Target classification in SAR images

Srinivas Arunteja Gottipati, Tennessee State University

Abstract

Synthetic Aperture Radar (SAR) is an active, all-weather remote sensor that can operate during day or night and penetrates cloud cover. It can construct high resolution images of the ground from airborne or space borne platforms. This project develops automatic target classification system that can classify targets in SAR images. Such a system is especially useful for military applications. The research goal was to design, implement, and test an automatic target classification system that can classify some targets of interest into several categories, such as BTR60, T72, D7, 2S1, BRDM2, ZIL131 and ZSU234 in SAR images. A number of feature extraction techniques including Principle Component Analysis and the Hu-invariant moments are employed. Support Vector Machines (SVM) is used as the main classifier. Neural Network is also employed for performance comparison. The MSTAR database is utilized for development and system testing.

Subject Area

Computer Engineering

Recommended Citation

Srinivas Arunteja Gottipati, "Target classification in SAR images" (2009). ETD Collection for Tennessee State University. Paper AAI1473385.
https://digitalscholarship.tnstate.edu/dissertations/AAI1473385

Share

COinS