Yang H.,Mississippi State University |
Du Q.,Mississippi State University |
Chen G.,DCM Research Resources, LLC
IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
The high dimensionality of hyperspectral imagery challenges image processing and analysis. Band selection is a common technique for dimensionality reduction. When the desired object information is unknown, an unsupervised band selection approach is employed to select the most distinctive and informative bands. Although band selection can significantly alleviate the computational burden in the following data processing and analysis, the process itself may induce additional computation complexity, especially when the image spatial size is large; it may be time-consuming for unsupervised band selection methods that need to take all pixels into consideration. Parallel computing techniques are widely adopted to alleviate the computational burden and to achieve real-time processing of data with vast volume. In this paper, we propose parallel implementations via emerging general-purpose graphics processing units (GPUs) for band selection without changing band selection result. Its speedup performance is comparable to the cluster-based parallel implementation. We also propose an approach to using several selected pixels for unsupervised band selection and the number of pixels needed can be equal to the number of selected bands minus one. With whitened pixel signatures (not the original pixels), band selection performance can be comparable to or even better than that from using all the pixels. For this approach, parallel computing is implemented for pixel selection only, since computational complexity in band selection has been greatly reduced. © 2011 IEEE. Source
Liu K.,Mississippi State University |
Ma B.,Mississippi State University |
Du Q.,Mississippi State University |
Chen G.,DCM Research Resources, LLC
Journal of Applied Remote Sensing
In our previous work, we proposed a joint optical flow and principal component analysis (PCA) approach to improve the performance of optical flow based detection, where PCA is applied on the calculated two-dimensional optical flow image, and motion detection is accomplished by a metric derived from the two eigenvalues. To reduce the computational time when processing airborne videos, parallel computing using graphic processing unit (GPU) is implemented on NVIDIA GeForce GTX480. Experimental results demonstrate that our approach can efficiently improve detection performance even with dynamic background, and processing time can be greatly reduced with parallel computing on GPU. © 2012 Society of Photo-Optical Instrumentation Engineers. Source
Agency: Department of Defense | Branch: Air Force | Program: SBIR | Phase: Phase I | Award Amount: 100.00K | Year: 2008
We propose a highly innovative modeling and analysis framework for distributed collaborative systems with aid of data-mining and data-fusion concept. Our approach has two major highly coupled parts: 1) Data fusion module. Mutliple-sensor data are fused to perform situation awareness. To generate the intent inference of the targets, we will extend Markov games via incorporating Multi-Cumulant Pareto Nash strategies and Graph concept. From a perspective of distributed decision making problem, we dynamically adapt Hierarchical Task Network and Auction algorithm for optimal meta-task decomposition and assignment. A collaborative search and tracking oriented sensor management algorithm based on Pursuit-Evasion game will be exploited to improve the performance of the multi-layered sensing system. 2) Data-mining module. Adaptation and pattern/feature recognition are carried out to dynamically select (or mine) appropriate features or feature sets and quickly associate them with the adversary intent and executable actions. In some time-critical scenarios, firstly, a primitive adversary intent estimation and the associated friendly force collaborative response actions can be quickly provided by the Data-mining module. Then, a refinement based on data-fusion will be carried out to improve the performance of the decision aids tool. Additionally, we will incorporate a semantic and textual processing technique to convert meta-tasks into actionable fusion processes BENEFIT: The proposed game theoretic decision aid tool for cooperative system modeling, simulation, and analysis has tremendous applications potential in many military applications. It can also be directly used for developing of advanced mission planning and emergency preparedness decision support systems such as Space Situational Awareness Fusion Intelligent Research Environment [SAFIRE] program, BMDS system, Future Combat System (FCS), Joint Strike Fighter (JSF) program, JSSEO program, and AEGIS program. DCM and Lockheed Martin are building a mentor-protégé program. We will leverage this relationship to identify the end customer, and work with these teams to transition our technoloy into their program. The DOD contact who knows the details of our work and who knows the above programs is Dr. Erik Blasch from AFRL. Other potential commercial applications include inter-satellite communications, multi-layered sensing, disaster assessment, air traffic control system, the national weather service, physical security systems, law enforcement agency, emergency control center, border and coast patrol, pollution monitoring, remote sensing and global awareness. We expect the aggregated market size will be similar to that of military applications.
