One of the important indicators of the optical performance of windshields is optical distortion. Detecting optical distortion and making timely improvements is a key step in providing feedback on performance. A method based on feature extraction, the Lucas-Kanade (LK) optical flow distortion detection method, is proposed in this paper. The method involves simultaneously down sampling the reference grid and the distorted grid, extracting feature corner points, calculating the deformation field of corresponding points, mapping the calculation results back to the original image layer, and performing a secondary optical flow calculation to obtain an accurate deformation field. This process effectively replaces the multi-level iteration of the optical flow pyramid, reducing detection time by 20% while ensuring detection accuracy and improving the efficiency of the optical flow method.