Study On Three Dimensional Reconstruction Algorithm Based On Kruppa Equation

Computer vision is a comprehensive discipline which mainly includes computer science and engineering, pattern recognition, image processing and understanding, applied mathematics and statistics and other subjects. The main content of it researches is how to make a computer or related equipment to simulate human vision. That is to say how to make computers and other equipments to have the functions which human have, for example, to identify or track or measure a target object. The main task of the computer vision is to obtain the corresponding three-dimensional information by processing the two-dimensional image or video segment, and that is the main task of human vision.The contents of this paper studies is an important issue in the computer vision field, which is three-dimensional (3D) reconstruction. The main study of 3D reconstruction is how the 3D information of the target object in 3D space obtained from the 2D information, and this is very important to reconstruction the 3D model of the target object. Now 3D reconstruction has been successfully used in many fields. 3D reconstruction technique includes the extraction of image features, image matching, camera calibration, view geometry theory, 3D modeling and so on.This paper researches the 3D reconstruction algorithm which based on kruppa equation, and puts forward the theory of the second match to improve the matching precision of feature points and obtain the fundamental matrix F and the poles e′which have high robustness. And at last it enhances the robustness of the camera's intrinsic parameters obtained from kruppa equation to get accurate 3D information.Specifically: joined a search window on the sift algorithm for the initial match to reduce the amount of tedious calculations. It can increase the matching rate under the same matching accuracy. Then combined the genetic algorithm and the random sample consensus algorithm for the second match to eliminate false matches effectively and greater the matching accuracy. All of this provides the basis for camera calibration by kruppa equation with high robustness.At last, this paper designed and completed the whole 3D reconstruction system, and given a detailed elaboration to the various parts of the system. The experimental results show that the system is feasible and it can extract the 3D information accurately from the target object, and then improve the accuracy of 3D reconstruction.