The goal of this research is to develop a general deep learning solution for atmospheric correction and target detection using multiple hyperspectral scenes. It is assumed that the scenes differ only in range and viewing angles, that they are acquired in rapid sequence using an airborne sensor orbiting a target, and that the target and the atmosphere remain invariant within the time scale of the collection. Several hundred thousand hyperspectral simulations were performed using the MODTRAN model and were used to train the deep learning solution, as well as to validate the proposed method. The input to the deep learning solution is a matrix of the simulated radiances at the sensor as function of wavelength and elevation angles. The output is atmospheric upwelling, downwelling, and transmission. This solution is repeated for all or a subset of pixels in the scene. We focus on emissive properties of targets, and simulations are performed in the longwave infrared between 7.5 and 12 μm. Results show that the proposed method is computationally efficient and it can characterize the atmosphere and retrieve the target spectral emissivity within one order of magnitude errors or less when compared with the original MODTRAN simulations.
On-going research to improve hyperspectral target detection generally focuses on statistical detector performance, reduction of background or environmental contributions to at-sensor radiance, dimension reduction and many other mathematical or physical techniques. These efforts are all aimed at improving target identification in a single scene or data cube. This focus on single scene performance is driven directly by the airborne collection concept of operations (CONOPS) of a single pass per target location. Today's pushbroom and whiskbroom sensors easily achieve single passes and single collects over a target location. If multiple passes are flown for multiple collects on the same location, the time scale for revisit is several minutes.
Emerging gimbaled hyperspectral imagers have the capability to collect multiple scans over the same target location in a time scale of seconds. The ability to scan the same location from slightly different collection geometries below the time scale of significant solar and atmospheric change forces us to reexamine the methods for target detection via the fundamental radiance equation. By expanding the radiance equation in the spatial and temporal dimensions, data from multiple hyperspectral images is used simultaneously for determining at-sensor radiance and surface leaving radiance with the ultimate goal of improving target detection.
This research reexamines the fundamental radiance equation for multiple scan collection geometries expanding both the spatial and temporal domains. In addition, our assumptions for determining at-sensor radiance are revisited in light of the increased dimensionality. The expanded radiance equation is then applied to data collected by a gimbaled long wave infrared hyperspectral imager. Initial results and future work are discussed.
Automated hyperspectral image processing enables rapid detection and identification of important military targets from
hyperspectral surveillance and reconnaissance images. The majority of this processing is done using ground-based
CPUs on hyperspectral data after it has been manually exfiltrated from the mobile sensor platform. However, by
utilizing high-performance, on-board processing hardware, the data can be immediately processed, and the exploitation
results can be distributed over a low-bandwidth downlink, allowing rapid responses to situations as they unfold.
Additionally, transitioning to higher-performance and more-compact processing architectures such as GPUs, DSPs, and
FPGAs will allow the size, weight, and power (SWaP) demands of the system to be reduced. This will allow the next
generation of hyperspectral imaging and processing systems to be deployed on a much wider range of smaller manned
and unmanned vehicles.
In this paper, we present results on the development of an automated, near-real-time hyperspectral processing system
using a commercially available NVIDIA® Telsa™ GPU. The processing chain utilizes GPU-optimized implementations
of well-known atmospheric-correction, anomaly-detection, and target-detection algorithms in order to identify targetmaterial
spectra from a hyperspectral image. We demonstrate that the system can return target-detection results for
HYDICE data with 308×1280 pixels and 145 bands against 30 target spectra in less than four seconds.
Canopy cover is a significant factor in assessing the performance of target detection algorithms in forested environments.
