Evaluation of ENVI feature extraction for geological mapping in arid environments
Object-Based Land Analysis Meeting, University of Nottingham, 7-8 April 2009 Arid environments provide ideal conditions for evaluating emerging technologies for geological mapping. This is due mainly to the lack of vegetation that otherwise covers the geological materials and prevents them from dire...
| Main Authors: | , |
|---|---|
| Format: | Text |
| Language: | English |
| Published: |
2009
|
| Subjects: | |
| Online Access: | http://nora.nerc.ac.uk/id/eprint/502917/ https://nora.nerc.ac.uk/id/eprint/502917/1/RSPSoc_Tragheim_Smith_2009_OBLA_Presentation.pdf |
| Summary: | Object-Based Land Analysis Meeting, University of Nottingham, 7-8 April 2009 Arid environments provide ideal conditions for evaluating emerging technologies for geological mapping. This is due mainly to the lack of vegetation that otherwise covers the geological materials and prevents them from direct observation. The most common approach to geological mapping by remote sensing involves: (i) calibration of image DN values to radiance; (ii) conversion to spectral reflectance and (iii) spectral analysis, commonly by Principal Component techniques or by comparing spectral signatures of image pixels with spectral libraries of known geological materials. Although the PC technique provides good spectral separation it is often difficult to generate suitable boundaries between the spectrally diverse units as they often become disjointed with the introduction of errors on application of region smoothing algorithms. Spectral signature analysis is often more temperamental where slight variations in spectral signatures from the reference libraries can result in small inliers of unclassified material that are clearly from the same geological unit. The ENVI Feature Extraction module enables segmentation of multi-band images into spectrally diverse components by including both spectral characteristics and surface texture in the analysis. There is also an option to include topographic information. Multiband spectral images have been obtained for several geological settings, mainly from arid environments: (i) Morocco, Saudi Arabia & Afghanistan, in a complex setting of sedimentary and metamorphic geology with igneous intrusions; (ii) Algeria, over a sedimentary sequence; (iii) Ethiopia, over rifting volcanic terrain, and (iv) Tristan-da-Cunha, a volcanic island in the South Atlantic Ocean. ASTER images for each of these areas have been ortho-rectified and atmospheric effects removed from the radiometrically calibrated data using the FLAASH algorithm, with the resultant reflectance images used to evaluate the ENVI Feature ... |
|---|