Calculating the orientation of a rectangular target in SAR imagery

Four algorithms that estimate the orientation of a target before classification are presented. Two of the algorithms are based on a widely used box-spinning method, and the other two are based on linear regression techniques. The algorithms are evaluated using synthetic aperture radar (SAR) imagery of a large civilian vehicle. These data consist of two sets: one in which the vehicle is deployed in an open field, the other in which the vehicle is deployed along a treeline. The data were collected using the MIT Lincoln Laboratory millimeter-wave SAR, a high-resolution (0.3 m×0.3 m), fully polarimetric SAR with a center frequency of 33.56 GHz. For the vehicle deployed in an open field, three of the four algorithms estimate the orientation with a similar degree of accuracy. The linear regression algorithm that is based on a reweighted least squares (RLS) technique is shown to provide the best performance when the vehicle is deployed along the treeline