Initialization‐Driven Endmember Extraction Algorithms (ID‐EEAs)
Chein‐I Chang · 2013
This chapter introduces a new type of EEAs, called initialization-driven EEAs (ID EEAs), which use specific initial conditions for algorithm initialization. Depending on how initial conditions are generated, ID-EEAs can be further categorized into initial endmember-driven EEAs (IED-EEAs) which are essentially designed for those EEAs that only require the first initial endmember to initialize the algorithms and endmember initialization algorithm (EIA)-driven EEAs (EIAD-EEAs) which are primarily designed for those EEAs that generally require a complete set of p initial endmembers for algorithm initialization. This chapter assumes that the value of p is estimated by virtual dimensionality (VD), and this VD-estimated p will be used to determine how many endmembers are required for an EEA to generate. This chapter addresses the issue in the use of random initial endmembers and further derives ID-EEAs which implement specific initial endmembers for EEAs.