This CAREER project will advance the ability to extract spatial information from digital aerial imagery for which prior or complementary information already exists. Examples of such information are pre-existing digital maps, digital terrain models, and spatial information systems in general. We use geospatial databases to support and guide digital image analysis methods for object extraction. The work will accomplish both research and educational objectives. The research program develops novel concepts and algorithms for the extraction of information from digital images. Emphasis is put on matching images to abstract representations of reality, change detection, analysis of scale differences between various images and images and databases, and structuring the extracted information in order to be suitable for incorporation into geospatial databases for their updating. A workshop is also planned with the participation of leading experts from a variety of disciplines relevant to this project. The educational initiatives will incorporate the research advancements made in this project into our graduate and undergraduate curriculum through the development of a new course and the modification of two existing courses. Furthermore, findings of the project will be incorporated in our high school outreach program, to expose high school junior classes to the challenging and evolving role of digital images in spatial information engineering.

Our objective is to improve information extraction processes from digital aerial images by having image analysis methods supported and guided by pre-existing/complementary spatial information.

The research program develops novel concepts and algorithms for the extraction of information from digital images. The research plan is organized in complementary tasks; more specifically:

• matching/comparing images to existing databases for change detection, i.e. identifying new, modified, or no longer existing objects; (this task has commenced and will be addressed during years 1 and 2 of this project);
• analysis of scale differences between database information and images, and the impact these variations have on the relevant image analysis techniques; (this task will commence on the second year of this project) and
• structuring the extracted information in order to be suitable for incorporation into the geospatial databases for their proper updating, and using metadata to enhance and improve the databases to support digital image analysis processes (this will be addressed during years 3 and 4 of this project).

During the second year of the project we will hold a 3-day workshop with the participation of approximately 10 experts from the fields of digital image analysis, photogrammetry, GIS, databases, and digital libraries.

Parallel to the aforementioned research activities, this project includes educational initiatives, designed to take advantage of and incorporate the research advancements made in this project. More specifically:

• a new undergraduate course will be developed, on the "Integration of Digital Image Processing and Geospatial Databases";
• on graduate level, experiences acquired in this project will be incorporated in the current courses of  Digital Image Processing & Analysis, and Analytical & Digital Photogrammetry, taught by the PI; and
• the high school outreach program of the Department of Spatial Information Science and Engineering will be enriched and expanded as a result of this project, to expose high school junior classes to the challenging and evolving role of digital images in spatial information engineering.

As mentioned before, we are still at a very early stage in this project, i.e. only halfway through year 1. In general, Year 1 focuses on the development of a matching scheme which uses as input database object models and is applied on digital imagery for object extraction and change detection. At this stage, the application focuses on monoscopic scenaria (processing a single image against a database). Year 1 of the project also addresses the initial development of material for a new undergraduate course to be offered by the PI in Year 3 of the project.

Both issues proceed on schedule. We have developed the basic theoretical part of our matching scheme and continue with its implementation and extension. We have been able to extend matching from an image-to-image operation into an image-to-information operation.

Combined, the issues addressed in this project (mentioned above under "Goals, Objectives, and Targeted Activities")  will substantially advance science in digital image processing and analysis, and will complement parallel advancements in a variety of related disciplines, most notably digital libraries, geographic information systems, and remote sensor technology.  The project will offer a fresh view to both digital image analysis and GIS database management and updating. It puts emphasis on information, and promises novel theoretical approaches to its extraction from digital imagery and management within GIS.

The project will advance science in two manners, i.e. by developing both

• new algorithms, and a
• new approach

• significant progress towards image understanding for topographic applications by improving information flow among various geospatial databases;
• substantial betterment in automating digital image processing by employing photogrammetric concepts and spatial information environments; and
• advancements in spatial information engineering by strengthening the role of digital imagery as a data collection tool.

