Research On 3D Reconstruction Based On Sequence Images

With the development of information processing technology and the emergence of computer vision,three-dimensional modeling can be used to process the three-dimensional information of objects.The comprehensive and intuitive information experience of the three-dimensional models bring people growing demand for the three-dimensional model,and also give birth to the development of many modeling techniques.Among the three-dimensional models,the image-based modeling technology has the unique advantages,such as low cost,simple and systematic operation,and reconstructing the sense of authenticity.It is widely used in various fields,such as military,agriculture,medicine and so on.It has increasingly become the focus of current social research.This article starts with the reconstruction of the surface model of the target object and conducts corresponding research on each module in the reconstruction system.In the design of each module in this reconstruction system,we first focusd on classical feature extraction algorithms,and proposed an edge feature-based template feature point extraction algorithm,which correlates image edge information with template angle information and extracts images.The feature points were described in binary code.Zhang's calibration method was used to calibrate the camera and to solve the projection matrix transformed between image points and space points.In this paper,a new method for solving spatial points based on nearest neighbor image matching was proposed.A small number of matching feature points were used to complete the constraint on the spatial points,which can greatly reduce the computation dimension.After solving the space point coordinates,the minimum optimization algorithm for sum of distance squares was conducted to complete the precise spatial location of the calculated space points and to get the space point cloud data of the whole object surface.For the acquired point cloud data,a grid method was used to topologically connect two-dimensional plane point with the point cloud in the back projection space,and to complete the transformation from point to surface for the three dimensional point cloud data.Finally,Texture mapping and rendering of the obtained model surface was achieved.A three dimensional surface model with texture and color rendering was obtained at last.