Successive Projections Algorithm for Feature Bands Extraction of Suspended Sediment Concentration from Airborne Hyperspectral

Abstract

Suspended sediment concentration (SSC) is an important indicator in water monitoring. Airborne hyperspectral remote sensing can well acquire the spectral features of ground objects to accurately extract water quality parameters, but also has the problems of serious information overlap, invalid information, and redundant bands. In this paper, a successive projections algorithm (SPA) based method for feature bands extraction of suspended sediment concentration from airborne hyperspectral is proposed. The distribution of the feature bands extracted based on the SPA method and the Correlation Coefficient (CC) approach is compared and analyzed, and the SSC retrieval models based on the three-band combination model and the Multiple Linear Regression (MLR) model are constructed using the bands extracted from the study and applied to the North Channel of the Yangtze Estuary. The results show that the SSC retrieval model based on the feature bands extracted by the SPA method has higher accuracy.