Abstract: It is a challenge to identify lithological type by satellite hyperspectral remote sensing under dense vegetation cover. In this study, Huma area, Heilongjiang Province is taken as an example. The Hyperion data from EO-1 satellite are pre-processed to estimate vegetation indexes. The vegetation cover information, background, rock-soil distribution information and anomaly, are clearly separated on the 2D scattering points diagram generated by employing principal component analysis (PCA) method. By the field spectrum measurements, after the continuum removal of field spectrum from Hyperion image spectrum, main lithological types are identified by spectral feature fit (SFF) method. A technical flow chart is brought forward to classify lithology from hyperspectral data. Applying these methods, the deduced lithologic classification and boundary lines of different geologic units in the study area coincide with the known geology map, proving the ability of the methods in the paper to do lithology classification from hyperspectral data at dense vegetation cover area.

Key words: dense vegetation cover area, hyperspectral remote sensing, principal component analysis, spectral feature fit (SFF), lithologic classification