Precise land cover classification in complex scene based on ultra-hyperspectral data from AisaIBIS sensor

The advancement of ultra-hyperspectral imaging technology, exemplified by the AisaIBIS sensor, has enabled a leap from hyperspectral data (hundreds bands) to (thousands bands). It provides immense potential for precise ground object recognition within intricate scenes. However, complexities inherent features objects, coupled with copious redundant information data, pose substantial challenges accurate recognition. Therefore, this paper proposed comprehensive framework explore optimal classification strategy in complex scenes (12 vegetation and non-vegetation classes). (a) Our investigation delves into influence diverse feature subsets range machine learning classifiers on precision objects is up an overall accuracy 88.44%, effectively avoiding curse dimension, significantly enhancing capability recognize objects. (b) Furthermore, based simulation images different spectral resolutions, we compared results (0.11 nm) datasets (10 nm, 5 1 methods. Compared datasets, improved 5.30–6.38%. This substantiates pronounced advantages land cover classification. study valuable reference application scenes, urban monitoring.