diorthosis, reverse engineering and rapid prototyping manufacture, ergonomics and garment design and manufacture, 3D animation and virtual reality etc.With 3D data set, especially human body contour data obtained from Laser 3D-scanning system as research objects the data pre-processing is first carried out, and then suface mesh reconstruction based on local-flatness property and projection is studied in detail. After surface mesh editing and optimization, including mesh smoothing, mesh hole filling and mesh simplification, the algorithm for skeleton extraction from 3D mesh models using radial basis functions and gradient rapid descent is presented, and the human animation on the basis of joint-skeleton links is performed. As the last key content of the subject human key sizes auto-measurement is implemented.The main innovations are as follows:(1) Pre-processing human body contour point data set. The method based on distance fuzzy logic judgement to delete noise and stray points from point cloud is presented and the algorithm for auto-adjust human orientation and the human auto-partition based on human feature points are introduced. Iterative corresponding points (ICP) algorithm is used to registrate point cloud from multi-sensor.(2) A 3D Data surface meshes reconstruction algorithm based on projection that makes use of local-flatness property of 3D surface is proposed. This method can recognize the holes and boundaries automatically, and by taking full advantage of characteristics of human contour data reconstruction time can be shortened. After surface reconstruction is completed, the mesh editing and optimization methods are discussed, including mesh-smoothing algorithm based on spring model and lower-pass filter, hole-filling algorithm using grey prediction GM(1,1), and mesh simplification algorithm based on mesh cell contraction under the rule of quadric error metric.(3) The 3D skeleton extraction from surface mesh models and human model animation based on skeleton are implemented. Firstly apply radial basis functions to implicit represent 3D mesh models, and by using gradient rapid descent method 3D skeletons can be computed and obtained. Locate the joints on the human body skeleton by manual and set the properties of each joint and so build up joint-skeleton links. After binding the skin vertices to corresponding joints by flexible model the human animation is implemented.(4) Human body key sizes automatic measurement is implemented. First use animation algorithm to change body posture to stand posture and squat posture in order to satisfy measurement requirements. Algorithms for locating feature points based on human suface mesh model are introduced. After 9 surface feature points and 4 feature joints are found out, the 38 key sizes, which are very useful in ergonomics and garment design and manufacture fields are measured and the results for male and female models are acquired respectively.