2020-09-21, 13:15–13:45, Room 2
Dual Attention Residual Band Selection Network for Spectral-Spatial Hyperspectral Image
Classification
Due to the existence of noise and spectral redundancies in hyperspectral images (HSI),the band selection
is highly required and can be achieved through the attention mechanism.However, existing band selection (BS) methods
fail to consider global interaction between the spectral and spatial information in a non-linear fashion. In this letter, we
propose an end-to-end unsupervised dual attention reconstruction network for band selection (DARecNet-BS). The
proposed network employs a dual attention mechanism, i.e., position attention module (PAM) and channel attention
module (CAM), to recalibrate the feature maps and subsequently uses a 3D reconstruction network to restore the
original HSI. This way, the long range nonlinear contextual information in spectral and spatial directions is captured and
the informative band subset can be selected. Experiments are conducted on three well-known hyperspectral datasets, i.e.,
IP,UP and SA, to compare existing band selection approaches, and the proposed DARecNet-BS can effectively select less
redundant bands with comparable or better classification accuracy