Modeling Spatial Data: Challenges and Applications to Solar Imagery

Mike Turmon (JPL/Caltech) (S)

Statistically-motivated data-mining schemes are particularly suited for science data analysis. We discuss several schemes for locating features of interest in, or generative modeling of, scientific imagery. Some examples include location of objects like craters and volcanoes by template matching, identification of sunspots and faculae in solar imagery using Markov random fields driven by learned Gaussian mixtures, and tracking the evolving path and characteristics of moving objects with Kalman filters. The explicit use of statistical models allows their encapsulation in a definition language for re-use and exchange.

Presentation (PDF File)

Back to Mathematical Challenges in Scientific Data Mining