| Popis: |
In this paper, we suggest optimal flat fields for calibration of hyperspectral data (EO1) — predominantly of urban nature. We compare the suitability of various flat field candidates by performing vegetation, soil, and impervious surface (VIS) classification. Flat Field methods provide 90% average overall accuracy (with best 100% and 77% worst accuracies) over number of conducted experiments. Flat fields show marginal improvement over IAR. Most common urban land covers such as play grounds, concrete parking lots, and especially industrial roof covers have been found adequate for the image calibration. |
