Research On Key Technology Of 3D Reconstruction Of Human Body Model Based On SFM Algorithm

In recent years,with the rapid development of computer technology,three dimensional reconstruction has gradually entered people's lives,and has received more and more attention.The three-dimensional reconstruction of the human body model is one of the very important research topic,but also a challenging field of engineering research.However,due to the increasing demand for computers to automatically acquire three-dimensional information of the surrounding environment,the practical application requirements for three-dimensional reconstruction technology are also increasing.Therefore,how to quickly and efficiently reconstruct a high-precision human body 3D model according to specific applications has become an important research topic in the field of computer vision.In view of this practical application problem,from the perspective of structure from motion(SFM)algorithm in this paper,the three dimensional reconstruction technology based on human body model is studied in detail.IN this paper,a study for a number of key technologies in the human body three-dimensional reconstruction process,the specific content is as follows:First of all,the imaging principle of the camera is studied and the camera is calibrated.The internal and external parameters of the camera are calculated by analysis,and the distortion is corrected by using the camera parameters;Secondly,the characteristics of the acquired image,using the matching method based on scale invariant feature transform the image feature points extracted and matched.A modified RANSAC algorithm is proposed to detect and eliminate the matching points of the error,so that the matching points obtained are relatively accurate and reliable.Thereby a point in the physical space corresponds to a projection point in a different image.The establishment of the correspondence between the pairs of images is completed,that is,the imaging points of the same physical space point in two different images are in one-to-one correspondence.This method improves the SIFT algorithm by 25%,increases the accuracy by 2% compared with the traditional RANCAC algorithm,and greatly reduces the rejection of correct matching points.Thirdly,the spatial point three-dimensional information is reconstructed by the SFM algorithm,and the camera projection matrix is obtained and the three-dimensional coordinates of the spatial point are solved in tandem with the reliable matching points obtained in the previous one to obtain a sparse three-dimensional point cloud.Finally,the front sparse point cloud is further generated into a dense point cloud by using the CMVS/PMVS algorithm.For the generated dense point cloud,the PCL library is used to filter the point cloud to further reduce the number of noise points and improve the accuracy.And a Poisson surface reconstruction is performed to finally obtain a human body model.At the same time,the reconstructed results are compared with the reconstructed results of the traditional methods,and the reconstruction accuracy of the two results is evaluated.The comparison results show that the reconstruction result of the optimized method is better and more accurate than the traditional algorithm.