Adaptive Multisensor Acquisition via Spatial Contextual Information for Compressive Spectral Image Classification

Spectral image classification uses the huge amount of information provided by spectral images to identify objects in scene interest. In this sense, typically contain redundant that is removed later processing stages. To overcome drawback, compressive imaging (CSI) has emerged as an alternative acquisition approach captures relevant using a reduced number measurements. Various methods classify from projections have been recently reported whose measurements are captured non-adaptive, or adaptive schemes discarding any contextual may help reduce projections. paper, method for proposed. particular, we adaptively design coded aperture patterns dual-arm CSI architecture, where first system obtains multispectral and second arm registers hyperspectral snapshots. The proposed exploits spatial coding such subsequent snapshots acquire scene's complementary improving performance. Results extensive simulations shown two state-of-the-art databases: Pavia University Indian Pines. Furthermore, experimental setup performs sensing was built test performance on real data set. exhibits superior with respect other