Finke River: Results
The integration of multiresolution DEM's are beneficial for a variety of reasons. Multiresolution DEM's are necessary to provide accurate coverage on a regional basis and it is a useful methods for removing errors related to different sensors. In addition, development of a multiresolution DEM provides information at multiple scales of importance, thus reducing storage requirements.
Qualitative comparison of three multiresolution DEM's.
The three DEM's have been combined by the application of the Dynamic Adaptive Grid Hierarchy (DAGH) data structure [2]. DAGH was originally designed for solving partial differential equations (PDE) with the capability of adaptive mesh refinement. Whenever local errors from the integration of the PDE on the solution grid exceed the predetermined threshold, it adaptively refines the area to reiterate the solution process locally. The DAGH system maps N-dimensional subspaces of different resolution into a 1-dimensional space using space-filling curves which achieve locality preserving mappings. By maintaining records of the relative coordinates and the instantaneous resolution of subgrids, DAGH allows manipulation of data across different levels while preserving resolution locally.
This study was the first investigation of the capability of the data structure for multiresolution integration of remotely sensed data. Herein, DAGH is applied in an inverse manner in the sense that the locations of refinement are predetermined. Once a locality of multiple resolution data is defined, its structure is preserved for subsequent analysis. Variable resolution data are represented with localized subgrids at different resolutions. The Adaptive Mesh Refinement - Hierarchical Data Format (AMR-HDF) is adopted to store and retrieve a series of multi-dimensional, multiple-sized, locally variable resolution data in one file. There is very little computational overhead required for describing the characteristics of individual data segments.
The whole process of variable resolution data representation can be summarized as:
- Image registration and Geocoding
- Removal of planar error
- Import data into DAGH
- Set flags at the unreliable pixels at all Levels
- Create list of Bounding
Boxes (BBox) separating reliable pixels from unreliable pixels
- Propagate the BBox to coarser Level
- Define the BBoxes on the coarse level, within the propagated BBox and outside the outer limit of finer level data boundary
- Update BBox list of reliable pixels cluster.
- Visualization or Storage in AMR-HDF
|
Last Modified: Wed Apr 14, 1999
CSR/TSGC Team Web
|
