Online Human 3D Modeling Method Based On Skeleton

3D human modeling is an important issue in computer vision and computer graphics.It has been widely applied in biological medicine,virtual fitting,film and television animation and other fields.Currently there are two categories for 3D human modeling methods,including direct modeling from depth information and data-driven 3D human modeling.The former is limited by the precision of the equipment and thus requires high quality acquisition environment and human pose.The latter depends on the features of the template data set,which has high complexity and poor generality.In addition,in real applications human pose is often dynamic and multi-scale,and it is difficult to directly apply the existing methods to obtain appropriate results.Focusing on dynamic and multi-scale challenges in the scene,we propose an online human 3D modeling framework for online close-range(1.2m-3.0m)and long-range(3.0m-8.0m)scenes based on skeleton.Specifically,we leverage the depth camera to obtain the skeleton and reconstruct the 3D model in real-time in the close-range scene,and propose a texture mapping method which is further applied to the operating room scene.Aiming at the long-range scene,we present a method of using binocular camera to extract the skeleton to reconstruct the human model.We can efficiently reconstruct the 3D human model by exploring the mapping correlation between binocular skeletons.The main contributions of this paper are as follows:1)Proposing an integrated framework for online human 3D modeling based on skeleton,which can be used for 3d human modeling of single and multiple people in different distance scenes and applied to operating room and other scenes.2)Proposing a binocular vision-based extraction method for 3D human skeleton,which realizes efficient extraction based on image information at a distance scale.3)Accomplishing a real-time reconstruction based on an arbitrary skeleton and presenting an efficient texture mapping method.