SBIR-STTR Award

Multimodal Sensor Fusion and Exploitation for Proliferation Detection
Award last edited on: 11/19/2018

Sponsored Program
SBIR
Awarding Agency
DOE
Total Award Amount
$1,164,930
Award Phase
2
Solicitation Topic Code
01b
Principal Investigator
Stephanie Higgins

Company Information

Arete Associates

9301 Corbin Avenue Suite 2000
Northridge, CA 91396
   (818) 885-2200
   info@arete.com
   www.arete.com
Location: Multiple
Congr. District: 32
County: Los Angeles

Phase I

Contract Number: DE-SC0017775
Start Date: 00/00/00    Completed: 00/00/00
Phase I year
2017
Phase I Amount
$154,933
The Department of Energy’s National Nuclear Security Administration Office of Defense Nuclear Nonproliferation is tasked with detection and identification of new and emerging sources of illicit production and transportation of nuclear-related materials and technology. A variety of sensor modalities providing data at various spatial, spectral, and temporal resolutions exist to aid the detection of targets of interest such as nuclear fuel production sites. Exploiting literal and non-literal data at multiple spatial, spectral, and temporal resolutions requires expertise in many technical aspects of imagery analysis, and is challenging and time-consuming. Many sensors have specialized algorithms unique to their system, requiring specialized analysis and visualization tools. Additionally, some non-literal systems like Synthetic Aperture Radar and Hyperspectral Imaging require special training and a unique set of algorithms for interpretation. To address the challenges presented by these federated systems. Automated fusion algorithms that could manage different sensor modalities collected at different times and from different geometries would greatly enhance analysts’ current ability to extract actionable intelligence from this variety of sensor data. Under this effort, a rigorous approach to multimodal data fusion will be developed as the scientific basis for combining data streams from multiple imaging sensors to address this problem. The algorithms and framework developed here will provide the basis for combining and exploiting a variety of sensor modalities such as synthetic aperture radar, hyper-spectral and panchromatic imagery, video imagery, lidar pointclouds, or other sensor types that may become available. The fused data picture that the framework will produce will improve and accelerate the current analysis chain by improving automated detection and classification algorithms relevant to the Office of Proliferation Detection as well as providing a more informative and intuitive visualization and data interface for Proliferation Detection analysts. In Phase-I the data fusion algorithms will be developed and implemented in a prototype software environment that will be demonstrated against relevant government provided data. Based on this success, a Phase-II effort would be used to mature and refine the algorithms and implement them into an end-to-end fusion framework. The work done under this effort will benefit proliferation detection analysts by making them more effective and efficient at detecting and monitoring nuclear fuel production facilities. The same algorithms would benefit a number of commercial and government programs that use commercial or satellite imagery.

Phase II

Contract Number: DE-SC0017775
Start Date: 00/00/00    Completed: 00/00/00
Phase II year
2018
Phase II Amount
$1,009,997
The Department of Energy’s National Nuclear Security Administration Office of Defense Nuclear Nonproliferation is tasked with detection and identification of new and emerging sources of illicit production and transportation of nuclear-related materials and technology. A variety of sensor modalities providing data at various spatial, spectral, and temporal resolutions exist to aid the detection of targets of interest, such as nuclear fuel production sites. Exploiting literal and non-literal data at multiple spatial, spectral, and temporal resolutions requires expertise in many technical aspects of imagery analysis, and is challenging and time-consuming. Many sensors have specialized algorithms unique to their system, requiring specialized analysis and visualization tools. Additionally, some non-literal systems like Synthetic Aperture Radar and Hyperspectral Imaging require special training and a unique set of algorithms for interpretation. To address these challenges, Areté is developing automated fusion algorithms that can manage different sensor modalities collected at different times and from different geometries. These algorithms will enhance analysts’ ability to extract actionable intelligence from a variety of government and commercial satellite imagery. Under this effort, a rigorous mathematical approach to multimodal data fusion is being developed as the scientific basis for combining data streams from multiple imaging sensors. The algorithms and framework developed here will provide the basis for combining and exploiting a variety of sensor modalities such as synthetic aperture radar, hyper-spectral and panchromatic imagery, video imagery, lidar point clouds, and other sensor types that may become available. The fused data products will improve and accelerate the current analysis chain by improving automated detection and classification algorithms relevant to the Office of Proliferation Detection, as well as providing a more informative and intuitive visualization and data interface for Proliferation Detection analysts. In Phase-I, the data fusion algorithms were developed and demonstrated against local data sets. Based on this success, a Phase-II effort is proposed that will be used to mature and refine the algorithms and implement them into an end-to-end fusion framework. This framework will be developed towards transition to PD analyst end users. Components of the framework will simultaneously be commercialized as GeoKit, as fusion module integrated as a plug-in for satellite image provider platforms. The work done under this effort will benefit proliferation detection analysts by making them more effective and efficient at detecting and monitoring nuclear fuel production facilities. The same algorithms will benefit a number of commercial and government programs that use commercial or satellite imagery.