Career Interests:

Research and teaching in digital image processing, computer vision, and computer architecture. Areas of particular interest are automated and interactive object recognition, three dimensional structure and motion estimation in visible imagery and radar, feature extraction in remote sensing, and applications of image processing and computer vision in science and industry.

Education:

Employment History

Senior Group Engineer, Lockheed Martin Astronautics November 1995 to present

Staff Engineer, Lockheed Martin Astronautics October 1990 to November 1995

Graduate Research Assistant, University of Illinois August 1981 to October 1990

Undergraduate Hourly, University of Illinois July 1977 to August 1981

Research Experience

Interactive Exploitation Tools for Image Analysts
A small classified program and the INIMEX program, both managed by Dr. Debrunner, have incorporated components of the ARAGTAP and MSTAR model-based automatic target recognition (ATR) systems into interactive target recognition tools for the intelligence imagery analyst. The focus-of-attention (FOA) modules were incorporated as a screening tool which alerted the analyst to image locations containing likely targets. The prediction and matching portions of the hypothesize and test module will be adapted to use mixed human-machine control to guide the search. The user can select the exact orientation, configuration, and articulation of the model to be predicted, and will receive feedback in the form of a simulated image and a mismatch map, showing the regions of poor agreement between the simulation and the real image. In addition, since it is often difficult for inexperienced image analysts to relate features in a SAR image to object parts, the system provides visual tie-back from the simulated image to an optical rendering of the target.

Model-Based ATR in Synthetic Aperture Radar
The ARAGTAP and MSTAR programs develop ATR systems which use an on-line target image simulation capability to predict the target in any possible orientation, configuration, and articulation. This provides a complete and nearly continuous coverage of the space of target images without the cost of storing sample images at all possible target orientations, configurations, and articulations. The system designs consisted of a focus-of-attention module to select image regions containing potential targets, an indexing module to produce a set of possible target types, and a predict-extract-match module to refine the target hypotheses provided by the indexer. Dr. Debrunner’s role on these programs has been to support feature extraction, system design, evaluation strategies, geometric modeling of the SAR sensor, and indexer design.

Image Understanding for 3D Model-Based Fixed-Site Exploitation
Under the MOSTAR program, Dr. Debrunner and his team incorporated context information available from 3D site models into the ATR process to improve ATR performance. Until its completion in June of 1996, Dr. Debrunner also managed the Denver component (focusing on SAR exploitation) of the DARPA/ORD RADIUS II program. The goal of this program was to demonstrate that mature image understanding technology together with 3D site models can aid image analysts in the exploitation of overhead imagery. The final version of the RADIUS testbed was delivered and successfully demonstrated to the government in July of 1996. Under the D03R internal research and development (IRAD) project at Lockheed Martin, Dr. Debrunner investigated applications of ARAGTAP target recognition technology in other sensor domains, developed an image-to-3D-model registration system for synthetic aperture radar imagery.

Geometric Modeling of Imaging Sensors
Any algorithm which seeks to exploit 3D models of objects in imagery needs a mathematical model of the projection from 3D space to 2D image space. Dr. Debrunner developed a such a model of a synthetic aperture radar sensor which was incorporated into the RADIUS testbed under the MOSTAR contract and was later enhanced under the INIMEX project.

3D Motion and Structure Estimation
Dr. Debrunner's Ph.D. research addressed the problem of recovering the 3D structure (i.e. shape) and motion of moving objects from a long sequence of images. The work involved tracking feature points over many images, partitioning the feature points into sets corresponding to the different objects, and calculating the structure and motion parameters of the objects. This information could be used by an object recognition system to identify particular objects, or by a robot control system to position the robot's gripper for grasping the object.

Image Processing Algorithms for Pyramid Architectures
Dr. Debrunner's M.S. research involved design, analysis and simulation of computer vision and image processing related algorithms for pyramid architecture multi-processors.

