Imaging through curved glass: Windshield optical impact on automotive cameras

The state-of-the-art position of cameras in forward-facing Advanced Driver Assistance Systems (ADAS) is behind the windshield, integrated within the rearview mirror holder. In this position, the quality of the windshield as an optical element directly impacts the quality of the camera image. With increasing camera resolution and narrow field of view optics required for large object detection distances, the optical impact of the windshield becomes increasingly important. We suggest a method based on computer graphics for evaluating the optical performance of windshields in front of ADAS cameras. Using a ray tracing framework, we produce quantitative simulations of the light transport through the windshield. To represent the geometry of the windshield, we fit ellipsoid models to measurements of its inner and outer surfaces produced using a chromatic white light sensor in a coordinate measuring machine. The ellipsoid fits enable accurate ray intersections with the windshield even for cameras positioned close to the windshield surface. Additionally, we investigate the windshield microgeometry using optical profilometry and find that the microstructure is smaller than 200 nm. Thus, the microgeometry can only cause a very slight diffractive blur, and we consider the ellipsoidal macrogeometry sufficient for evaluating the light transport. In simulation experiments, we evaluate the impact of the windshield on a forward-facing ADAS camera by computing the modulation transfer function degradation of the camera image. In our experiments, we vary camera aperture and resolution as well as distance and angle of the windshield to the camera. To validate our results, we reconstruct the angle variation experiment in the lab.