Ensuring the quality of automotive components requires precise sheet metal forming simulation using accurate parameters of material models including yield function and strain hardening laws. Conventionally, the parameter identification has been performed based on multiple simple material tests. However, the deformation states obtained from the conventonal material tests are limited, and many tests are required to achieve high identification accuracy. In this study, we develop a Bayesian data assimilation method that uses a specially-designed “D-shaped”specimen to generate multiaxial deformation states [1] for accurately estimating material model parameters. Furthermore, this method focuses on the selection of the region of interest (ROI) for the strain measurement using the digital image correlation (DIC), which significantly affects the accuracy of parameter estimation. The data assimilation enables the estimation of material model parameters by accounting for uncertainties associated with both experimental and simulation results. In this work, we conduct uniaxial tensile test simulations on A6061 aluminum alloy and estimate the anisotropy parameters of the Yld2000-2d yield function [2]. The D-shaped specimen features a parallel section corresponding to a uniaxial stress state, in addition to regions generating multiaxial stress states. We investigate the influence of ROI selection on estimation accuracy by comparing results obtained with and without the data from this parallel section. The results show that excluding the parallel section from the observation data reproduces the experimentally obtained strain distribution more accurately than when it is included. Moreover, these results confirm that the estimated parameters accurately reproduce the geometric shape of the yield surface representing the material's plastic anisotropy.