Computational model offers insight into mechanisms of drug-coated balloons.
Argus and Rover may be doing for urban map-making what the McNallys did for traditional cartography a century ago.
Argus, named after the 100-eyed creature in Greek mythology, is a high-resolution digital camera mounted on a small mobile platform and wheeled around campus by researchers and students from the MIT City Scanning Project. Argus incorporates specialized instrumentation to estimate, for each image acquired, the position and orientation of the camera in a single, earth-relative coordinate system. A new image is recorded every few seconds.
A smaller companion vehicle nicknamed Rover is used to acquire geo-referenced video images of interiors and exteriors. Argus and Rover are the prime tools in a multiyear project to develop an automated system for mapping urban environments.
Having experimented by recording images of buildings surrounding its headquarters in the Laboratory for Computer Science (LCS) in Technology Square, the project is now working with Facilities to develop a a three-dimensional model of the entire MIT campus.
The camera records images from many vantage points around the object at many times of day, in many types of weather and in different climatic conditions at various times of the year.
These many observations allow the construction of a 3D model of the region, consisting of geometry and appearance information. Observers can thus view the model under differing simulated lighting and atmospheric conditions, as well as from any perspective.
The real-world applications are many. Three-dimensional models could be a boon to tourists, emergency response workers, architects, assessors, urban planners, real estate professionals, builders and the military. Haptic or tactile interfaces to the model could be created to help sightless people work their way around a strange environment.
To develop the map of the campus, aerial views were necessary to incorporate rooftops. A single-engine propeller airplane equipped with the digital camera and navigational instruments buzzed the campus at an altitude of 1,000 feet for several days last December. The plane, owned by Earth Data Aviation of Gaithersburg, MD, made several hundred passes over the campus and acquired more than 1,200 images.
Among the devices on board were a prototype high-resolution (16-megapixel) digital camera manufactured by Lockheed-Martin, and navigation instrumentation prototyped by the Ohio State University Center for Mapping, with whom MIT's group is collaborating.
The principal investigator on the project is Seth Teller of the LCS Computer Graphics Group, the X-Consortium Career Development Associate Professor of Computer Science and Engineering who joined the Department of Electrical Engineering and Computer Science in 1994.
Noting that traditional map-making is tedious and requires "significant human effort," Professor Teller said the project will make it possible to substitute sensors and computers in the process, making it more efficient, faster and more flexible. Acquiring enormous quantities of imagery and navigation data and applying computer vision techniques will enable the automated production of 3D maps, or computerized models, of the world, he said.
This approach "will eliminate much of the sweat that goes into authoring 3D models of urban areas," said Professor Teller, who studied physics at Wesleyan University and did graduate work in computer graphics at the University of California at Berkeley, where he created a model for a new building that "allowed design evaluation to be done through virtual inspection." He did postdoctoral research in Israel and at Princeton before coming to MIT.
A number of UROP and graduate students have contributed to the City Scanning Project since its inception. They include current EECS graduate students Matt Antone, Mike Bosse, George Chou (SM 1993), Barbara Cutler, Manish Jethwa, J.P. Mellor, Laughton Stanley and Qixiang Sun; EECS senios Adam Kropp and Ning Ye; EECS junior Deborah Tran; EECS sophomores Johann Burgert and Avishai Geller; Jesus Orihuela, a sophomore in architecture; and freshman Eric Cohen. Other who have been involved are Eric Amram (SM 1998), Douglas De Couto (SB 1997), Satyan Coorg (PhD 1998), Tara Schenkel (SB 1998, SM), research scientist Neel Master and visiting scientist Stefano Totaro.
Funding for the project has been provided by Lincoln Laboratory's Advanced Concepts Committee, the Office of Naval Research, the Defense Advanced Research Projects Agency, Intel Corp. and Interval Corp.
A version of this article appeared in MIT Tech Talk on February 24, 1999.