Recursive Hyperspectral Band Processing of Linear Spectral Mixture Analysis

In previous chapters, recursive hyperspectral band processing (RHBP) was developed for subpxiel detection, RHBP of constrained energy minimization in Chap. 13, RHBP of anomaly detection in Chap. 14, unsupervised target detection, RHBP of an automatic target generation process in Chap. 15, and mixed pixel detection/classification, RHBP of orthogonal subspace projection in Chap. 16. This chapter concludes the last piece of RHBP’s applications to the well-known technique linear spectral mixture analysis (LSMA). It develops a new approach, called RHBP of LSMA (RHBP-LSMA), which can process data unmixing according to the band-sequential (BSQ) format. This new concept is different from band selection (BS), which must select bands from a fully collected band set according to a band optimization criterion. There are several advantages of using RHBP-LSMA over BS. In particular, it allows users to perform LSMA using available bands without waiting for a complete collection of full bands. In doing so, an innovation information update recursive equation is further derived and can process LSMA progressively as well as recursively as its band processing taking place. The resulting LSMA becomes RHBP-LSMA. To be more specific, RHBP-LSMA can be carried out by updating LSMA results recursively band by band in the same way that a Kalman filter does in updating data information in a recursive fashion. Consequently, RHBP-LSMA can provide progressive band-varying profiles of LSMA results so that significant bands can also be detected and identified by inter-band changes in LSMA results without BS.