Blue tab
Green tab
Brown Tab

Preliminary results (examples)



Test Site 1. - Sokolov - Mineral and physical soil property mapping


The following includes different work that has been done on the Sokolov data.

Endmember extraction

The 2010 HyMap airborne data of Sokolov, Czech Republic, was use to derive a set of image endmembers to be used for subsequent classification. Endmembers were extracted from the reflectance imagery that was not geocorrected and had no cross-track illumination or BRDF correction applied. Of the 125 atmospherically corrected bands (ATCOR, R. Richter., 2010), 14 of those occurring near the main 1.4 and 1.9 μm atmospheric water absorption features and noisy bands longer than 2.45 μm were removed from the data leaving 110 bands. For endmember extraction no masking of the data was applied.

Sixty-five ground samples that were collected in 2010 by DLR for only a limited number of soil, clay, and iron-oxide mixtures.  The spectra were compiled into a single library and re-sampled to the 110 HyMap bands for subsequent comparison and evaluation with endmembers derived using OSP SS and SSEE SS endmember extraction tools. Both tools were adapted from OSP (Harsanyi and Chang 1994) and SSEE (Rogge et al. 2007) at DLR to run on large multi- flight-line hyperspectral surveys (Rogge et al. 2012) using a spatial-spectral implementation.


The results from OSP SS and SSEE SS endmember extraction were heavily influenced by the spectral diversity of the many urban materials within the scene. However, the SSEE SS approach was still successful in extracting many of the key geological materials known to be in the scene and matched samples collected in the field in 2010 (Rogge et al., 2012) (see following figures). These endmembers were subsequently used for generating a preliminary thematic map prior to the 2011 campaign to show the broad distribution of all land cover classes, but specifically geological materials.

endmember spectra images
SSEE SS endmember spectra derived from the 2010 HyMAP data for natural non-geological surface materials.


Endmember spectra graphs
Selection of SSEE SS endmember spectra (dotted and dashed lines) derived from the 2010 HyMAP data for natural geological surface materials compared with ASD field spectra (solid line) of samples collected in the field in 2010.
Note: Gypsum match is taken from the USGS spectral library.

Land Cover and geological material mapping

As noted above the endmembers derived from SSEE SS were subsequently used for generating a preliminary thematic map showing the broad distribution of all land cover classes, but specifically geological materials. To generate this preliminary map we use an adapted version of spectral angle mapper (SAM) developed at DLR. In the adapted method each of the reference spectra (endmembers) are compared to the image data using 4 spectral ranges to better asses specific spectral features inherent to the given materials. These spectral ranges use the full spectral range (0.4542 – 2.4538 nm), which assess the broad shape of the spectra. The other three ranges include: 0.4542 – 1.3097 nm; 1.4998 – 1.8050 nm; and 2.0099 – 2.4538 nm. The SAM results for the 4 spectral ranges are combined, resulting in a better classification of the reference materials as compared to using only 1 spectral range.

A thematic map for the full 2010 HyMAP area was generated and shows broad surface material classes grouped together (see following figures). A more detailed map of the complete 2010 image data was also generated for only the geological reference materials (see last figure). This map only classified pixels with a strong spectral match, such that a high degree of confidence was retained. Pixels with a high degree of spectral mixing were not classified in this map. However, more advanced classification methods, such as spectral mixture analysis, are being implemented to improve our ability to map geological materials that are not the dominant pixel component. This will allow for a better assessment of the distribution of important geological materials, such as iron oxides that may indicate acid mine drainage.

Click on image to see it full-sized.

Thematic map
Thematic map of the 2010 HyMAP data showing broad classes of surface materials: black – water (and highly shaded pixels); white – urban (includes clouds); dark green – forest; medium green – low standing vegetation; light green – vegetation with minor soil; coral – soil with minor vegetation; purple – coal; light yellow – Fe-rich clay soil; dark yellow – Fe-poor clay soil; cyan – kaolinite-rich soils;
red – iron oxides.
overburden dump site
Zoom of detailed geological map showing the distribution of various materials on a overburden dump site north of the active mine areas. Only pixels with high SAM values are mapped. Grey and white areas are pixels with low SAM values owing to spectral mixing.

It has to be noted that the different methodological approaches have been test on different sub-areas. No final maps have been created so far due to the current product development tasks.