We test this premise and explore representation spaces from a single deep convolutional network and their visualization to argue for a novel unified feature extraction framework. The objective is to utilize and re-purpose trained feature extractors without the need for network retraining on three remote sensing tasks i.e. superpixel mapping, pixel-level segmentation and semantic based image visualization. By leveraging the same convolutional feature extractors and viewing them as visual information extractors that encode different image representation spaces, we demonstrate a preliminary inductive transfer learning potential on multiscale experiments that incorporate edge-level details up to semantic-level information.Index Termssettlement mapping, segmentation, representation learning, convolutional neural networks, inductive transfer learning.
Passive radiation detection remains one of the most acceptable means of ascertaining the presence of illicit nuclear materials. In maritime applications it is most effective against small to moderately sized vessels, where attenuation in the target vessel is of less concern. Unfortunately, imaging methods that can remove source confusion, localize a source, and avoid other systematic detection issues cannot be easily applied in ship-to-ship inspections because relative motion of the vessels blurs the results over many pixels, significantly reducing system sensitivity. This is particularly true for the smaller watercraft, where passive inspections are most valuable. We have developed a combined gamma-ray, stereo visible-light imaging system that addresses this problem. Data from the stereo imager are used to track the relative location and orientation of the target vessel in the field of view of a coded-aperture gamma-ray imager. Using this information, short-exposure gamma-ray images are projected onto the target vessel using simple tomographic backprojection techniques, revealing the location of any sources within the target. The complex autonomous tracking and image reconstruction system runs in real time on a 48-core workstation that deploys with the system.
Explosion of spatial data from satellite to citizen sensors has posed the critical challenge of Big Spatial Data integration, analysis, and visualization. This article focuses on research and development activities at Oak Ridge National Laboratory (ORNL) that are addressing end-user applications utilizing high performance computing based geospatial science and technology solutions to optimize the analysis, modeling, and multi-megapixel scale visualization of the geospatial data. Specifically we highlight recent developments and successes in the areas of high resolution settlement mapping, transportation and mobility analysis, and effective monitoring of biomass for energy and food security.
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.