Adaptive subspace detection in signal-dependent interference and non-Gaussian clutter

This paper investigates the adaptive subspace detection problem in signal-dependent interference and non- Gaussian clutter. The clutter is modeled as a spherical invariant random vector (SIRV), which is the product of a generalized inverse Gaussian distribution texture and complex Gaussian distribution speckle components. To solve the adaptive detector design in signal-dependent interference and generalized inverse Gaussian texture clutter, a singular value decomposition (SVD) approach is presented to modify the detection problem. Based on the two-step generalized likelihood ratio test (GLRT) and maximum aposteriori (MAP) GLRT, two novel detectors are designed, where the test statistics of the detectors are derived by assuming the speckle covariance matrix is known in the first step. Then, the estimation of the covariance matrix is inserted into the test statistics to obtain the adaptive detectors. Besides, the proof of the constant false alarm ratio (CFAR) properties of the proposed subspace detectors is developed. The numerical simulation experiments in both simulated clutter and measured clutter data demonstrate the effectiveness of the proposed detectors.