Research On Key Technologies Of 3D Reconstruction Of Human Body Via Finite Element Method

3D reconstruction of human body is that a complete digitized expression of human body is established through using the local or global bodily observation data.In recent years,with the commercialization popularization of RGBD camera devices and the development of the non-rigid 3D reconstruction technology,more and more researchers are focus on the reconstructing a complete dynamic human shape under the condition of limited RGBD observation data.However,there is often the strong non-rigid deformation compared with the bodily other regions around the joints during human motion.Some techniques for reconstructing a full human body shape often smoothing extremely and deform unnaturally around the joints.Aiming at solving the above problem of the existing technologies,we make a lot of researchs about the motion reconstruction of the region around the joints as follows:The essence of motion reconstruction of the region around the joints is to solve its nonrigid deformation.We analyze the deformation mechanism around a joint from the perspective of motion structure,and explain the deformation distribution of different strength around the joint.For the complex deformation distribution characteristics around the joints,we innovatively introduce the Finite Element Method.The problem of solving the non-rigid deformation around a joint is transformed into the problem of solving the elastic deformation of mutiple element.Firet of all,the region around a joint is divided into multiple elements;Secondly,the elastic deformation produced by each element is analyzed according to the distribution of material inside the element and the joint rotation angle.Finally,the pose of all elements are transformed according to the elastic deformation degree of each element,so that the discrete elastic deformation of all elements is fused into continuous non-rigid deformation of the region.In addition,we establish a general wrinkle deformation model for adjusting the geometric shape of the region around the joint.The precondition of the above work is that obtaining the point cloud around joints from local or global bodily observation data.An intuitive idea is to use the current advanced human pose tracking technologies to obtain the 3D position of joints.However,the fast movement of the human body makes the joint positions often deviate from the point cloud of human region.We expands the research to solve this problem.We add the secondary locating based on human pose tracking technology,and use particle swarm optimization algorithm based on swarm intelligence to search 3D position of joints.We was conducted a series of experiments in several datasets.The experimental results verify that the research in this paper is reasonable and effective.The research has significant advantages for motion reconstruction of the region around joints under the condition of extremely sparse RGBD observation data.