• Human Resources: This project supports the graduate studies of two Ph.D. candidates. Furthermore, educational developments resulting from this work will affect one undergraduate and two graduate courses, with an average yearly participation of approximately 15 students per class. Work covered by this project is also part of the high school outreach program ("Spatial Horizons") of our department. This program is typically attended by 100 or more high school students per year. Another high school outreach program affected by this project is the University-organized "Expanding your Horizons" program, which is directly addressing female high school students, attempting to increase their participation in science and engineering disciplines.
• Education and curriculum development: This project includes educational initiatives, designed to take advantage of and incorporate the research advancements made in this project. More specifically:
• a new undergraduate course will be developed, on the "Integration of Digital Image Processing and Geospatial Databases";
• on graduate level, experiences acquired in this project will be incorporated in the current courses of  Digital Image Processing & Analysis, and Analytical & Digital Photogrammetry, taught by the PI; and
• the high school outreach program of the Department of Spatial Information Science and Engineering will be enriched and expanded as a result of this project, to expose high school junior classes to the challenging and evolving role of digital images in spatial information engineering.
• Department/institution infrastructure: The project further supports collaboration between the PI and GIS-oriented faculty within the PI's Department, and GIS-relevant faculty within the University of Maine. One such collaboration resulted in the awarding of a Center of Excellence in Remote Sensing Applications grant by NASA. The awarded grant, of approximately $300,000, supports the improvement of our digital image analysis laboratory through the acquisition of substantial modern computing facilities to complement the existing ones. The grant was augmented by software donations from digital photogrammetric companies, valued at approximately $150,000.
• Industry: We are still at a very early stage regarding this project. However, some initial discussions with leading industrial firms (e.g. Leica/Helava, Vision International) in the fields of digital image analysis and softcopy photogrammetry have confirmed their interest in technology transfer of our work.

The project is still at an early stage. The following existing references offer an introduction to the project, while the planned ones offer further details on some specific project tasks:

• A. Stefanidis and P. Agouris: "Digital Photogrammetry: The Integration Challenge", GIM International Journal of Geomatics, Vol. 11, No. 1, January 1997, pp. 67-69.
• P. Agouris, A. Stefanidis and K. Beard: "Digital Photogrammetric Techniques for Transportation Data Acquisition and Management", Transportation Research Record No. 1599, pp. 111-117, November 1997.
• P. Agouris, S. Gyftakis, and A. Stefanidis: "Using a Fuzzy Supervisor for Object Extraction within an Integrated Geospatial Environment". Accepted for publication in the International Archives of Photogrammetry & Remote Sensing. To be presented at the Symposium on Object Recognition and Scene Classification from Multispectral and Multisensor Pixels, organized by the International Society for Photogrammetry & Remote Sensing, Columbus, Ohio, July 1998.
• P. Agouris, A. Stefanidis, and J. Carswell: "Intelligent Retrieval of Digital Images from Large Geospatial Databases". Accepted for publication in the International Archives of Photogrammetry & Remote Sensing. To be presented at the Symposium on Object Recognition and Scene Classification from Multispectral and Multisensor Pixels, organized by the International Society for Photogrammetry & Remote Sensing, Columbus, Ohio, July 1998.

Digital photogrammetry is an important field of digital image analysis. It deals with the processing of digital images for the extraction of metric quality information on objects/scenes depicted in them, without having to get into physical contact with the object/phenomenon under study. Typical digital photogrammetric applications employ aerial, space, or even close-range imagery.  Applications range from mapping and environmental monitoring to medical and biological image processing, intelligence and defense activities,  and even flow monitoring. In this project we deal with aerial images representing scenes for which prior or complementary information already exists in geospatial databases. Geospatial databases (e.g. digital maps, topographic databases etc.) contain qualitative and quantitative information on the precise location, function, and interrelationships of man-made objects (e.g. roads, buildings) and natural objects (e.g. bodies of water, or even the terrain itself). Aerial images are clearly ideal complements to geospatial databases as data capture sources, since image-captured terrain scenes can be used to extract accurate and up-to-date qualitative and quantitative spatial information. The issue of object extraction from digital images is a major research topic in computer vision and digital photogrammetry.

• A. Gruen, O. Kuebler, and P. Agouris (eds.) "Automatic Extraction of Man-Made Objects from Aerial and Space Images", Birkhaeuser Verlag, Boston, 1995
• A. Gruen, E.P. Baltsavias, and O. Henricsson (eds.) "Automatic Extraction of Man-Made Objects from Aerial and Space Images Vol. II", Birkhaeuser Verlag, Boston, 1997
• Proceedings of the recent ARPA Image Understanding Workshops

• Retrieval of digital images and other types of GIS and cartographic data  from integrated spatial databases using matching techniques, topological  relations, metadata, and database management systems.
• Object extraction from digital imagery in medical and biological applications, processing CAT scans, mammograms, X-rays etc.
• Rapid scene modeling and updating using hand-held cameras in dynamic  environments. Potential for support by wireless communications for defense and intelligence applications.