マルチスペクトル画像及び周辺のハイパースペクトル画像を用いた鉱物識別におけるWorldView-3/SWIRの有効性

抄録

A hyperspectral (HS) imager allows us to discriminate minerals more effectively than a multispectral (MS) imager. The spatial coverage of HS images, however, is limited in comparison to that of MS images. Thus, Kruse and Perry proposed a method that uses coincident HS imaging and MS imaging data to extend mineral mapping to larger areas. Hirai and Tonooka modified this method to provide robustness against spectral inconsistency between HS and MS images, and validated their modified method using shortwave infrared (SWIR) images taken over the Cuprite area (Nevada, USA) by an HS sensor, the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS), and an MS sensor, the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER). In the present study, SWIR images of the Cuprite area from two MS sensors, ASTER and WorldView-3 (WV-3), were compared as input under Hirai and Tonooka’s method. The test involved an evaluation of the producer’s accuracy (PA) and the user’s accuracy (UA) for five minerals (muscovite, calcite, buddingtonite, alunite, and kaolinite) in five cases: [A] AVIRIS (30m) and ASTER bands 5 to 9, [B] AVIRIS (30m) and ASTER bands 5 to 8, [C] AVIRIS (15.7m) and WV-3 (15.7m), [D] AVIRIS (30m) and WV-3 (30m), and [E] AVIRIS (30m) and WV-3 (30m in endmember determination and 15.7m in map generation), where 30m and 15.7m are ground resolutions. Our results demonstrate that WV-3/SWIR performs better than ASTER/SWIR mainly due to higher image quality (less noises). In addition, the results from Case E indicate that endmember spectra derived from HS and MS images with a lower spatial resolution can be applied to an MS image with a higher spatial resolution, suggesting a future alternative procedure under the modified method.