Machelle Wilson

Savannah River Ecology Laboratory

Support Vector Classification of Hyper-Spectral Data on Plants Exposed to Metal Toxicity

Remote sensing technologies with high spatial and spectral resolution show a great deal of promise in addressing critical environmental monitoring issues, but the ability to analyze and interpret the data lags behind the technology. Robust analytical methods are required before the wealth of data available through remote sensing can be applied to a wide range of environmental problems for which remote detection is the best method. In this talk, I introduce the basic theory behind support vector machines (SVM) and discuss their effectiveness in classifying plants that have been exposed (or not) to varying levels of heavy metal toxicity using leaf-level hyperspectral (reflectance) data. If SVM work well on leaf-level data, then there is some hope that they will also work well on data from airborne and space-borne platforms, which are significantly more complicated. We discuss the probabilistically derived bounds on the prediction error, as well as leave-one-out cross validation to assess the effectiveness of SVMs on two data sets from a controlled greenhouse study.


Back to Colloquium Series