Effects and Benefits of Discrete Cosine Transform on Polarimetric Decompositions: Application to Man-made Object Recognition

In this project, the intern will use a recently developed interpolation technique based on the Lie group theory to enhance the quality of the classification of Synthetic Apertur Radar images. He will evaluate the effects and discuss the benefits of this interpolation on the complete set of polarimetric features extracted from a fully polarimetric SAR image. The effects on the polarimetric features will impact the performance of target recognition algorithms applied to the detection of man-made ground targets. The recognition process will be achieved by using a neural network. These techniques will be applied on the MSTAR data set available online from the DARPA agency and made from a collection of a large variety of land-based targets under different poses relative to the SAR sensor.

Faculty Supervisor:

Dr. Jiri Patera

Student:

Matthieu Voorons

Partner:

Lockheed Martin Canada

Discipline:

Mathematics

Sector:

Aerospace and defense

University:

Université de Montréal