PhD Thesis Defense

Title: Time-frequency Methods for Vibration Estimation Using Synthetic Aperture Radar

By: Mr. Qi Wang

Advisor: Dr. Majeed Hayat

Date: Aug 20th 2012, 11:00 AM

Location: ECE, Room 118

Vibration signatures associated with various structures bear vital information about these structures; therefore, it is key to have the capability of estimating the vibration signatures. A lack of physical access to these structures typically makes the problem of detecting such activities challenging by current means. Synthetic aperture radar (SAR) has already been proven as a highly effective remote-imaging technique. Moreover, it is inherently capable of sensing Doppler shifts in the electromagnetic returns from objects, thereby allowing us to detect vibrations.

The target vibration can be estimated through successive chirp-rate estimations using the discrete fractional Fourier transform (DFrFT), which is inherently geared toward chirp-rate estimation. In this dissertation, two DFrFT-based methods, termed the range method and the cross-range method, are developed for SAR-based vibration estimation. The range method is capable of estimating vibrations with large and rapid changes in velocity. The cross-range method is capable of estimating realistic low-level vibrations (e.g., 1 mm, 5 Hz). The capability of the cross-range method is verified by experiments using the Lynx SAR system.

Clutter signal prevents the cross-range method to yield reliable results. A subspace method has been incorporated into the cross-range method to enhance its performance in clutter. Simulations and experiments have shown that the cross-range method now yields reliable results for signal-to-clutter ratios (SCRs) > 8 dB.

In strong clutter, i.e., SCR = 3 dB or lower, the clutter signal can be removed entirely by taking the difference of the two signals collects by a dual-beam SAR system. Three methods, termed the magnitude method, the extended Kalman filter (EKF) method, and the maximum-likelihood estimation (MLE) method, are proposed to estimate the vibration using the dual-beam SAR system. The magnitude method is simple and capable of estimating the frequency of single-component vibrations. The EKF method is capable of tracking the instantaneous velocities and displacements of all types of vibrations. By parameterizing the vibration displacement, the MLE method is capable of estimating the vibration parameters. The performance of the three methods is studied by means of simulation. The mean-square-error of the MLE method is compared to the Cramer-Rao lower bound.

Finally, in SAR images, target vibrations introduce localized artifacts that prevent SAR to form focused images of the vibrating object. By exploiting the vibration estimation results developed in this dissertation, a method based on phase demodulation has been developed to reduce the artifacts. The capability of this method is verified by experiments using the Lynx SAR system.

The vibration-estimation methods considered in this dissertation extract the vibrational information from the SAR and dual-beam SAR data without hardware modifications to the current radar systems. It therefore adds a new capability to existing SAR systems. Moreover, the research provides guidelines for future radar systems to make them more specialized in detecting vibrations and identifying objects of interest.