Stereovision 3D Reconstruction Technology Research And Achieve

The technology of three-dimensional reconstruction based on the stereo vision is an important direction of the computer vision. It can help realizing three-dimensional untouched measurement and rapid three-dimensional modeling. It has been played an increasingly important role in robot vision, satellite navigation and mapping, military reconnaissance, medical imaging and industrial inspection, owing to high efficiency, low cost and easy to carry.In this paper, the system of three-dimensional reconstruction based on stereo vision has been researched. The main idea is to calculate coordinates of the target based on image coordinates getting by image matching, associate with camera parameter matrices, and then to realize three-dimensional reconstruction. Research works of this thesis are as follows:Kruppa equation is solved for auto camera calibration, which is used to calculate the camera parameter matrices. To improve the accuracy and speed of image matching, an algorithm of Scale Invariant Feature Transform (SIFT) feature extraction combined with Normalized Cross Correlation (NCC) matching is applied. Because sparse matching can not pick up sufficient matching points, the dense matching based on regional growth is put forward to collect seed points for regional growth, so that enough matching points have been gotten. To avoid error matching points, a method of doing the secondary elimination is proposed in this paper, which is different from the traditional method and reduces the harmful effects of mismatched points better.Based on a lot of study on algorithms, a software system of stereo vision three-dimensional reconstruction has been designed. It can regain three-dimensional appearance of objects according to two images of different angles, followed by a series of operations. The system can reconstruct the images, which meet some certain conditions.