From Raindrops To Pixels: A Novel Model to Predict ADAS Camera Image Degradation in Rain

The development of Advanced Driver Assistance Systems (ADAS) has led to safer roads through the implementation of automated features such as blind spot monitoring, lane departure assist, and collision mitigation braking. In ADAS, cameras are one of the sensor types required to capture environmental information. However, camera performance degrades in adverse weather, especially in rain, due to droplet impacts, which cause changes to the light trajectory, thus decreasing object recognition and detection quality. While it is commonly understood that camera image quality directly correlates to rain intensity, understanding is limited in the exact relationship between incoming rain and image degradation in a dynamic environment due to the complexity of the multiphase interactions. In this study, a novel wind-driven rain system is deployed in conjunction with a spray-nozzle system at the ACE Climatic Aerodynamic Wind Tunnel. Using the controlled and repeatable testing methodology, objective camera image quality metrics are correlated with the incoming rain conditions through theoretical modeling. The model shows consistent trends across a variety of test conditions with different surface materials. The results prove the validity of the proposed model, which provides the impetus for further development to include additional physical interactions, such as droplet dynamics and vehicle aerodynamics.