Synthetic Aperture Radar 3D Feature Extraction for Arbitrary Flight Paths

Rethinking vehicle classification with wide-angle polarimetric SAR

“…It has been shown that when the class of ASC models is used joint ly with frequency, polarization, and wide-angle apertures, the primitive class ambiguity is significantly reduced [7], [8], [15], [16]. Then, ASC primitives become highly desirable tar get features.…”

“…For large 3600 im ages (over 512 x 512 pixels), this will likely take several hours. However, the extracted features are immediately available for im age-based parameter estimation methods such as in [16]. The com putational burden of conduct ing nonlinear optimization is not clear because of the sophistication of the algorithms and the size of the available search library, but it (Left) Joint frequency and polarization feature vector [8].…”

“…Note the feature set may be expanded to include additional information about a primi tive [16]. In general, the ASC model estimation problem is to identify the number of primitives N that represent a target as well as the type r m and underlying parameters e m (e.g., 3D location, size, and pose) of each primitive.…”

“…For example even-bounce objects may be dihedral or top-hat reflectors. However, the joint frequency-polarization feature space re moves the ambiguity by intersecting the physics captured by the a parameter and the physics captured by the polarizaare coupled, and optimization is a high-dimensional, nonlinear, nonconvex problem [16]. Accu rate initialization of primitive classes can greatly reduce the search time for a solution.…”

Sparsity-Driven Synthetic Aperture Radar Imaging: Reconstruction, autofocusing, moving targets, and compressed sensing

“…Sparsity-driven 3D image formation has also been used to initialize the process of geometric feature extraction from SAR data collected over arbitrary, monostatic or bistatic SAR apertures [21].…”