Automotive Integration of Smart Road Condition Detection System

We introduce the integration of an innovative approach for optically assessing hazardous road conditions, such as detecting ice and wet surfaces. Our cutting-edge method effortlessly pairs with commercially available CMOS image sensors, expanding the use of standard optical filters and significantly boosting the safety features of selected Advanced Driver Assistance Systems (ADAS). The detection system has been designed with cost-effectiveness and high accuracy in mind, leveraging the physical optical properties of water in conjunction with information obtained from RGB camera recordings, rather than relying solely on the latter. The detection system has been designed to function without the need for additional stimulation, utilizing only naturally available light sources or artificial illumination such as headlights or streetlights. Rather than relying on absolute measurements, it evaluates road conditions through the utilization of spectral absorption contrast ratio and polarization contrast ratio. A simple CMOS sensor, devoid of the Bayer filter, captures the light scattered off the road surface in two spectral bands (VIS and NIR) and two polarization planes. A machine learning-based classification algorithm, utilizing a trained Support Vector Machine (SVM) model, is employed for data inference to classify the road condition. The integrated system has undergone initial field tests, demonstrating good selectivity and robustness in the presence of movement and changing ambient light conditions. This paper provides an overview of the system and details its integration into a conventional passenger vehicle, located behind the windshield window. The initial results and encountered challenges will be presented and discussed.