This dissertation covers the development of a geometry-based sensor model for a specific monostatic spotlight synthetic aperture radar (SAR) system—referred to as the ExSAR (for experimental SAR). This sensor model facilitates single- and multiple-image geopositioning with error analysis. It allows for the use of known ground control points in refining the collection geometry parameters (a process called image resection) and for the subsequent geopositioning of other points using the resected image. Theoretically, the model also allows for the potential recovery of bias-like, persistent errors common across multiple images. The model also includes multi-image correspondence equations to aid in the cross-image identification of conjugate points.

The sensor model development begins with a generic, theoretical approach to the modeling of spotlight SAR. A closed-form solution to the range and range-rate condition equations and the corresponding error propagation equation are presented. (The SAR condition equations have traditionally been solved iteratively.) The application of the closed-form solution in the image-to-ground and ground-to-image transformations is documented. The theoretical work also includes a preliminary error sensitivity analysis and a treatment of the spotlight SAR resection process.

The ExSAR-specific model is established and assessed with an extensive set of images collected using the experimental radar over arrays of ground control points. Using this set, the imagery metadata elements are assessed, and the optimal element set for geopositioning is determined. The ExSAR imagery is shown to be transformed to the ground plane in only one dimension.

The eventual ExSAR sensor model is used with known elevations and single-image geopositioning to show a horizontal accuracy of 8.23 m (rms). With resection using five ground-surveyed control points per image, the horizontal accuracy of reserved check points is 0.45 m (rms). Resections using the same points, but recovering fewer geometry parameters produced nearly equal accuracies.

Three-dimensional geopositioning was shown feasible using the model and multiple images. Correspondence curve equations were developed and tested. These define the expected location of a conjugate point in overlapping imagery.

This work includes extensive explanatory appendices, including preliminary work with bistatic SAR modeling. The experimental data set and MATLAB ^® code are available.