3d Scan Information To Obtain The Depth Of The Image Registration Algorithm Design And Development

Cultural relics are precious heritage of civilization of human being since we came to the world thousands of years ago. They have undoubted value for research and cannot be reproduced ever. This makes the protection for them a continuous hard work. However, the threatening not only comes from human, but also some natural hazard such as weathering. Even those relics kept in museums could not be fully protected from the air around. The digitizing technology is a method which can turn the cultural relics from a real world object to the information represented in computers. Thus, they can be kept forever. 3D scanning is one of the main digitizing methods. For a long time, the key technologies of 3D scanning were mastered only by a few foreign companies. The devices are too expensive to be adapted in the most common digitizing projects in China. This gives us a reason to develop a cheaper but relatively high-resolution 3D scanner. The work in this paper contributes to the post processing part of the whole system. That is to develop the registration environment for the range images acquired.First, we analyzed the traits of the structured light scanning system and the features of our target cultural relics. After comparing the widely used registration methods, we came to a three-stage registration pipeline. At last, a global registration method was discussed.Based on the analysis, the common ICP algorithm can be adapted to our system, but it's not an efficient one. This due to the linear convergence and the corresponding points calculation in the algorithm. These bottlenecks decrease the performance dramatically. As another successful method, the squared distance function method solved the problem by choosing another target function and putting the highest calculation work in a preprocess step. However, it has drawbacks too. In order to get rid of the corresponding points calculation step, the value of the squared distance function has to be fitted in every space partitions. This adds extra error to the algorithm. That is why we came to a three-stage pipeline.After the pipeline was fully discussed, this paper compared the pipeline algorithm with the accelerated ICP algorithm from the convergence aspect and the successful initial configuration aspect. Though it gave a better result, there still exist some deficient parts that should be considered in the future.