By generalizing the clutter-clustered estimation method and considering the normalized sample covariance matrix (NSCM), a generalized NSCM(GNSCM) is proposed for covariance matrix structure estimation in correlated compound-Gaussian clutter. A maximum likelihood recursive estimation process of covariance matrix structure is derived in generalized clutter-clustered background. A generalized approximate maximum likelihood (GAML) estimator is then obtained by using GNSCM as the initialized estimation estimated matrix to recursive. GAML is an extension of the existing methods the approximate maximum likelihood (AML) and the constrained recursive clutter-clustered estimator (CRCCE). Simulation results show that, compared with
the two previous methods, GAML has higher estimation accuracy, and the corresponding adaptive detector has better constant false alarm ratio (CFAR) property and detection performance.