Although hyperspectral images capture very detailed information about the scanned objects, their efficient analysis, transfer, and storage are still important practical challenges due to their large volume. Classifying and segmenting such imagery are the pivotal steps in virtually all applications, hence developing new techniques for these tasks is a vital research area. Here, deep learning has established the current state of the art. However, deploying large-capacity deep models on-board an Earth observation satellite poses additional technological challenges concerned with their memory footprints, energy consumption requirements, and robustness against varying-quality image data, with the last problem being under-researched. In this paper, we tackle this issue, and propose a set of simulation scenarios that reflect a range of atmospheric conditions and noise contamination that may ultimately happen on-board an imaging satellite. We verify their impact on the generalization capabilities of spectral and spectral-spatial convolutional neural networks for hyperspectral image segmentation. Our experimental analysis, coupled with various visualizations, sheds more light on the robustness of the deep models and indicate that specific noise distributions can significantly deteriorate their performance. Additionally, we show that simulating atmospheric conditions is key to obtaining the learners that generalize well over image data acquired in different imaging settings.
Time series of weekly and daily solutions for coordinates of permanent GNSS stations may indicate local deformations in Earth's crust or local seasonal changes in the atmosphere and hydrosphere. The errors of the determined changes are relatively large, frequently at the level of the signal. Satellite radar interferometry and especially Persistent Scatterer Interferometry (PSI) is a method of a very high accuracy. Its weakness is a relative nature of measurements as well as accumulation of errors which may occur in the case of PSI processing of large areas. It is thus benefi cial to confront the results of PSI measurements with those from other techniques, such as GNSS and precise levelling. PSI and GNSS results were jointly processed recreating the history of surface deformation of the area of Warsaw metropolitan with the use of radar images from Envisat and CosmoSkyMed satellites. GNSS data from Borowa Gora and Jozefoslaw observatories as well as from WAT1 and CBKA permanent GNSS stations were used to validate the obtained results. Observations from 2000-2015 were processed with the Bernese v.5.0 software. Relative height changes between the GNSS stations were determined from GNSS data and relative height changes between the persistent scatterers located on the objects with GNSS stations were determined from the interferometric results. The consistency of results of the two methods was 3 to 4 times better than the theoretical accuracy of each. The joint use of both methods allows to extract a very small height change below the level of measurement error.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.