David A. Maluf

Bayesian super-resolved surface reconstruction from images

Bayesian inference has been used successfully for many problems where the aim is to infer the parameters of a model of interest. In this paper we formulate the three dimensional reconstruction problem as the problem of inferring the parameters of a surface model from image data, and show how Bayesian methods can be used to estimate the parameters of this model given the image data. Thus we recover the three dimensional description of the scene. This approach also gives great flexibility. We can specify the geometrical properties of the model to suit our purpose, and can also use different models for how the surface reflects the light incident upon it. In common with other Bayesian inference problems, the estimation methodology requires that we can simulate the data that would have been recorded for any values of the model parameters. In this application this means that if we have image data we must be able to render the surface model. However it also means that we can infer the parameters of a model whose resolution can be chosen irrespective of the resolution of the images, and may be super-resolved. We present results of the inference of surface models from simulated aerial photographs for the case of super-resolution, where many surface elements project into a single pixel in the low-resolution images.

Abstraction of Representation for Interoperation

When combining data from distinct sources, there is a need to share meta-data and other knowledge about various source domains. Due to semantic inconsistencies, problems arise when combining know edge across domains and the knowledge is simply merged. Also, knowledge that is irrelevant to the task of interoperation will be included, making the result unnecessarily complex. An algebra over ontologies has been proposed to support disciplined manipulation of domain knowledge re sources. However, if one tries to interoperate directly with the knowledge bases, semantic problems arise due to heterogeneity of representations. This heterogeneity problem can be eliminated by using an intermediate model that controls the knowledge translation from a source knowledge base. The intermediate model we have developed is based on the concept of abstract knowledge representation and has two components: a modeling behavior which separates the knowledge from its implementation, and a performative behavior which establishes context abstraction rules over the knowledge.

NETMARK: A schema-less extension for relational databases for managing semi-structured data dynamically

Object-Relational database management system is an integrated hybrid cooperative approach to combine the best practices of both the relational model utilizing SQL queries and the object-oriented, semantic paradigm for supporting complex data creation. In this paper, a highly scalable, information on demand database framework, called NETMARK, is introduced. NETMARK takes advantages of the Oracle 8i object-relational database using physical addresses data types for very efficient keyword search of records spanning across both context and content. NETMARK was originally developed in early 2000 as a research and development prototype to solve the vast amounts of unstructured and semi-structured documents existing within NASA enterprises. Today, NETMARK is a flexible, high-throughput open database framework for managing, storing, and searching unstructured or semi-structured arbitrary hierarchal models, such as XML and HTML.

Articulation management for intelligent integration of information

When combining data from distinct sources, there is a need to share meta-data and other knowledge about various source domains. Due to semantic inconsistencies and heterogeneity of representations, problems arise in combining multiple domains when the domains are merged. The knowledge that is irrelevant to the task of interoperation will be included, making the result unnecessarily complex. This heterogeneity problem can be eliminated by mediating the conflicts and managing the intersections of the domains. For interoperation and intelligent access to heterogeneous information, the focus is on the intersection of the knowledge. An algebra over domain is proposed which uses articulation rules to support disciplined manipulation of domain knowledge resources. The objective of a domain algebra is to provide the capability for interrogating many domain knowledge resources. The algebra formally supports the tasks of selecting, combining, extending, specializing, and modifying components from a diverse set of domains. The paper presents a domain algebra and demonstrates the use of articulation rules to link declarative interfaces for Internet and enterprise applications. In particular, it discusses the articulation implementation as part of a production system capable of operating over the domain described by the interface description language (IDL) of objects registered in multiple CORBA servers.

On Space Exploration And Human Error - A Paper on Reliability and Safety

NASA space exploration should largely address a problem class in reliability and risk management stemming primarily from human error, system risk and multi-objective trade-off analysis, by conducting research into system complexity, risk characterization and modeling, and system reasoning. In general, in every mission we can distinguish risk in three possible ways: a) known-known, b) known-unknown, and c) unknown-unknown. It is probable almost certain that space exploration will partially experience similar known or unknown risks embedded in the Apollo missions, Shuttle or Station unless something alters how NASA will perceive and manage safety and reliability.

A new uncertainty measure for belief networks with applications to optimal evidential inferencing

We are concerned with the problem of measuring the uncertainty in a broad class of belief networks, as encountered in evidential reasoning applications. In our discussion, we give an explicit account of the networks concerned, and call them the Dempster-Shafer (D-S) belief networks. We examine the essence and the requirement of such an uncertainty measure based on well-defined discrete event dynamical systems concepts. Furthermore, we extend the notion of entropy for the D-S belief networks in order to obtain an improved optimal dynamical observer. The significance and generality of the proposed dynamical observer of measuring uncertainty for the D-S belief networks lie in that it can serve as a performance estimator as well as a feedback for improving both the efficiency and the quality of the D-S belief network-based evidential inferencing. We demonstrate, with Monte Carlo simulation, the implementation and the effectiveness of the proposed dynamical observer in solving the problem of evidential inferencing with optimal evidence node selection.

A Bayesian approach to high resolution 3D surface reconstruction from multiple images

We present a radically different approach to the recovery of the three dimensional geometric and reflectance properties of a surface from image data. We pose the problem in a Bayesian framework, and proceed to infer the parameters of the model describing the surface. This allows great flexibility in the specification of the model, in terms of how both the geometrical properties and surface reflectance are specified. In the usual manner for Bayesian approaches it requires that we can simulate the data that would have been recorded for any state of the model in order to infer the model. The theoretical aspects are thus very general. We present rules for one type of surface geometry (the triangular mesh) and for the Lambertian model of light scattering. Our framework also allows the easy incorporation of data from multiple sensing modalities.

Intelligent Information Integration: Reclaiming the Intelligence

The authors present their work in the conceptualization, design, implementation, and application of “lean†information integration systems. They present a new data integration approach based on a schema-less data management and integration paradigm, which enables developing cost-effective large scale integration applications. They have designed and developed a highly scalable, information-on-demand system called NETMARK, which facilitates information access and integration based on a theory of articulation management and a context sensitive paradigm. NETMARK has been widely deployed for managing, storing, and searching unstructured or semi-structured arbitrary XML and HTML information at the National Aeronautics Space Administration (NASA). In this paper the authors describe the theory, design and implementation of our system, present experimental benchmark evaluations, and validate our approach through real-world applications in the NASA enterprise.

NASA's Aviation System Monitoring and Modeling Project

Within NASA’s Aviation Safety Program, the Aviation System Monitoring and Modeling (ASMM) Project addresses the need to provide decision makers with the tools to identify and evaluate predisposing conditions that could lead to accidents. This Project is developing a set of automated tools to facilitate efficient, comprehensive, and accurate analyses of data collected in large, heterogeneous databases throughout the National Aviation System. This report is a brief overview of the ASMM Project as an introduction to the rest of the presentations in this session on one of its key elements-the Performance Data Analysis and Reporting System (PDARS).

Dramatic Improvements to Feature Based Stereo

The camera registration extracted from feature based stereo is usually considered sufficient to accurately localize the 3D points. However, for natural scenes the feature localization is not as precise as in man-made environments. This results in small camera registration errors. We show that even very small registration errors result in large errors in dense surface reconstruction.We describe a method for registering entire images to the inaccurate surface model. This gives small, but crucially important improvements to the camera parameters. The new registration gives dramatically better dense surface reconstruction.