Single Data Set Calibration and Imaging with Uncooperative Electromagnetic Inversion Systems

Abstract

After the production process, various types of grains are stored in storage facilities for years, and it is important to maintain humidity and temperature at values that ensure safe storage. Recently, Electromagnetic Imaging (EMI) systems for grain storage monitoring have been developed as an alternative method to monitor moisture content and temperature in grains while providing more sensitivity than other approaches. EMI systems use a large number of co-resident antennas usually connected to a Vector Network Analyzer via a switch. A numerical model is used to model the physical electromagnetic problem and an inversion algorithm is used to invert the collected data to produce an image of the target. However, before the computer model can be used, the raw VNA measurements must be calibrated to bridge the computational model and the true system physics. Traditional calibration approaches usually require two data sets: a data set measured from a known target for calibration purposes and a data set measured for the unknown target. We introduce and evaluate a new calibration method to calibrate and image using a single S-parameter measurement of the unknown target only. We apply this method to EMI inside of grain bins. This proposed calibration workflow: (1) estimates the bulk contents of the grain bin using a parametric inversion method and(2) uses the bulk results to subsequently estimate per-channel calibration coefficients for both the transmit and receive paths to each antenna. The novel calibration procedure is demonstrated via both synthetic and experimental results. The single-data set calibration can provide similar quality results as traditional two-data set calibration for a lab-scale experiment setup, but the technique could not produce acceptable images on the larger experiment. i ContributionsThe main contribution of this thesis is to show that, in the domain of Electromagnetic Imaging, both calibration and imaging with single data set is possible. In some sense, this could be called calibration of a measurement system without reference measurements. In grain bins, like many other electromagnetic imaging systems in other application areas, it is common practise to use two (or more) sets of measurements with a known and unknown target to calibrate the system. To the authors knowledge, calibration and imaging with only a single set of measurements has not been completed previously.This thesis is based on the work published in the following papers: [