SLCRF: Subspace Learning With Conditional Random Field for Hyperspectral Image Classification

Subspace learning (SL) plays an important role in hyperspectral image (HSI) classification, since it can provide effective solution to reduce the redundant information pixels of HSIs. Previous works about SL aim improve accuracy HSI recognition. Using a large number labeled samples, related methods train parameters proposed solutions obtain better representations pixels. However, data instances may not be sufficient enough learn precise model for classification real applications. Moreover, is well-known that takes much time, labor and human expertise label images. To avoid aforementioned problems, novel method includes probability assumption called subspace with conditional random field (SLCRF) developed. In SLCRF, first, 3D convolutional autoencoder (3DCAE) introduced remove addition, relationships are also constructed using spectral-spatial among adjacent Then, (CRF) framework further embedded into procedure semi-supervised approach. Through linearized alternating direction termed LADMAP, objective function SLCRF optimized defined iterative algorithm. The comprehensively evaluated challenging public datasets. We achieve stateof-the-art performance these sets.