George O. Ramseyer

Imagery Pattern Recognition and Pub/Sub Information Management

Imagery pattern recognition (IPR) is becoming more challenging as image resolutions continue to improve and algorithms become more complex. Systems for analysis are being developed on increasingly diverse and heterogeneous platforms, from hardware support using field programmable gate arrays, graphics processing units, and Cell processors to high performance computers, featuring multi-core processors and shared memory communications. Given the complexity of modern IPR systems, neither analytical nor adaptive methods are sufficient to understand the shortfalls of an approach to IPR for any particular architecture. Our test harness implements an adaptive, experimental approach that can be used to quickly estimate performance and perform validation. A case study is implemented that implements the test harness for a publish/subscribe architecture, and in which several approaches for brokering are compared in a pub/sub implementation.

Using globus grid objects to extend a corba-based object-oriented system

This paper, along with the accompanying demonstration, describes an object-oriented framework for high-performance computing. The original system [1] was developed with Common Object Request Broker Architecture (Corba) [2] objects to implement the framework interfaces to code-dependent methods. To add new codes to the initial system required the implementation of new Corba interface objects. Globus Grid objects [3] have extended that original system, and have resulted in added flexibility in the design for the addition of new codes. High-performance computing agent services were also implemented as Globus Grid objects, resulting in enhanced communications between the framework user interface services and the agents responsible for executing the codes. The management functions for remote agents have also been improved.

A FrameWork for high-performance image exploitation

This paper describes an object-oriented system that is designed to provide access to databases and high-performance computing systems to process images and other objects found in the databases. The system features a Web-based front-end and CORBA middleware to provide convenient access and management of services. The Kerberos authentication system is used for security to avoid IP address-based approaches that restrict user mobility.

Information Management Support for Interactive High-Performance Computing Frameworks

The information management capabilities of the Joint Battlespace Infosphere (JBI) and the evolving 100X JBI were evaluated to support a network of remote computers. The network consisted of an interactive testbed of distributed high performance computers (HPCs) across the Defense Research and Engineering Network (DREN). This Distributed Interactive High Performance Computing Testbed (DIHT) was developed to enable researchers to operate in near real time in a diverse HPC environment. This interactive environment included access to six remote Linux clusters, and allowed the rapid execution of parallelized software in real time. At peak times the DIHT is producing hundreds to thousands of information objects per second

Application Development Framework for the Rapid Integration of High Performance Image Processing Algorithms

The application development framework (ADF) provided developers with a unique environment that supported the rapid integration and testing of image processing algorithms on high performance computers (HPCs). Using object-oriented middleware for the base of the system, along with Web technologies, allowed considerable flexibility in extending the system to a broad range of system development tools and components. The pub/sub system at the core of the ADF was the foundation that provided rapid system integration.

Redeveloping a high-Performance computing framework

This paper describes our experience in redeveloping an object-oriented system. Experience working with the system along with new requirements that were introduced as the system matured motivated a significant redesign and reimplementation effort. The initial object-oriented design helped to allow extension as well as restructuring of system objects. Experience in integrating new objects into the system and reimplementing objects in different languages drove a refactoring process. The Web interface to the system was reimplemented as a Grid portal.

Using metrics to optimize a high performance intelligent image processing code

Optimizing the execution of intelligent codes on high performance computers (HPCs) has become more challenging as the numbers of processors increases. Single processors in many HPCs have been replaced with dual processors, and more recently multiprocessors. This, combined with the inherent complexities of multi-core processors, has made the processing of intelligent codes even more complex on the latest HPCs. The coming availability of thousands of processors in more affordable medium sized HPCs offers the potential for improved performance for codes that can scale sufficiently to take advantage of hundreds of teraflops. Additionally, techniques for harnessing the performance potential of multi-code processors require the appropriate location of data in shared memories, or even shared level-2 caches, and can afford additional orders of magnitude performance increases. The key to designing code that uses the available teraflops wisely is an understanding of the applications behavior. For intelligent systems, whose behavior may depend on heuristics evaluated at runtime, measurements and profiling runs provide the basis for system design decisions, regarding distribution of data and processing. This paper focuses on the metrics needed to optimize intelligent codes, and how a specific image processing code was instrumented to produce the required metrics.

Fault tolerant integrated information management support for physically constrained iterative deconvolution

Multiple image processing algorithms are often required to process computer vision inputs. The rapid processing of complex image streams requires more computing power than is found in a typical PC based computer or workstation, and the processing power of high-performance computers (HPCs) and Linux clusters have been required to do this type of rapid massive processing. Emerging multicore processors offer the possibility of doing these types of processing at the PC level in real time. The physically constrained iterative deconvolution (PCID) algorithm is a multi-frame blind deconvolution (MFBD) parallel algorithm that allows the extraction of simple and complex information from multiple images. Massive computing power is required to use this algorithm in real time. Message passing interface (MPI) is normally used with PCID for communications between processors in multiprocessor systems. However, MPI has fault tolerant issues. A tool to replace MPI for multiprocessor communications has been developed that supports a high degree of fault-tolerance, and facilitates multiple image processing by integration with a publication/subscription infrastructure. This tool is demonstrated here for the PCID algorithm. Other attributes of MPI and this tools publication/subscription information management support for PCID are compared and contrasted.

Information management for high performance autonomous intelligent systems

The publish/subscribe model for information management is particularly well suited for use in intelligent autonomous systems, ranging from robots to tactical communication systems. Information management systems that support pub/sub inherently provide a high degree of autonomy for users and communicating systems. The pub/sub paradigm can allow autonomous intelligent systems to communicate without requiring connection to a centralized brokering system. Each system is responsible for part of the overall brokering function, which imposes a cost for local system resources and proportionally diminishes the intelligence that can be expressed by each node. This raises the question of whether there exist controls that each intelligent autonomous system can use to avoid over-committing resources for publication brokering, such that node intelligence is uncompromised. Issues which affect autonomy in a pub/sub system that is currently under development are addressed.

Chemical and electrical analysis of CdS interlayers on InP and related materials

Cadmium sulfide (CdS) layers were deposited from aqueous solutions of thiourea, cadmium sulfate, and ammonia on (100) InP, InGaAs, and InAlAs. X-ray photoelectron spectroscopy (XPS), Auger electron spectroscopy (AES), and atomic force microscopy (AFM) were used to investigate the structural and chemical nature of the deposited CdS layer and the CdS/semiconductor interface. XPS showed that the deposition process effectively removes existing native oxides on InP and InAlAs before CdS growth occurs. Capacitance-voltage measurements of metal-insulator- semiconductor (MIS) capacitors were used to investigate the interface- state density of samples with and without CdS films between InP and a deposited insulator. CdS interlayers were found to reduce both the hysteresis and the interface-state density of the MIS capacitors. Applications of CdS interlayers for various photonic devices will be discussed.