20 Apr 2020

PUBLICATION NOTIFICATION: Local Spatial Dispersion for Multiscale Modeling of Geospatial Data: Exploring Dispersion Measures to Determine Optimal Raster Data Sample Sizes

ABSTRACT: Scale, or spatial resolution, plays a key role in interpreting the spatial structure of remote sensing imagery or other geospatially dependent data. These data are provided at various spatial scales. Determination of an optimal sample or pixel size can benefit geospatial models and environmental algorithms for information extraction that require multiple datasets at different resolutions. To address this, an analysis was conducted of multiple scale factors of spatial resolution to determine an optimal sample size for a geospatial dataset. Under the NET-CMO project at ERDC-GRL, a new approach was developed and implemented for determining optimal pixel sizes for images with disparate and heterogeneous spatial structure. The application of local spatial dispersion was investigated as a three-dimensional function to be optimized in a resampled image space. Images were resampled to progressively coarser spatial resolutions and stacked to create an image space within which pixel-level maxima of dispersion was mapped. A weighted mean of dispersion and sample sizes associated with the set of local maxima was calculated to determine a single optimal sample size for an image or dataset. This size best represents the spatial structure present in the data and is optimal for further geospatial modeling.

14 Apr 2020

PUBLICATION NOTIFICATION: Coincidence Processing of Photon-Sensitive Mapping Lidar Data

Link: http://dx.doi.org/10.21079/11681/35599  Report Number: ERDC/GRL TR-20-1 Title: Coincidence Processing of Photon-Sensitive Mapping Lidar Data By Christian Marchant, Ryan Kirkpatrick, and David Ober Approved for Public Release; Distribution is Unlimited February 2020 Abstract: Photon-sensitive mapping lidar systems are able to image at greater collection area rates and ranges than linear-mode systems. However, these systems also experience greater noise levels due to shot noise, image blur, and dark current, which must be filtered out before the imagery can be exploited. Described in this report is a synthetic test data set of imagery from a notional airborne Geiger-mode lidar. Also described is the Bridge Sign algorithm, which uses a least-squares technique for noise filtering. The algorithm’s performance was validated using synthetic test imagery of both a toy scene and of a realistic scene, which were generated using the parameters of a notional airborne Geiger-mode system. Analysis of the results shows the technique effectively removes noise and preserves fine details with good fidelity. 30 pages / 1.568 Mb