Using Data Indexing For Remote Visualization Of Point Cloud Data
Using Data Indexing For Remote Visualization Of Point Cloud Data |
Abstract
For decades, point cloud-based datasets have been critical to a number of research communities, including graphics, robotics, and CAD. In many ways, points better represent many data sources than do triangulations. Take a laser scan for example. The laser collects a series of point that represent a surface. Then, other algorithms take over to produce surfaces, such as triangulations, NURBS, etc. These surfaces must make assumptions about continuity or connectedness that may or may not be valid. Point clouds also have the advantage of being more naturally representable in multiple resolutions. For example, clusters of points can be culled to achieve a desired resolution. Visualizing points on remote devices remains somewhat problematic to this day. Very large datasets have the problem of needing to be transmitted, have any data structures built locally, and rendered on the local machine. Even with advances in low-power computing, such as tablets, the trend in computing is to compute as little as possible locally and rely on cloud-based systems to store and process large data. For point-based data this approach is highly relevant. In fact, data should only be transmitted at a resolution visible on the output display. What’s the point in rendering billions of points, when the display only has 2 million pixels?
Downloads
Citation
Paul Rosen, and Les Piegl. Using Data Indexing For Remote Visualization Of Point Cloud Data. Computer-Aided Design and Applications, 2017.
Bibtex
@article{rosen2017using, title = {Using Data Indexing for Remote Visualization of Point Cloud Data}, author = {Rosen, Paul and Piegl, Les}, journal = {Computer-Aided Design and Applications}, volume = {14}, pages = {789--795}, year = {2017}, keywords = {point cloud, visualization, data indexing, data structures, remote visualization}, note = {textit{Presented at the CAD Conference and Exhibition 2016.}}, abstract = {For decades, point cloud-based datasets have been critical to a number of research communities, including graphics, robotics, and CAD. In many ways, points better represent many data sources than do triangulations. Take a laser scan for example. The laser collects a series of point that represent a surface. Then, other algorithms take over to produce surfaces, such as triangulations, NURBS, etc. These surfaces must make assumptions about continuity or connectedness that may or may not be valid. Point clouds also have the advantage of being more naturally representable in multiple resolutions. For example, clusters of points can be culled to achieve a desired resolution. Visualizing points on remote devices remains somewhat problematic to this day. Very large datasets have the problem of needing to be transmitted, have any data structures built locally, and rendered on the local machine. Even with advances in low-power computing, such as tablets, the trend in computing is to compute as little as possible locally and rely on cloud-based systems to store and process large data. For point-based data this approach is highly relevant. In fact, data should only be transmitted at a resolution visible on the output display. What's the point in rendering billions of points, when the display only has 2 million pixels?} }