In new energy vehicles, aluminum alloy has gained prominence for its ability to achieve superior lightweight properties. During the automotive design phase, accurately predicting and simulating structural performance can effectively reduce costs and enhance efficiency. Nevertheless, the acquisition of accurate material parameters for precise predictive simulations presents a substantial challenge. The Johnson-Cook model is widely utilized in the automotive industry for impact and molding applications due to its simplicity and effectiveness. However, variations in material composition, processing techniques, and manufacturing methods of aluminum alloy can lead to differences in material properties. Additionally, components are constantly subjected to complex stress states during actual service. Conventional parameter calibration methods primarily rely on quasi-static and dynamic tensile tests, offering limited scope in addressing compression scenarios. This paper proposes an inversion calibration method tailored for 6082-T6 aluminum alloy, employing both tensile and compression tests to refine parameter estimation. The calibration process integrates results from both tensile and compressive stress states into the Non-dominated Sorting Genetic Algorithm-II (NSGA-II) data set. The initial model parameters are derived from uniaxial tensile test results. Subsequently, uniaxial compression test results are used as target curves, with the optimal calibration parameters identified through the minimization of the standard deviation between these target curves and corresponding simulated curves. Finally, the accuracy of the method is verified by tensile tests, compression tests and drop hammer impact tests. Throughout the process of parameter calibration, a more suitable material model can be obtained by adjusting the content of target curves and the specific gravity between target curves.