A Networked Computing Environment for the Manipulation and Visualization of Geometric Data
Lawrence Wolff, Yair Amir, Michael Goodrich, Rao Kosaraju, Subodh Kumar, Roberto Tomassia(Brown Univ.), Russ Taylor, David Yarowsky
Department of Computer Science
Johns Hopkins University was founded in 1879 and was the first United States institution to implement the "German system" of formal graduate training as a research apprenticeship under the tutelage of a mentor. A few years later Harvard followed the lead of Johns Hopkins University to become the second US institution to implement such a system. The Department of Computer Science resides in the Whiting School of Engineering and was founded in 1986 with Gerald M. Masson the founding and current chairman.
Manipulation and computation involving large datasets of geometric information is becoming an increasingly integral part of a number of research areas in Computer Science. A diversity of workstations and PCs exist today that provide a wide range of capabilities for geometric computation and graphics visualization. A main goal of this project is to develop a core infrastructure for geometric computing providing a unified environment for five research laboratories in the Department of Computer Science at Johns Hopkins University to perform: (i) independent and collaborative research within a given laboratory, (ii) inter-laboratory collaborative research and/or share software resources, both within the Computer Science Department and between Computer Science and other parts of the University, (iii) inter-laboratory collaboration between the Johns Hopkins Computer Science Department and other universities on the World Wide Web. This geometric computing environment would not only provide state-of-the-art computation and visualization tools, but would also facilitate interfacing existing workstations in respective laboratories with such state-of-the-art tools, and allow the interfacing of additional purchased equipment in the future. This project started August 1997 and details can be found on the web at http:\\www.cs.jhu.edu\\ri .
One research area being explored by the Center for Geometric Computation (Michael Goodrich, Rao Kosaraju, Roberto Tamassia) is the development of an interactive interface for geometric computation over the internet. The basic idea is that geometric computing servers execute a variety of geometric algorithms on behalf of remote clients supporting applications which can include execution of algorithms, checking consistency of geometric structures, experimental study and comparison of algorithms, design of new algorithms, interactive electronic books, educational demonstrations, and electronic commerce for CAD engineering companies. The research in the Computer Graphics Laboratory (Subodh Kumar in collaboration with Yair Amir) focuses on interactive visualization of large and complex geometric environments. It includes investigation of high-order geometry representation, efficient visibility computation, high-dimensional space visualization, collaborative design paradigms, and distributed graphics and intuitive user navigation. Efficient utilization of a network and distributed computing resources is made by dynamically allocating computation to processors. Such associative computing achieves much higher utilization of resources than is readily achieved if each workstation is tied to one application program.
The Computer Vision Laboratory (Lawrence Wolff) is involved with research projects requiring both very large geometric datasets and very intensive computational operations on some of these datasets. Research in Automatic Target Recognition requires visualization of signatures produced from ray traced rendering of CAD models of targets utilizing sophisticated extended illumination and reflectance modeling. 3-D geometric feature extraction methodologies on large volumetric
datasets obtained from stacked 2-D CAT scans of lung and brain anatomy are being used for mapping out complex lung bronchial and vascular tree structures, and, complex folding of tissue in brain anatomy, as well as for researching the correlation of anatomical geometry with disease.
There is much interaction with clinicians in the Department of Radiology at the Johns Hopkins School of Medicine who are providing empirical datasets.
The Computer Integrated Surgery Laboratory (Russ Taylor) has a critical need for high performance geometric computation and visualization. One important area being researched is the development of techniques for image and model registration that use various 2D imaging modalities (video, x-ray, fluoroscopy, ultrasound) to accurately locate 3D anatomical structures in the operating room. Geometric algorithms are also being used for treatment planning in optimizing a pattern of radiation seeds to be implanted in diseased tissue. These incorporate numerical finite element simulations. Navigation and guidance during neurosurgery and lung surgery is another important research area involving complex inter-dependent geometric relationships between tools used by the surgeon, and patient anatomy.
A central problem of Natural Language Processing (David Yarowsky) is visualizing the relationships between datapoints in a high-dimensional space. This includes document-space visualization for information retrieval, clustering algorithms for high-dimensional word and document spaces, and, 3-D visualization tools in high-dimensional word and document spaces.
This past year we have purchased with the NSF RI grant a variety of SGI workstations (Octanes, O2's) centered about a powerful Onyx2 with 4 R10000 processors. These have been networked into a single unified system shared by 5 laboratories in the department of computer science. This has had a huge impact particularly on our ability to perform geometric computations, particularly those using sophisticated computer graphics and real-time video overlays. In addition we have achieved our first year goal of upgrading our departmental network from 10mbs to 100mbs.
A diversity of researchers currently utilize this networked system, which beyond the PIs include postdocs, PhD students, MS students and undergraduates. A number of class projects and on-going summer projects heavily utilize this system.
With Johns Hopkins matching funds we have hired a system administrator/manager who has aided in the strategic planning, design, purchase and assembly of our networked system of SGIs. Alot of time and effort was expended by the PIs in conjunction with the system administrator in deciding upon the right equipment for the right networked environment. A number of PhD students also aided in assembling the network.
The NSF CISE Research Infrastructure grant was key in recruiting Jon Cohen (PhD UNC 1998) this year as a new assistant professor and strengthening our computer graphics group in the department of computer science. The RI grant also significantly contributed to the proposal effort for the Computer Integrated Surgery, Engineering Research Center.
The continuing development of GeomNet, a web based system being developed by the Center for Geometric Computation facilitating easy access to Computational Geometry algorithms, is exemplary of a growing trend towards collaborative research on the internet. This in turn is influencing how the research community is working together in ever more effective ways. The expectation someday is to hold meetings, workshops and conferences without ever leaving your office and yet experience the full fidelity of physically attending such an event.
Continuing developments in the Computer Integrated Surgery Laboratory are revolutionizing the way surgery is performed in the operating room. The capabilities of a surgeon augmented by robotic/visualization devices controlled by efficient real-time computational algorithms will immensely impact health care worldwide in reduced costs for treatment and recovery, reduced mortality and better quality of life.