Research On Range Images Registration Technology And Its Application Of The Digital Display On Cultural Heritage Sites

With three-dimensional laser scanning technology becoming more sophisticated, model reconstruction technology based on range image reconstruction technology has become a hot area of the computer graphics. The key issue of reconstruction of big scene models is how to quickly realize the accurate registration of range image. This paper focus on the research of registration technology of range image, and the result is applied to 3d reconstruction of ancient site. The main work of this paper is summarized as follows:1) In order to solve large amount of data and noise problems of large scene range image, the model should be preprocessed before range image registration. First, the paper uses KNN algorithm to remove noises of models. Then, it simplifies models using a mixed sample algorithm which included random sampling and curvature sampling. And it provides an amount of robustness and smaller data models for registration technology.2) This paper proposes an improved registration algorithm of multiple feature regions by using planar feature. This algorithm fits the model into several planes via the plane segmentation of Stamos'. Then, it composes a feature area by a plane and two other adjacent planes. Then use multiple sets of feature areas for rough registration. This algorithm is applied to the digital display of Zhang Shi-Han tombs, and experiments show that this algorithm has high precision, short time-consuming.3) In order to improve accuracy of rang images'rough registration and reduce the time-consuming of the ICP algorithm, this paper proposes an improved ICP algorithm based on blocked method. First, after rough registration, this algorithm divided the overlap region of range images into small blocks. Then, the algorithm selects multiple blocks, and each block use ICP algorithm to get the transformation parameters. At last, the optimal solution is the transformation parameter which has the shortest distance from the other parameters. Experiments show that this algorithm based on ICP algorithm effectively raise the efficient of accurate registration. 4) For the problems of existing overlap redundant data after range images registration, and details are not rich enough in range images of the big scene. This paper uses image fusion technology to eliminate the overlap redundant data after range images, and enrich the details in range images of the big scene.Base on the algorithm proposed in this paper, considering the features of range images registration of the big scene, this paper designs and implements the system of registration on range images. The system has implemented pretreatment, registration, fusion of range images.