3D Object Reconstruction Research Based On Raster Projection

Computer vision technology is a comprehensive technology. With the rapid development of computer’s computing, computer vision technology is influencing and changing human life increasingly, one of the most prominent is the 3D technology. 3D technology affects all aspects of human life, from the 3D games, animation, 3Dmovie to computer aided manufacture, 3D printing, medical imaging, automatic driving, etc. all of them reflect the application of 3D technology. The three dimensional reconstruction as a key technology of 3D technology, has been praised by the attention of people. In this paper, we research the optical three dimensional reconstruction algorithm deeply, to achieve a stable and efficient optical 3d scanner system.On this basis, in order to achieve panoramic reconstruction of 360 degree, we study motion visual. Then we completed the trajectory drawing based on motion vision, and provided an important basis of theory and practice for panoramic reconstruction of 360 degree.In this paper, we used camera calibrate method from Zhang, and used black and white checkerboard as reference of camera calibration, and independent designed projection checkerboard pattern needed in the projector calibration. using OpenCV open-source computer vision library and MATLAB to realize the calibration of the camera and calibration of the projector. After completion of device calibration, this paper introduced the advantages and disadvantages of various coded structured light, and emphatically introduced the gray coded structured process of structure light, finally through computer generated independently gray code images based on time-multiplexing. After getting gray code projection, we can accord to the device calibration parameters and structure of gray code decoding strategy of light to complete 3D reconstruction on the surface of the object.In this paper, we design two methods in trajectory drawing: one is the monocular camera trajectory drawing, the other is binocular camera trajectory drawing. In the process of monocular camera trajectory drawing, firstly, the camera collect image sequence in the scene. Secondly, detecting the characteristic points in the image sequence based on SIFT feature descriptor. Thirdly, using RANSAC algorithm to delete error matching points. Finally, completing camera pose estimation according to the geometry of the pole line constraint. Binocular camera trajectory drawing process and monocular camera trajectory drawing are similar, only difference in the camera pose estimation. This paper experiment monocular camera through the real outdoor scene and binocular camera tracks trajectory drawing through simulation indoor scene drawn by 3Dmax.