The Key Technologies Of 3D Scanner Based On Multiple View Geometry

The project designed in this article researched and designed a spatial digitizer system based on the multiple-view geometric principle. The spatial digitizer comprises 54 tables of CMOS based on the Linux operation system which may operate indepently, and the center computer used for the fusion calculation and the displaying the human-computer interaction interface. The CMOS may be connected with the center computer via the Ethernet, and also may control the feature points supplement system and the illuminating system at the same time. The center computer may send the instruction on shooting to the CMOS, and the CMOS will perform two times of photograph once receiving the instruction, and send the photos back to the center comouter. Then the center computer will refine the feature points of every photos and match them one for another, and may calculate the three-dimensional coordinate of the matched points and form the 3 d point clouds, then perform the reconstruction of Poisson surface and the chartlet reconstruction. The spatial digitizer designed by this article applies to the scanning of the big objects, such as human body or the sculpture by means of the way of multiple-view reconstruction. Especially, since the multiple-view technology may create the three-dimensional model based on the photo of single width, it applies to the threedimensional reconstruction for the objects of which the forms are moving continuously such as human body and alive animal.A hardware system is added into the project of this article, which has the advantage of opening the source, convenience for construction and reasonable organization structure, and of which the designation of use interface allows the users to use the system for the 3 d image fusion more convenient. At the same time, the flow path of collecting the data and the fusion of the system are relatively independent, which leads to the more easy later stage extension. At the same time, coordination with the reasonable calibration algorithm for CMOS, separate the calibration process and the fusion process of the three-dimensional model, thus may reduces the time needed for synthesizing the model.Directing to the features of the multiple-view fusion algorithmic, the author of this article constructed the feature points supplement system which enhanced the amount and the precision of the point clouds generated during the fusion greatly. The reasonable use coordinating with UDP and TCP protocol guaranteed the success ratio of taking photo by the CMOS in synchronization and photo transferring. This article introduced the functional framework of this system, and performed rounded analysis and designation on the key technology of this system based on the three-dimensional fusion algorithm analysis according to the multiple-view geometric principle, and discussed the main soft program of this system and the accomplishment of the hardware structure.