Our research on remote sensing data
analysis is conducted in the context of the USDA’s Conservation
Reserve Program (CRP) that seeks to encourage farmland owners to adopt
sustainable management practices. There are two CRP-related research
issues that involves remotely sensed Landsat imagery, i.e., CRP mapping and compliance monitoring. The first one
focuses on the delineation of CRP tracts from the remote sensing data,
and the second one is concerned with the compliance of given CRP tracts.
We have applied both the decision tree and support vector machine (SVM)
approaches to the first problem, and we have shown the benefit to
involve the multi-source GIS data for CRP mapping. We also develop
a new one-class SVM method for CRP compliance monitoring that
modifies the original one-class SVM approach by introducing a
self-supervised training scheme.
Detected CRP tracts in the Texas County, Oklahoma imposed
on the Landsat TM imagery (left) and the reference CRP map (right).
X. Song, G. Fan and M. Rao, “Edited-Bootstrapped
Support Vector Machines for One-class Remote Sensing Data Analysis”,
IEEE Geoscience and Remote Sensing Letters, April 2008.
X. Song, G. Fan, and M. Rao, “Automated CRP Mapping
using Non-parametric Machine Learning Approaches”, IEEE Trans. on
Geoscience and Remote Sensing, Vol. 43, No. 4, pp888-897, April, 2005.
X. Song, G. Cherian, and G. Fan, “A $\nu$-insensitive
SVM Approach for Compliance Monitoring of the Conservation Reserve
Program”, IEEE Geoscience and Remote Sensing Letters, Vol. 2, No. 2,
pp99-103, April, 2005.
X. Song, G. Fan, and M. Rao,
"Machine Learning Approaches for Multisource Geospatial Data
Classification with Application to CRP Mapping in Texas County,
Oklahoma", in Proc. IEEE Workshop on Advances in Techniques for
Analysis of Remotely Sensed Data, NASA Goddard Visitor Center,
Washington DC, October 27/28, 2003.