We address the estimation of the scatter matrix of a scale mixture of Gaussian stationary autoregressive vectors. This is equivalent to consider the estimation of a structured scatter matrix of a Spherically Invariant Random Vector (SIRV) whose structure comes from an autoregressive modelization. The Toeplitz structure representative of stationary models is a particular case for the class of structures we consider. For Gaussian autoregressive processes, Burg method is often used in case of stationarity for its efficiency when few samples are available. Unfortunately, if we directly apply these methods to estimate the common scatter matrix of N vectors coming from a non-Gaussian distribution, their efficiency will strongly decrease. We propose then to adapt these methods to scale mixtures of autoregressive vectors by changing the energy functional minimized in the Burg algorithm. Moreover, we study several approaches of robust modification of the introduced Burg algorithms, based on Fréchet medians defined for the Euclidean or the Poincaré metric, in presence of outliers or contaminating distributions. The considered structured modelization is motivated by radar applications, the performances of our methods will then be compared to the very popular Fixed Point estimator and OS-CFAR detector through radar simulated scenarios.