3D Point Cloud Compression And Webglbased Visualization Research

With the increasing of space data acquisition method, the 3D laser scanning technology to obtain geometric information on the surface of the object has gained a rapid development. Laser radar is an optical remote sensing technology, which uses the laser to intensively sample an object. Thus it produces high precision x, y, z values(maybe also color value), that is the point cloud data. And the acquired point cloud data often stored by a variety of file formats, such as txt, pcd, las. All these data format only store uncompressed point cloud data, and requires a particularly large storage space. So it's very necessary and important to compress the point cloud data.Then, the achieved point cloud data must do visualization processing, so that researchers or users can view the data. The common practice is to process the point cloud data and create a grid model. But it takes a long time and can't directly render the origin point cloud data. It also needs to install Open GL graphics libraries, which is difficult for ordinary user. Delightfully, Web GL does not require a browser to install additional plug-ins. So it is a very meaningful thing to use Web GL in browser to implement visualization of the point cloud data.This dissertation introduces the concept of data compression and 3D visualization, including arithmetic coding, three-dimensional transformation and summarizes the currently used multi-resolution LOD technology. Then we improve the compression algorithms, point cloud organizations and visualization algorithm for point cloud data. The main work is as follows:(1) Study and improve the method to compress point cloud data. In view of the traditional general compression algorithm, it only use dictionary methods and statistical methods and arithmetic coding, without considering point cloud coordinate relations. This dissertation adopted the predictive coding based on point cloud coordinate relations, combination of variable length coding and adaptive arithmetic coding algorithm to solve this problem.(2) Study and improve the octree structure of point model, make its nodes sorted according to their maximum visibility distance, thus the linear point octree model is formed. So it can take advantage of the GPU high-speed rendering ability. In view of the shortcomings of visualization method based on RAM, this dissertation design and realize a double octree structure and out-of-core rendering method. The construction method of the structure doesn't need long time of post-processing, and also doesn't need point cloud to have a specific sampling density and normal vector of the points.(3) Realize the visualization of the point cloud based on the browser. The visualization is based on rendering the double octree structure of point cloud data on the web. The main idea of the rendering method is that only the point cloud data in the range of the visual cone can be rendered, and only at a certain level of detail. We also show how to adapt the size of the point cloud to avoid the hole and how to hide piont density of different levels of detail.Finally, we design and realize the 3D point cloud compression and visualization system based on browser, and validate the feasibility and effectiveness of the proposed point cloud compression algorithm and visualization methods.