The 3D Reconstruction Algorithm Research And Technology Based On Binocular Camera

Recently, with the development of technology, machine vision is widely applied in human daily life. This technology is becoming an essential part of Biomedical Technology, Virtual Reality Technology and Industrial Inspection. Even in the field of space remote sensing detection, machine vision is an obbligato applied technology as well. Binocular stereo visual method is the method which simulates human eyes to detect the real objects and rebuild them in 3D model in the terminal.In order to restructure a 3D object, a novel method would be presented in this paper. Initially, two viewpoints would be simulated by binocular camera, and then machine would take sample form one subject through those two viewpoints. The images which sample from two viewpoints are different; therefore, there are 5 steps would be executed to obtain 3D vision and depth information of the subject. Those 5 steps are obtaining image information through binocular camera, building imaging model, camera calibration, feature extraction, binocular stereo matching, and three-dimensional reconstruction. In this article, author analyzed the advantage and disadvantage of the classic image model, and then presented distortion considered image model which is based on the classic pinhole model. After evaluating various different kinds of algorithms, in the step of camera calibration, Tsai two-step calibration algorithm would be used in the experiment, And finally, the parameters of binocular camera could be obtained through the experiment.In the step of binocular stereo matching, according to 3D imaging principle, a polar constraint, which based on classic SIFT algorithm, could be introduced into SIFT algorithm. Therefore, an advanced SIFT feature matching algorithm, named polar constraint introduced SIFT feature matching algorithm, would be presented in this paper. After comparing SIFT feature matching algorithm and polar constraint introduced SIFT feature matching algorithm through 3D matching experiment. Polar constraint introduced SIFT feature matching algorithm could be proved that it is an advanced algorithm for 3D matching. Finally, 3D reconstruction would be executed by using the feature points which are generated by 3D matching; the Parallax image and 3D reconstruction image would be generated in terminal.