Theory and Application of an Eye-Point Dependent Metric for Multiresolution Terrain Models
Abstract
We further developed our interactive visualization for terrain change detection and investigated a broader range of data structures for probabilistic surface modeling. Detecting changes in LIDAR scans creates an overwhelming number of change models. These models should be grouped into meaningful higher level events. For example, a set of 100 removed trees grouped into a "deforestation" event. These events would provide a semantic index into the set of change models. Both computer and human analyses are needed to identify events. To help the user identify change objects, we integrate road and building permit data into the visualization. To allow the user to see change trends at any scale, we develop a three-tier, level-of-detail rendering algorithm. Finally, to show the distribution of change model size and to filter the displayed models, we develop an interactive heat-map that plots shape area versus height versus occurrence frequency. Our system does not use the permission grid data structure used in our prior work. We identified several problems with using this structure for change detection. We are examining non-computer graphics probabilistic surface models that are based on continuous, rather than binary, scalar fields. We are comparing these works? field datum and probabilistic metrics.
Document Details
- Document Type
- Technical Report
- Publication Date
- Oct 20, 2008
- Accession Number
- ADA499813
Entities
People
- William Ribarsky
- Zachary Wartell
Organizations
- University of North Carolina at Charlotte