Abstract: An improved method of fundamental solutions is presented, and it is applied to the research of the inverse identification of Cauchy boundary conditions for 2D anisotropic potential thin body problems. In the process, the truncated singular value decomposition is used to solve the resulting matrix equation which is highly ill-conditioned, and the method of generalized cross validation criterion is used to select the truncated number associated with useful singular values. Owing to the use of regularization method, the selection range of the distance between the source point and the real boundary was greatly expanded, and the sensibility of accuracy of the numerical solutions resulting from the choice of the distance is also reduced. Numerical examples are tested to demonstrate the feasibility and accuracy of the proposed technique, even if the thickness of a thin body structure down to 1E-09, the method still can obtain a very highly accurate numerical solution. It provides a new approach for dealing with such problems. Meanwhile, it extends the application field of the method for fundamental solutions (MFS).