Computer Music
As a member of the Computer Music group at the Computer-Based Education Research Laboratory, University of Illinois, Dr. Debrunner designed and built the electronics for a velocity-sensitive 88-key music keyboard. This included designing circuitry for reading the optical switches which detected key motion and circuitry for a Z80-based controller and interface board. It also involved writing the Z80 assembly code which detected the key changes and communicated with the Intel 8080 in the host PLATO V terminal. In addition, Dr. Debrunner contributed to the preliminary design of a sum-of-sines synthesizer and wrote assembly language and Tutor language utility programs for the IMS music editing and synthesis system.

Teaching Experience

Adjunct Professor Fall 1996, Fall 1997
Colorado School of Mines
Taught a graduate course on computer vision (EGES 512) based on Dr. William Hoff’s course supplemented with additional topics including projective geometry and motion factorization. In fall of 1997 the teaching was shared with Dr. Hoff.

Instructor Fall 1993
Martin Marietta (now Lockheed Martin) Evening Institute
Developed and taught an advanced course titled Computer Vision and Image Processing (much of the computer vision material was taken from Dr. William Hoff’s course at University of Colorado at Denver). The course covered a wide range of topics in image processing and computer vision including compression, sampling, filtering, enhancement, tomographic imaging, feature extraction, edge detection, segmentation, motion, stereo, 3D imaging, and object recognition.

Management Experience

INIMEX Program Manager July 1997 to present
Managed technical effort to develop an image analyst’s workstation. Responsible for all aspects of running the program, including budgeting, supervising a staff of five, managing a subcontractor, presenting plans and results to government sponsors, and technical direction.

MOSTAR Program Manager August 1995 to October 1997
In response to a sole source solicitation from the RADIUS customer, wrote a proposal and negotiated a statement of work. Managed all aspects of the contract, including soliciting, negotiating, and managing a subcontract, supervising a staff of three, holding reviews with the government, budgeting, and planning. The government customer rated the team's performance as excellent in a formal evaluation at the conclusion of the contract.

RADIUS Phase II Program Manager February 1995 to June 1996
Managed a subcontract from Lockheed Martin Management and Data Systems. Responsibilities included supervision of a technical staff of two, budgeting, planning and holding reviews with the customer.

IRAD Principal Investigator May 1993 to December 1996
Managed the D03R internal research and development (IRAD) project supervising a technical staff of up to two, presenting results to Lockheed Martin management twice yearly, and justifying continued funding.

Classified Project Technical Lead and Program Manager July 1993 to July 1994
Functioned as technical lead for most of project, and took over as manager for the last three months of the project, with responsibility for technical direction, supervision of a staff of three, budgeting, planning, and presentations to the government customer.

Proposal Experience

Contributed to numerous winning proposals at Lockheed Martin including:

INIMEX proposal
Responsible for the technical volume. This proposal described plans for an image analyst’s tool which incorporates many advanced capabilities including automatic target recognition and interactive target recognition tools based on MSTAR technologies. A government reviewer remarked that this was the best proposal he had seen in a long time.

RADIUS II proposal
Managed the proposal effort in support of Lockheed Martin Management and Data Systems, including a demonstration of capabilities shortly after submission. This lead to an award on RADIUS II (for which I managed the subcontract) and later to a sole source solicitation from the same customer in 1995 for the MOSTAR proposal.

GOLDPAN 93 proposal
Responsible for portions of the technical volume of a proposal to incorporate ATR technologies into a field demonstration.

Professional Activities

Associate Editor for the journal Pattern Recognition (1994 to present)

Member of the IEEE Computer Society (1983 to present)

Reviewer for several conferences and journals (1984 to present) including:

Organizer of the first Lockheed Martin Conference on Intelligence Imagery Analysis

Hobbies

Underwater Hockey (represented U.S. in world competition, South Africa, 1996).

Woodworking and remodeling.

Hiking, bicycling, and cross-county skiing.