Taking account of the difficulty of shape estimation for the extende d targets,a new measurement model and a shape estimation approach based on Gaussian surface feature matrix are proposed in this paper.First,the structural points are establiched,wh ich are able to reflect the true shape of the extend target,and these points are used to construct some Gaus sian s urfaces.Then these Gaussian surfaces are fitted to yield the suitable measuremen t spatial distribution models. Furthermore,the feature of measurement spatial distribution is described by usin g a matrix with Gaussian surface fitting approach,and a mapping relationship between the matrix coordinates and the Cartesian coordinates is established by a suitable mapping f unction.Finally,the Gaussian surface feature matrix is updated by Bayesian filtering method.Compared with the conventional algorithms,the proposed algorithm can estimate the true shape of extended targe t in the high-noisy environment,even the preset shape is inaccurate.Moreover,it can b e used to estimate any irregular shape,even the hollow shapes,without knowing the preset shape i nformation. Simulation results show that the plane shapes,hollow shapes and group targets a re estimated accurately in case of the preset shape parameters are inaccurate,which demonstr ates that the proposed algorithm has good performance for shape estimation of any extended t arget with a strong robustness.