Comparison of Methods and Approaches for Combining MultiTemporal Raster Data in Geographic Informati

1
Comparison of Methods and Approaches for
Combining Multi-Temporal Raster Data in
Geographic Information Systems

• Gabe Emerson
• Chris Heuer
• CSci 8715

2
Motivation

• No literature surveys on raster-based approaches
• Novelty Create classification scheme for
  multi-temporal raster techniques
• Evaluation of techniques based on
• Efficiency
• Cross-application portability
• Usefulness
• Identify similarities in techniques
• Suggest combinations of methodologies to yield
  additional improvements

3
Methodology Classification Scheme

• Snapshot
• Raster data from multiple times as static images
• Simple, uses little processing
• Visualization
• Simplest form animates snapshots in a sequence
• More complex forms can show cause and effect
  relationships
• Temporal Averaging
• Data from multiple times averaged to portray mean
  for time period being studied
• Average on pixel level or on larger scale by
  averaging spatial coverage of like-valued regions
• Compositing
• Combing multi-temporal data sets for maximum
  imagery detail
• Allows features visible at one time to be
  displayed with features that may obscure them at
  a different time

4
End-use Classification Scheme

• Tracking
• Multi-temporal data is about tracking change over
  time
• Useful by itself but can be used to produce the
  following
• Prediction
• Show probable future values based on history of
  previous values
• Change in data over time can be traced to show
  patterns and dynamics
• Completing
• Fills gaps in data for areas invisible or
  obscured at one time by inference from another
  time
• Does not always hold for night/daytime completing
  but works well for land obscured by clouds
• Abstracting
• Abstracting or simplifying multi-temporal data
  into a generalization about the features being
  studied
• Data can be used to make statements about overall
  change trends, patterns can be derived, or
  statistical data can be produced that does not
  require detailed knowledge of original data
  elements

5
Snapshot

• Remote Sensing and GIS methods for dynamics
  studies in the Caspian Sea Coastal Zone by
  Baldina and Labutina 1
• Authors utilized date from aerial/satellite
  images and maps
• Georeferencing difficult as sources had variation
  in angular distortion, spatial resolution, and
  coverage extent
• Created common vector base map with most easily
  identifiable geographic features

6
Visualization

• GIS and Remote Sensing techniques for the
  assembly of a database characterizing the
  spatio-temporal character of the South Australian
  continental shelf environments by Bryan, Kinloch,
  and Gerner 3
• Attempt to measure biological activity through
  quantity of chlorophyll a in the water

• Temporal Zooming by Hornsby 8
• Temporal data can be displayed at varying levels
  of detail/resolution in the same way spatial data
  can be zoomed in and out
• Operations defining object creation, destruction,
  "reincarnation", splits, joins, and other
  possible behaviors of objects can be expressed
  and stored

7
Temporal Averaging

• Bryan, Kinloch, and Gerner again 3
• Used monthly averaging on sea-surface color data
  to produce yearly composites of cloud-free
  imagery
• Mapping the land cover of the forested area of
  Canada with Landsat data by Wulder 5
• Mosaic of images combining leaf-on and leaf-off
  should be produced in order to aid seasonal
  land-cover classification
• Averaging the set of images over time means
  short-term effects of weather can be reduced and
  a greater number of cells can be assigned
  near-true values

• Spatio-Temporal Aggregates over Raster Image Data
  by Zhang, Gertz, and Aksoy 10
• Continuous stream of raster data (such as a live
  video or image stream transmitted by a remote
  sensor) can be averaged into discrete time or
  duration-based values
• Discuss applications which take averages based on
  the stream aggregation

8
Composition

• Integration of multi-seasonal remotely-sensed
  images for improved landuse classification of a
  hilly watershed using geographical information
  systems by Adinarayana and Rama Krishna 2
• Monsoon season caused features to be obscured
• Developed classification rules for assigning
  values to land cover based on supervised
  classification raster set from monsoon season and
  raster set from post-monsoon season

• Comparison of Methods for Estimation of Kyoto
  Protocol Products of Forests From Multitemporal
  Landsat by Goodenough, Bhogal, Chen, and Dyk 7
• Used leaf-on and leaf-off images to separate
  mature deciduous trees from 2nd-growth vegetation
• Ground-vegetation classification accuracy of
  91-95 compared to 81-90 for leaf-on only data
  sets

9
Composition cont.

• Photogrammetric and GIS techniques for the
  development of vegetation databases of
  mountainous areas Great Smokey Mountains
  National Park by Welch, Madden, and Jordan 6
• Focuses on land cover and vegetation
  classification for a mountainous and largely
  inaccessible national park
• Used aerial photos from leaf-on and leaf-off
  periods to develop a detailed database of
  vegetation cover

