The University of Michigan - Ann Arbor

Electrical Engineering and Computer Science Department

Radar Image Processing Lab


Classification

Land-cover classification is one of the main uses for remotely-sensed imagery of nay kind and SAR is no exception. The ecological and climate-modeling communities use this data to produce models with higher fidelity than before. Others also use the data over smaller areas such as the U.S. Forest Service which needs estimates of forest growth and health. Potentially this could even be used for identifying health and type of crops. In every case, the image must first be classified to the species level before other details can be determined.

Some examples are shown below. The image on the left shows a classification using a composite of the ERS and JERS Satellite SAR data. The image on the right shows a portion of the same area but classified using multi-frequency and multi-polarization SIR-C SAR data. The 2 methods agree well with each other and with reality.

Classification of Raco Supersite using combination of ERS and JERS sensorsMulti-frequency and multi-polarization classification using SIR-C data.



Some of our publications in this area:


The web address for this document is:   http://www.eecs.umich.edu/RADLAB/sar_image_lab/classification.html
last update: 3-19-98

Any questions or comments should be directed to:    Leland Pierce <lep@eecs.umich.edu>

Radar Image Processing Lab
The University of Michigan - EECS Dept.
1301 Beal Ave
Ann Arbor, MI 48109-2122