hybridization: it proceeds by comparing nodes and prongs to
establish links between helix pairs (Fig. 5).
This involves the comparison of their corresponding records as
they were presented in Section 2. Similarities can be identified
in instances where the corresponding records (e.g. azimuth
values or positions) differ less than an acceptable threshold (see
e.g. the buffer zones in Fig. 6).
Buffer in
position
Buffer in
rotation
Figure 6: The effect of buffering in spatiotemporal helix
comparison when comparing two spatiotemporal
helixes (one represented by a continuous line, the
other by a dashed line).
A similarity metric S is then provided as a coincidence
percentage:
S — (duration of coincidence)/(duration of event), (3)
Where duration coincidence is the aggregate time during which
the two events were displaying similar properties (e.g. both
were pointing North). In order to support this comparison, the
range of values of each property may be tessellated in few
subsets. For example, azimuth information may be presented as
4 (N, W, S, and E) or even 8 (adding NE, SE, SW, NW) discrete
directions as opposed to 360 discrete degrees.
6. COMMENTS
In this paper we introduced the concept of spatiotemporal helix
as a model of spatiotemporal events. It allows us to model
efficiently changes in the location and extent of a phenomenon,
and supports the comparison of events to identify similarities
and complex relationships among them. This comparison of
spatiotemporal helixes allows us to produce meaningful
qualitative metrics to what up to this point have been considered
as quantitative queries. While our motivation is the analysis of
events as they are captured in motion imagery datasets, the
concept of the spatiotemporal helix can be applied to any type
of multitemporal datasets with spatially registered information
(e.g. land use patterns as they are depicted in a multitemporal
sequence of maps, disease spread as it is recorded in a GIS etc.).
REFERENCES
Agouris P., A. Stefanidis & S. Gyftakis, 2001. Differential
Snakes for Change Detection in Road Segments,
Photogrammetric Engineering & Remote Sensing, 67(12), pp.
1391-1399.
Kass M., A. Witkin, & D. Terzopoulos, 1987. Snakes: Active
contour models, in Proceedings of the Ist Int. Conf. on
Computer Vision, London, pp. 259-268
Kohonen, T., 1982. Self-organized Formation of Topologically
Correct Feature Maps. Biological Cybernetics: 59-69.
Kollios G., V. Tsotras, D. Gunopulos, A. Delis, & M.
Hadjieleftheriou, 2001. Indexing Animated Objects Using
Spatiotemporal Access Methods, /EEE TKDE, 13(5): 758-777.
Partsinevelos P., A. Stefanidis & P. Agouris, 2001. Automated
Spatiotemporal Scaling for Video Generalization, IEEE-ICIPOI
(International Conference on Image Processing), Thessaloniki,
Vol. 1, pp.177-180.
Pfoser D., & Y. Theodoridis, 2000. Generating Semantics-
Based Trajectories of Moving Objects, Intern. Workshop on
Emerging Technologies for Geo-Based Applications, Ascona.
Smith M. & T. Kanade, 1995. Video Skimming for Quick
Browsing based on Audio and Image Characterization, Tech.
report CMU-CS-95-186, Carnegie Mellon University.
Stefanidis A., P. Partsinevelos. K. Eickhorst & P. Agouris,
2001. Spatiotemporal Lifelines in Support of Video Queries,
Proceedings DEXA Workshop 2001, pp. 865-869, Munich.
Sistla A.P., O. Wolfson, S. Chamberlain, & S. Dao, 1997.
Modeling and Querying Moving Objects. /CDE, pp. 422-432
Tao Y., & D. Papadias: 2001. MV3R-Tree: A Spatio-Temporal
Access Method for Timestamp and Interval Queries. VLDB
2001: 431-440.
Vazirgiannis M., & O. Wolfson, 2001. A Spatiotemporal Query
Language for Moving Point Objects, SSTD '01, Los Angeles,
pp. 20-35.
(2)
Wolfson O., A.P.Sistla, B. Xu, J. Zhou, & S. Chamberlain:
DOMINO: Databases fOr MovINg Object tracking. SIGMOD
Conference 1999, pp. 547-549.
Yeung M., & B.L. Yeo, 1997. Video Visualization for Compact
Presentation and Fast Browsing of Pictorial Content, IEEE
Trans. on CSVT, 7(S), pp. 771-785.
ACKNOWLEDGEMENTS
This work was supported by the National Science Foundation
under grants number DG-9983445 and IIS-0121269. We would
like to thank Mr. Sotiris Gyftakis for providing Figure 3.
—518—