Study on the alteration minerals caused by oil and gas microseepage by extracting endmembers from hyperion

Currently, it is not adequate to extract and classify the hyperspectral image for oil and gas exploration. For mining the deeper remote sensing data about hydrocarbon microseepage, the technology of spectral unmixing can extract pure surface information and provide each pixel´s endmember abundance. However, endmember extraction is the key of spectral unmixing. The paper mainly researches a valid algorithm, which has already be proposed, for extracting endmembers caused by the hydrocarbon microseepage from Hyperion image. Comparing the different endmember algorithms by analyzing the correlations between endmember abundances and contents, the experimental results demonstrate that PPI algorithm is more robust and efficient for extracting the hydrocarbon microseepage information in the study area.