Spectral signature measurements and image processing for geological remote sensing

Sammanfattning: This work focuses on (i) spectral signatures of rocks and lichens, and on the use of satellite images for (ii) survey mapping of areas dominated by exposed bedrock, (iii) bedrock mapping and mineral prospecting.(i) Laboratory measurements demonstrate that a lichen cover can significantly mask the spectral reflectance of rocks. But reflectance measurements on young basalts show that the degree of lichen cover can be helpful when mapping basalts of different age. Measurements of rocks and lichens show the potential of the near infrared regions (0.98-1.11 µm and 1.20-131 µm), which are not covered by Landsat TM.(ii) Landsat TM and SPOT HRV provide good possibilities for mapping of areas dominated by exposed  bedrock to be presented  on the Swedish topographical map (scale 1:50 000). The classification errors can be reduced by the use of a slicing procedure on the training data using transformed data, and by combining the automatic classification with a subsequent manual improvement.(iii) The third part concerns image processing methods for geological studies in two different climatic zones of the world; a high mountain region in Scandinavia, and an arid region in Saudi Arabia. Recommendations of best spectral bands and various methods for image enhancement for geological purposes are discussed. The results demonstrate the importance of the spectral information in TMl , TM4, and TM? (or TMS). The Scandinavian study shows that by extracting bedrock areas using a preprocessing masking technique before additional image enhancement better results are achieved. Preprocessing methods for the removal of picture elements influenced by vegetation, deep shadows, snow and clouds are presented. A manual classification method tested was better than a M-L classification for bedrock mapping. The results from the Saudi Arabia study demonstrate   the potential for including Landsat TM's thermal information for geological mapping.  A further enhancement of high-pass filtered images can be achieved by the use of a spectral shading technique.