Ferris State University

 

 

 

National Geospatial-Intelligence Agency

MULTISOURCES DATA FUSION, A MATHMATICAL AND COMPUTATIONAL APPROACH

Supported by: The National Geospatial-Intelligence Agency

Research team:

Prof. Yaron Felus and Prof. Robert Burtch

Students: Chad M. Schaeding, Brian Romsek, and Krag Caverly

Surveying Engineering Department

Ferris State University

 

Abstract:

Multisource data fusion is a key issue in spatial information systems, concerned with how to combine data from different (and possible diverse) sensors in order to make inference about a physical event, activity or situation. Specifically, it is imperative that the processes’ result will maintain the highest accuracy and resolution of the original data.

The primary scientific objectives of the proposed investigation include:

(1) Developing a sequential Kriging algorithm, a mathematical and statistical model

of data integration, which uses accuracy information in the data fusion process. This accuracy information will be automatically estimated from the data using semivariogram modeling. Many data integration methods have been proposed including standard Least-Squares, Maximum Likelihood and Kalman filtering, to our best knowledge, a mathematical model of sequential Kriging has not yet been formulated and used.

(2) Investigation of an efficient data structure for sequential kriging

      interpolation of massive data sets. The Delaunay Triangulation (DT)  

      incremental algorithm, with  three levels neighborhood structure was

      successfully used in Kriging  [Felus et al., 2002]. TIN with incremental

      algorithms seems to be an ideal method for updating and data fusion

      tasks.

(3) Developing criteria for the optimal DT neighborhood size to be used in

        the sequential Kriging interpolation.

(4) Updating USGS/NOAA bathymetric data using the new TIN structured

sequential kriging and evaluating the performance of the proposed innovative model. Applications of the proposed mathematical solution and algorithm may readily be extended to other scenarios and spatial data sources (Shuttle Radar Topography Mission (SRTM), laser scanning, radar, imaging etc).

 

Technical publications so far:

  1. Felus A. Y., Saalfeld A. and Schaffrin B. (2005) Delaunay Triangulation Structured Kriging for Surface Interpolation. Journal of Surveying and Land Information Science, Vol. 65 (1) pp. 27-36.
  2. Felus, A.Y. and Schaffrin, B. (2005) Performing Similarity Transformations Using the Error-In-Variables Model. American Society for Photogrammetry and Remote Sensing (ASPRS) Annual Meeting, Baltimore, Maryland.
  3. Schaffrin, B. and Felus, A.Y. (2005), Generalizing the Total Least-Squares Estimator for Empirical Coordinate Transformations, SIAM Conference on Mathematical and Computational Issues in the Geosciences, Avignon, France.
  4. Felus, Y., and Schaffrin, B. (2005b). A Total-Least Squares approach for Semivariogram modeling of aeromagnetic data. GIS and Spatial Analysis, IAMG'05, Chang. Q. and Bonham-Carter G., eds., China University of Geosciences Printing House, Beijing, 215-221.
  5. Burtch R. R. Felus, A. Y. and Schaeding, C. M., (2006). Using Geostatistical Methods for Spatial Data Fusion. American Society for Photogrammetry and Remote Sensing (ASPRS) Annual Meeting, Reno Nevada.
  6. Burtch R. R. (2006), A comparison of methods used in rectangular to Geodetic Coordinates Transformations. American Congress for Surveying and Mapping Annual Conference. Orlando, Florida.
  7. Schaeding, C. M., Felus, A. Y. and Burtch R. R. (2006). DEM Decisions. Point of Beginning, Vol. 31 (11), pp. 28-33.
  8. Schaffrin, B. and Y. Felus (2006). A Multivariate Total Least – Squares Adjustment for Empirical Affine Transformations. In “Challenge and Role of Modern Geodesy". editor: Peiliang Xu. Proceeding of the VI Hotine-Marussi Symposium of Theoretical & Computational Geodesy:  29 May – 2 June 2006. Wuhan, China to be published by Springer-Verlag.
  9. Schaffrin, B. and  Felus, A.Y. (2006). Algorithms for Data Fitting in Error-in-Variables Models with Linear and Quadratic Constraints. International Workshop on Total Least Squares and Errors-in-Variables Modeling. Leuven, Belgium. Submitted to the journal of Computational Statistics and Data Analysis.

Other:

1.        Crimson & Gold newsmagazine article

2.        Baseline News

 

Intermediate reports and working documents:

1. First order polynomial trend surfaces and TIN interpolation

 

Pictures:

  1. Visit of the NGA research team at Ferris, November 2006.
  2. Visit of the NGA research team at Ferris, November 2005.
  3. From Left to Right: Dr. Yaron  Felus, Dr. Scott Loomer, Science Advisor for Geospatial Science, NGA, Dr. Paul Salamonowics, Sensor and Image Science, Team Lead, NGA, and Prof. Robert Burtch in NGA research Symposium, Washington DC, September 2004.
  4. The Katke Golf Course of Ferris State University
  5. Chad Schaeding mapping the golf course using a GPS
  6. Robert Burtch and Chad Schaeding mapping the golf course
  7. Yaron Felus and Chad Schaeding planning the golf course mapping

 

Research Data:

  1. Case Study 1: USGS bathymetric points at the Atlantic Ocean
  2. Case Study 2: Combining SRTM with Photogrammetric data
  3. Case Study 3: Combining Geoid undulation datasets from MDOT
  4. Case Study 4: Combining gravity datasets

References