This is true of electro-optical (EO), radar frequency (RF), light detection and ranging (LIDAR), multi/hyperspectral
(MSI/HSI), and other remote sensing methods. This research compares traditional ground based methods of estimating
canopy closure with estimates of canopy cover via spectral detection methods applied to VNIR/SWIR hyperspectral
imagery. This paper uses canopy cover and canopy closure as defined by Jennings, et al. [1]. In the Summer of 2009, a
pushbroom VNIR/SWIR hyperspectral sensor collected data over a forested region of the Naval Surface Warfare Center,
Dahlgren Division, Virginia. This forested region can be best described as single canopy cover with multiple tree
species. Hyperspectral imagery was collected over multiple days and at multiple altitudes in August and September,
2009. On the ground, densiometer measurements and hemispherical photography were used to estimate canopy closure
at 10 meter intervals across a 2500 m2 grid. Several spectral detection methods including vegetation indices, matched
filtering, linear un-mixing, and distance measures, are used to calculate canopy coverage at varying ground sample
distances and across multiple days. These multiple estimates are compared to the ground based measurements of canopy
closure. Results indicate that estimates of canopy coverage via VNIR/SWIR hyperspectral imagery compare well to the
ground based canopy closure estimates for this single canopy region. This would lead to the conclusion that it is possible
to use airborne VNIR/SWIR hyperspectral alone to provide an accurate estimate of canopy cover.
Identification of differing vegetation species has been a lauded ability of hyperspectral imagery and analysis but continues to be a challenging problem. Hyperspectral imagery has been used for years in applications such as vegetation analysis and delineation, terrain categorization, explosive mine detection, environmental impacts and effects, and agriculture and crop evaluation. Unlike applications which focus on detection of specific targets with constant spectral signatures, vegetation signatures continually vary across their growth cycle. In order to identify various vegetation species, either large collections of time-varying reference signatures are required, or ground truth/training data is needed. These are not always viable options and in many cases only in-scene data can be used. In this study we compare the performance of various spectral matching methods in separating vegetation at the species level. Parametric, non-parametric, derivative techniques, and other methods are compared. These methods are applied to a complex scene, the National Arboretum in Washington DC, which was imaged by an airborne hyperspectral sensor in August, 2008. This survey assesses performance of spectral matching methods for vegetation species delineation and makes recommendations for its application in hyperspectral data analysis.
Wavelet Packets have been used to detect trace gases in long-wave infrared hyperspectral imagery. Spectral features for
gases in the long-wave infrared can be characterized as lorentzian emission and absorption features. This is unlike
spectral features for materials in visible, near infrared, and short-wave infrared, which are dependent on both the source
illumination and the physical reflective properties of the surface material. In the reflective domain, features are
represented by a much greater variety of shapes and distributions. These types of features are ideal for an adaptive target
signature approach such as the Wavelet Packet Subspace (WPS). The WPS technique applies the wavelet packet
transform and selects a best basis for pattern matching. The wavelet packet transform is an extension of the wavelet
transform, which fully decomposes a signal into a library of wavelet packet bases. An orthogonal best basis is chosen
which best represents features in the target signature at multiple resolutions. This best basis is then used for target
detection. In this research, the Wavelet Packet Subspace technique is extended to reflective hyperspectral imagery.
Using hyperspectral imagery data with known ground truth, a quantitative comparison is made between the WPS
technique and other spectral matching methods. Spectral angle mapper, and clutter matched filter, and WPS are
compared. Initial results demonstrate that performance of the WPS technique for reflective hyperspectral imagery is
comparable to that of existing methods.
A method for trace gas detection in hyperspectral data is demonstrated using the wavelet packet transform. This new
method, the Wavelet Packet Subspace (WPS), applies the wavelet packet transform and selects a best basis for pattern
matching. The wavelet packet transform is an extension of the wavelet transform, which fully decomposes a signal into a
library of wavelet packet bases. Application of the wavelet packet transform to hyperspectral data for the detection of
trace gases takes advantage of the ability of the wavelet transform to locate spectral features in both scale and location.
By analyzing the wavelet packet tree of specific gas, nodes of the tree are selected which represent an orthogonal best
basis. The best basis represents the significant spectral features of that gas. This is then used to identify pixels in the
scene using existing matching algorithms such as spectral angle or matched filter. Using data from the Airborne
Hyperspectral Imager (AHI), this method is compared to traditional matched filter detection methods. Initial results
demonstrate a promising wavelet packet subspace technique for hyperspectral trace gas detection applications.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.