In this paper, we propose a new method that can extract aberrations using aerial image measurements and present its experimental results on lithographic tools. Based on physical simulation and statistical analysis, a linear regression matrix is obtained establishing a connection between principal component coefficients of specific aerial images and Zernike coefficients. In the application phase, the aberrations of the projection lens are solved via the use of this regression matrix. An engineering model is established based on an extension of theoretical model that incorporates all the significant systematic errors. The performance of the engineering model as applied on a 0.75 NA ArF scanner is reported. In the experiment, measurement marks oriented in orthogonal directions are used and aerial images at 9 field points are measured. To verify the repeatability of this technique, every point is measured 20 times. By inputting the aerial images into the engineering model, Zernike coefficients are solved and the results are analyzed. The wafer exposures were performed to evaluate the results of aberration correction.