Agency: Department of Defense | Branch: Air Force | Program: STTR | Phase: Phase I | Award Amount: 100.00K | Year: 2010
In this proposal, DCM Research Resources (DCM), LLC, and Syracuse University propose a highly innovative distributed pattern detection and classification approach, called Compressive Sensing aided Sequential Pattern Detection and Classification (CSASPDC) in Distributed Sensor Network. Our goal is to develop sophisticated approaches that can effectively detect or classify very weak distributed patterns that are undetectable in the local signatures at individual nodes. At the mean time, any solution to pattern detection and classification needs to take into account the very limited energy and communication bandwidth. We propose a pattern detection/classification framework that combines sophisticated techniques in several areas, including compressive sensing, distributed detection, game theoretic sensor selection and management for detection/estimation, and sequential detection/classification, and secure cognitive radio, leveraging our previous experiences in these areas BENEFIT: The proposed compressive sensing aided sequential pattern detection and classification (CSASPDC) algorithm for distributed sensor network is very important in many military (DoD) applications including reconnaissance and surveillance, homeland security, etc. It can be directly used for developing of advanced mission planning and emergency preparedness decision support systems such as CB agent defense, Space Situational Awareness Fusion Intelligent Research Environment [SAFIRE] program, JSPOC Situational Awareness Response System (JSARS), BMDS system, Future Combat System (FCS), Joint Strike Fighter (JSF) program, and JSSEO program. During the Phase I, we will work closely with Lockheed Martin MS2, who is prime contractor on the Aegis weapon system, the Littoral Combat Ship, and C2 lead for the DDG-1000 program. We have developed a strong and realistic plan to transition our technology to their programs (support letter attached). In addition, DCM and Lockheed Martin are building a mentor-protégé program. We will leverage this relationship to identify the end customer, and work with these teams to transition our Phase 2 technology into their program. The DOD contact who knows the details of our work and who knows the above programs is Dr. Erik Blasch from AFRL. The market for military applications is quite large.Other potential commercial applications include air traffic control system, network security intrusion detection, the national weather service, physical security systems, law enforcement agency, emergency control center, border and coast patrol, pollution monitoring, remote sensing, robotics, medical applications, and global awareness. The size of this market is not small and hard to estimate. We expect the aggregate market size will be similar to that of military applications.
Agency: Department of Defense | Branch: | Program: SBIR | Phase: Phase II | Award Amount: 748.75K | Year: 2010
In phase 1, we have developed a holistic approach to systematically employ distributed sensor management techniques for large scale networks, with technical innovations on network utility optimization, efficient distributed computational methods, and robust and scalable information propagation. To attain predictive and responsive space situation awareness (SSA), the DCM team design, Awareness-based Compressed Data Collection and Dynamic Resource Management (ACDC-DRM), jointly addresses compressive and active sensing with network resource management in a unified manner, which represents an emerging data collection paradigm that is important, and indispensable in many cases, to the success of resource-constrained, large-scale sensor networks monitoring dynamic and/or localized phenomena. Our phase 1 work consists of: 1) compressive and active sensing algorithms for efficient data acquisition and reconstruction of a large-scale complex target-field, 2) in-network cooperative multi-sensor searching and tracking algorithm with information-based awareness metrics; and 3) game-theoretic dynamic sensor resource allocation approach for intelligent targets. In addition, a prototype based on open-source software has been implemented to illustrate the algorithms. In Phase II, we plan to coordinate with government POCs, academic researchers and industrial partners on research and development, as well as updating various open-standard database collection routines. We will also refine the key algorithms in our ACDC-DRM design, extend the system capability using theoretical performance guidelines quantified under various operating conditions, and develop an executive prototype for realistic network scenarios. BENEFIT: The first potential commercialization application is JSTAR and SAFIRE program. The second potential application is other DoD application such as ARL. The third potential application is AEGIS program and other programs where LM is the Prime Contractor. Lockheed Martin MS2 is prime contractor on the Aegis weapon system, the Littoral Combat Ship, and C2 lead for the DDG-1000 program. We have developed a strong and realistic plan to transition our technology to their programs. During the first stage of Phase II, LM MS2 will study and quantify through simulation and analysis how the ACDC-DRM can enhance the performance of their network centric platform. Assuming a successful Phase II, during the Phase III LM MS2 will build on the results of the Phase II work to implement and test the technology in real-world systems. Beyond the AEGIS first application, the innovations we are developing will improve situation awareness, planning, decision support for many military applications and we will aggressively pursue these other applications. As a metric of success, the technology is also applicable to commercial systems. Our target application will focus on disaster management, intelligent air traffic control system, and network defense.