Research On 3D Model Reconstruction Technology Based On Terrestrial Laser Scanner

Terrestrial 3D laser scanning technology is a new technology of surveying and mapping, which can collect the massive point cloud data of the target’s surface with non-contact, high accuracy and high speed. These point cloud can reflect the target’s surface status truly and accurately, providing a new idea for the digital 3D model reconstruction of material object. Therefore terrestrial 3D laser scanning technology is widely used in digital city, cultural relics protection, archaeological surveying and mapping, and many other fields. The hot and difficult problems of the current research are how to deal with the massive point cloud data, obtain the high quality point cloud, and improve the efficiency of digital model reconstruction.Firstly, this paper introduces the operation principle of terrestrial 3D laser scanner, and then it studies on the pretreatment of point cloud using several examples, including visualization, point cloud registration, point cloud denoising, point cloud simplification. For the two kinds of objects with regular and irregular shapes, this paper proposes two sets of detailed 3D model reconstruction scheme: irregular triangular grid method and feature line extraction method. It achieves from the point cloud to the reconstruction of digital 3D model. Finally, this paper introduces the multi-view image model reconstruction technology, combining with point cloud data, studying on the fusion of 3D laser scanning and multi-view image reconstruction technology, and in the end the paper proposes a new fast fine model reconstruction idea. The main research result of this paper is the following:(1)The work process of 3D laser scanning technology is summarized, and the required hardware and software equipment are aggregated and classified. The paper also discusses outdoor process of point cloud collection and the matters needing attention.(2)Taking Guting point cloud as an example, the paper used four kinds of point cloud registration methods to carry out the experimental study. The accuracy, efficiency and applicability of the four registration methods are summarized and compared, and finding that the target registration method is fast and the precision can reach the mm level. At the same time, the paper analyzes the three main sources of the point cloud noise point and studied on the methods of point cloud denoising: the straight through denoising, voxel grid method, Statistical Outlier Removal and Geomagic soft denoising. Using the research results to Guting point cloud denoising, the high quality point cloud data was obtained in the end.(3)Aiming at the different structure shape of the scanning objects, combining with two examples of the GuTing and the modern building, the paper proposes two sets of 3D model reconstruction scheme: irregular triangular grid method and feature line extraction method, and completes model reconstruction. The two kinds of the modeling technology are compared and analyzed from the model precision, the model data quantity, the applicability and other aspects. The result shows that the triangular mesh modeling method is suitable for the irregular solid modeling, the average deviation of GuTing model is 1.2mm. The feature line extraction method is applicable to the regular objects model reconstruction, the maximum deviation of the building model is 1.4cm. Meanwhile, for the two different structure shapes of the 3D model, combining with 3D rendering software, this paper proposes two different texture mapping program to enhance the authenticity and appreciation of the model.(4)This paper introduces and studies the multi-view image model reconstruction technology, taking Song dynasty ancient horse as an example to complete model reconstruction experiment, and verifies the feasibility of this technology. Taking point cloud data as the combination point, in the face of some special scanning scene, this paper studies on the fusion of laser scanning technology and multi-view model reconstruction technology, and proposes a new fast fine modeling idea. Taking an archaeological site as test object, this paper completes model reconstruction and the model precision evaluation, the model error is 1.03 cm. Moreover, it is also a new idea that using multi-view image model reconstruction technology to repair the defect area of 3D laser scanning point cloud.