Department of Electrical & Computer Engineering Signal and Image Laboratory (SaIL) The University of Arizona®

Past Research

Automatic Target Recognition Using Millimeter-Wave Radar

Images

Student: Mary L. Cassabaum

A challenging problem in digital signal and image processing is that of automatic target recognition (ATR) and object classification. In defense operations in particular, rapid object discrimination is critical. While this problem has been studied extensively, it remains a challenge due to the complex and time-intensive methods of typical approaches. This study involved constructing features from the range profile signal returns, identifying the "best" set of features and performing ATR.

MMW radar returns for two different target classes:

Maturation of millimeter wave (MMW) radar technology has led to the potential for exploitation of range profile signatures in the detection and identification of targets from the surrounding background. This potential is realizable due to the MMW wavelengths and the corresponding high range resolution obtainable at these frequencies. An additional benefit is the capability of MMW energy at key frequency bands to penetrate rain, dust, fog and smoke and provide an associated narrow radar beamwidth pattern. The narrow radar beamwidth results in improved cross-range resolution with the accompanying increase in the signal-to-clutter ratio (SCR). This increase in SCR is a significant benefit to detection/identification of targets due to the inherent suppression of terrain-induced clutter present in air-to-ground scenarios. The SCR may be further improved by utilizing doppler beam sharpening (DBS). DBS refers to a technique that improves the cross-range resolution through a "poor-man's SAR," and provides the capability to increase cross-range resolution beyond that nominally available with the traditional real beam (RB) techniques. Additionally, DBS permits the exploration of image-based techniques for feature extraction and pattern recognition. Whether using RB or DBS, the approach under investigation relies on the development and selection of salient features that are constructed from the range profiles derived from radar target signatures. Feature selection involves developing a suite of statistics (in the broad sense) that summarize and maximize the discriminatory information of the original signals. If the set of features is "sufficient," the set provides all the necessary information content of the original signal. It is the goal of feature extraction to select a set of statistics that provide "sufficiency" for discriminating between the target classes of interest; these are not necessarily the same statistics required for target signal characterization.

In this investigation, the targets of interest were characterized using fully polarimetric radar returns. A target radar cross-section (RCS) is commonly described by the scattering matrix. The scattering matrix is usually represented by a circular or linear orthogonal basis.

Collecting the full polarimetric signal allows the investigator to exploit all the information relating to target RCS. The advantages of extracting the polarimetric nature of a particular target RCS for detection and identification have been previously documented. This investigation derived several features that require full polarimetric returns. This study considered the extraction and selection of features from MMW radar returns, providing the information necessary for automated target class identification from RB and DBS signals of scenes.

Publications:

  1. Mary L. Cassabaum, Donald E. Waagen, Jeffrey J. Rodriguez, and Harry A. Schmitt, "Unsupervised Optimization of Support Vector Machine Parameters," in Automatic Target Recognition XIV, Ivan Kadar, ed., Proc. of SPIE, vol. 5426(1), SPIE Defense \& Security Symposium, Orlando, FL, April 13-15, 2004, pp. 316-325. [ PDF ]

  2. Mary L. Cassabaum, Jeffrey J. Rodriguez, Jack G. Riddle, and Donald E. Waagen, "Feature Analysis Using Millimeter-Wave Real Beam and Doppler Beam Sharpening Techniques," in Proc. 2002 IEEE Southwest Symp. on Image Analysis and Interpretation, Santa Fe, NM, April 7-9, 2002, pp. 101-105.[ PDF ]

 1230 E. Speedway Blvd., P.O. Box 210104, Tucson, AZ 85721-0104
 ©2014 All Rights Reserved.  
 Contact webmaster                                  
Think ECE!