Algorithms Of Virtual Human Reconstruction And Skinned Mesh Animation Based On Depth Data

With the coming of network information technology times,3D virtual human modeling techniques have been utilized extensively in the fields of computer vision, film and television production, game designing, etc. So how to obtain a more realistic 3D virtual human model, and to simulate the human body movement effectively, are the research targets and hot topics in the area of computer graphics and computer vision currently. Current 3D human body modeling methods mainly include scanning reconstruction of the static models and animation of the models. It often requires expensive scanning equipment, and the operation is complex, which usually needs professional training. For the animation production of the model, it can be skinned and bound through 3D software, which is usually time-consuming. And it can be calculated directly according to the relative position of skeleton with the skin. The error is high. These problems greatly limit the development of 3D human body modeling technologies.Our methods overcome some current flaws in the 3D human body modeling technologies, and studied the virtual human model reconstruction and skinning animation based on depth data, which included two parts. The first part was the depth data sequence of body in a standard pose and different point of view, the other part was the depth data sequence of body in different pose and a point of view. And then we translate them to point cloud, and we can execute the pre-processing of point cloud data, including point cloud denoising, triangularization, simplification,holes mending, etc. And the point cloud data can be divided into two parts, one was used for model reconstruction in a standard pose, the other was used for skinning parameter optimization; In the model reconstruction we accomplish the alignment operation of the point cloud data sequence through improved iterative closest point algorithm. And with the improved truncation Signed Distance Functions algorithm, we process the fusion of point cloud in overlapping region. Then, after fusion, we can acquire a integrate model of 3D human body through point cloud surface topology reconstruction.In the end, we use the second part data for skinning parameters optimization. According to the LBS algorithm, we established 3D human body driven model.According to the CPD algorithm, we computed the corresponding relationship between the integrate model and the partial model sequence, we design the objective function between them.In addition, the weight parameters of integrate model and the pose parameters of the point cloud sequence were optimized alternately, and were also executed iteratively until the error of objective function converged. So we can acquire the more realistic virtual body model. The experimental results show that, through the reconstruction of the model and the parameter optimization, we can acquire a more realistic virtual human model. At the same time, the skinning animation of virtual human simulates the body movement of real human effectively.