• A Web-based browsing and spatial analysis system
  for regional natural resource analysis and
  mapping by Vatsavai, Burk, Wilson, and Shekhar
  4
• Pre-processing of common spatial and temporal
  analysis functions can reduce the overall
  per-request server-side and client network
  resource usage
• Zhang, Gertz, and Aksoy again 10
• Solves problem of partial/missing image data by
  computing a number of answers for segments of the
  query region in which some image data only
  contributes partially

10
Comparison
11
Validation

• End-use classification falls short due to
• Techniques utilized to reach end goal may be
  incompatible with available data
• Choosing a system based on end-use can limit
  functionality of the system
• Forced to pigeon-hole projects into one or more
  categories

• Methodology classification best choice for this
  survey
• Allows a higher degree of abstraction when
  comparing so that multiple projects can
  potentially be accurately classified in multiple
  ways
• Can show multiple types of research in a wide
  range of end uses, as well as a range of
  cross-method approaches

12
Conclusions

• Common Elements
• "Temporal Averaging papers attempt to reduce the
  effect of imaging outliers on the eventual raster
  cell assignments
• Two 3,5 deal with raw weather imagery
• Three Compositing papers 7,2,6 are working
  towards improving data context from dissimilar
  sets acquired in different seasons
• Creation of "all-season" classifier for land
  cover based on leaf-on and leaf-off requires
  complex rule-based analysis of the effects of
  different types and families of vegetation
  visible at different times
• 5 some similarities to 7,6 in that these
  methods are designed to handle combinations of
  leaf-on/leaf-off raster data
• Averaging approach used in 5 closer to 3
• Areas for Improvement
• Methods described in several of these papers
  could be enhanced by combination with other
  methods investigated

13
References

• Baldina and Labutina, (2002) Remote sensing and
  GIS methods use for dynamics studies in the
  Caspian Sea coastal zone in Geoscience and
  Remote Sensing Symposium, 2002. IGARSS '02. IEEE
  International. Volume 5. Pages 2838-2840
• Adinarayana and Krishna. (1996) Integration of
  multi-seasonal remotely-sensed images for
  improved landuse classification of a hilly
  watershed using geographical information systems
  In International Journal of Remote Sensing,
  Volume 17, Issue 9, Pages 1679 1688
• Bryan, Kinloch, and Gerner (2003) GIS and Remote
  Sensing techniques for the assembly of a database
  characterizing the spatio-temporal character of
  the South Australian continental shelf
  environments In Proceedings of Coastal GIS 2003
  7 July 2003, University of Wollongong,Wollongong,
  Australia.
• Ranga Raju Vatsavai, Thomas E. Burk, B. Tyler
  Wilson, Shashi Shekhar. (2000) A Web-based
  browsing and spatial analysis system for regional
  natural resource analysis and mapping, In
  Proceedings of the 8th ACM international
  symposium on Advances in geographic information
  systems, Pages 95-101.
• Wulder, M. (2002) "Mapping the land cover of the
  forested area of Canada with Landsat data", In
  Geoscience and Remote Sensing Symposium, Volume
  3, Pages 1303 - 1306.
• Welch, R. M. Madden, and T Jordan. (2002).
  "Photogrammetric and GIS techniques for the
  development of vegetation databases of
  mountainous areas Great Smokey Mountains
  National Park". In ISPRS Journal of
  Photogrammetry and Remote Sensing. Volume 57,
  Pages 53-68.
• Goodenough, D. G., A.S. Bhogal, H. Chen, and A.
  Dyk. (2001). "Comparison of Methods for
  Estimation of Kyoto Protocol Products of Forests
  From Multitemporal Landsat". In Geoscience and
  Remote Sensing Symposium, 2001. IEEE
  International.Volume 2, Pages 764 - 767
• Hornsby, K. (2001). "Temporal Zooming". In
  Transactions in GIS Volume 5, Issue 3, Pages
  255-272
• Jeong, S-H, A. Fernandex, N. W. Paton, and T.
  Griffiths (2005). "An Experimental Performance
  Evaluation of Spatio-Temporal Join Strategies" In
  Transactions in GIS Volume 9 Issue 2. Pages
  129-156
• Zhang, J. , M. Getz, and D. Aksoy (2004).
  "Spatio-Temporal Aggregates over Raster Image
  Data" In Proceedings of the 12th annual ACM
  international workshop on Geographic information
  systems. Pages 39-46.

14
Questions?