# Advances in Algorithms for Processing, CTIS Flash Hyperspectral Imagery - PowerPoint PPT Presentation

PPT – Advances in Algorithms for Processing, CTIS Flash Hyperspectral Imagery PowerPoint presentation | free to download - id: b68bc-ZDc1Z

The Adobe Flash plugin is needed to view this content

Get the plugin now

View by Category
Title:

## Advances in Algorithms for Processing, CTIS Flash Hyperspectral Imagery

Description:

### Advances in Algorithms for Processing, CTIS Flash Hyperspectral Imagery ... by a disperser (computer generated hologram) onto the spectrometer's focal plane ... – PowerPoint PPT presentation

Number of Views:23
Avg rating:3.0/5.0
Slides: 2
Provided by: Oo46
Category:
Tags:
Transcript and Presenter's Notes

Title: Advances in Algorithms for Processing, CTIS Flash Hyperspectral Imagery

1
Advances in Algorithms for Processing, CTIS Flash
Hyperspectral Imagery
Deirdre' Johnson Fisk University Research
Alliance in Math and Science Computer Science and
Mathematics Division Mentor Dr. Jacob
Barhen http//www.csm.ornl.gov/Internships/rams_0
7/abstracts/d_johnson.pdf
ABSTRACT The Missile Defense Agency (MDA)
requires Flash Hyperspectral Imagery (FHI) for
target kill assessment and spectral analysis of
high-impact scenarios. This work exploits data
collected by the CTIS sensor, an imaging
spectrometer developed by the University of
Arizona (UOA). CTIS collects a set of registered
spectrally contiguous images of a scenes spatial
and radiation distribution. The target data
acquired by CTIS form a three dimensional (x, y,
?) object referred to as the target object cube.
This object is actually projected by a disperser
(computer generated hologram) onto the
spectrometers focal plane array (FPA). In order
to recover the target object from the data
captured by the FPA a large scale systems of
linear equations must be solved. These equations
relate the objects voxels to the FPA pixels
through the Optical System Matrix (OSM). The
equations are corrupted by both signal-dependent
Poisson (photon) noise and post-detection
Gaussian noise. The goal of MDA is to explore
innovative approaches for solving the target
recovery problem in real time. Originally, MDA
considered the mixed expectation maximization
(MEM) stochastic algorithm. Unfortunately, MEM
exhibits oscillatory behavior and poor
convergence. Three alternate methods were
considered two were developed at ORNL, i.e.,
asymptotic attractor (AA) dynamics, sparse
conjugate gradient (SCG), and one was proposed by
UOA, i.e., multiplicative algebraic
reconstruction technique (MART). These approaches
assume that noise has been removed by
prefiltering. Results demonstrate that
considerable speed-up under monotonic convergence
is achieved.
University of Arizona Computer Tomography Imaging
Spectrometer
Reconstruction Approaches
Mixed Expectation Maximization
MART
Asymptotic Attractor Dynamics
• a nonlinear iterative computed tomography
• faster
• takes away the noise
• 3 matrix-vector multiplications per iteration
• algorithm exhibits oscillatory behavior
• convergence requires over 100 iterations
• UA stops at 10-20! 40m / run
• optical system matrix H is non-square,
non-symmetric, and singular
• no inversion of H required .
• sparsity of H is fully exploited, and no
transpose of H is used

MART
Asymptotic Attractor Dynamics
Mixed Expectation Maximization
Hyperspectral Object Reconstruction
Convergence to True Target Voxel Recovery Error
AA CG MART MEM
Conclusions A speed up factor in excess of 1,000
was observed
Future Research Implement the performed
algorithms on IBM Cell multicore processors
The Research Alliance in Math and Science program
Computing Research, U.S. Department of Energy.
The work was performed as part of a joint project
funded by Office of Naval Research Discovery and
Innovation Program, at the Oak Ridge National
Laboratory which is managed by UT-Battelle, LLC
under Contract No. De-AC05-00OR22725. This work
has been authored by a contractor of the U.S.
Government, accordingly, the U.S. Government
retains a nonexclusive, royalty-free license to
publish or reproduce the published form of this
contribution, or allow others to do so, for U.S.
Government purposes.
The author would like to thank Dr. Jacob Barhen
for the opportunity to work on this project. A
thank you goes out to Dr. Stephen Egarievwe and
Fisk University for the nurturing and scholastic
support of all opportunities.
OAK RIDGE NATIONAL LABORATORY U.S. DEPARTMENT OF
ENERGY