Electromagnetic Imaging (EMI) is a modality that uses electromagnetic waves to interrogate an object and reconstruct the internal electrical properties. Data is collected in an imaging chamber in which antennas transmit and measure the fields surrounding the object. Cables and a switching network are necessary to connect the antennas to a vector network analyzer (VNA) to collect data, however, they also introduce mismatch by attenuating and phase shifting the signal between the antennas and VNA. A second source of error is the simplifying assumptions EMI algorithms make to efficiently model the electromagnetic fields, leading to discrepancies between the physical system and simulated fields. This thesis focuses on the calibration process used to convert the raw measurements to the scattered field data used as the input of the EMI algorithm. The first contribution of this thesis addresses the signal mismatch between the VNA and antenna ports. Considering the application of grain bin EMI, in which EMI is performed in commercial grain bins to prevent grain spoilage, the long cables needed to surround the bin structure can lead to highly corrupted signals. Herein, we consider the case when the EMI system cannot be physically accessed, and only partial scattering parameters are available from the VNA. A de-embedding method is presented that estimates the unknown cable and VNA scattering parameters so that the measurements at the antenna ports can be recovered. The method is experimentally tested on an industrial sized grain bin, with the recovered data agreeing within the measurement error of the VNA. The second contribution of this thesis is a novel calibration procedure that models the antennas and scattered fields in the imaging chamber as 2-port networks. The parameters of the antenna model are calculated by measuring calibration targets for which the theoretical fields are calculated. After the antenna model is characterized, a simple de-embedding procedure can be applied to recover the scattered fields of unknown targets. The method is experimentally tested in a 2-dimensional near-field imaging system by comparing the calibration accuracy against a technique popular in the literature, showing modestly higher accuracy.