Imaging Software

The software that produces the imagery from spectrographic instruments is a very important part of the process. ESSI is working to produce real-time processing for its PROBE series of instruments. The figures in the following examples are examples of the type of imagery now produced:

Figure 1

Moving from assigning different values of colors to segments of the spectrum for highlighting purposes to application of hyperspectral interpretations requires knowledge of how a hyperspectral image is constructed. It starts with the spectrum itself . The PROBE-1 records 128 channels (or bands) between .4 and 2.5 micrometers of the electromagnetic spectrum. See Figure 1. Materials on the Earth’s surface typically have characteristic spectral features. Once the hyperspectral data (called radiance in its raw form) has been converted to reflectance (solar, atmospheric, and instrument effects removed) users can go to a spectral library and select what material or signature they wish to look for in the data. See Figure 2. See typical signature for water, vegetation and soil. Each image is actually a “cube of information” as represented in Figure 3 or in a flattened our version in Figure 4. The X and Y axes are the spatial dimension of the image under investigation. The Z axis is the spectral dimension. All 128 bands are captured for each pixel.

Figure 2

Figure 3

Figure 4

If one changes the scale of view one changes the pixel size identified. As in a camera, a close up adjusts the pixel size. A data capture from high altitude will provide a broader view but less detail in the image. Figure 5 shows this with even a small town and a bay in the image.

Figure 5

Mineral exploration has provided much of the growth in hyperspectral remote sensing technology. The technology is proven in the identification of common “pathfinder” surface minerals that are then used by geologists to determine specific points of interest for “ground truthing”.

Figure 6 shows what a typical mine area looks like from high altitude hyperspectral remote sensing. It begins with a natural color version similar to an airphoto and proceeds through a series of images with hyperspectral software analysis to highlight specific minerals. In this case it was relatively easy to prove the remote sensing results because the true ground minerals were well known.

Figure 6

Figure 7

Figure 7 is a first spectral interpretation of the mine image using some of the visible and near Infrared signatures of the spectrum where minerals of interest are located. (Green is vegetation, purple are disturbed soils, red is goethite, yellow is jarosite, orange is a material with a ferric iron signature.)

From the same data “cube” comes another hyperspectral interpretation using the short wave infrared part of the spectrum to map clay and carbonate minerals. See Figure 8. (red is kaolinite, magenta is halloysite, blue is carbonate, green is vegetation.)

Figure 8