Challenges in Object Detection Under Rainy Weather Conditions

Abstract

Intelligent vehicles use surround sensors which perceive their environment and therefore enable automatic vehicle control. As already small errors in sensor data measurement and interpretation could lead to severe accidents, future object detection algorithms must function safely and reliably. However, adverse weather conditions, illustrated here using the example of rain, attenuate the sensor signals and thus limit sensor performance. The indoor rain simulation facility at CARISSMA enables reproducible measurements of predefined scenarios under varying conditions of rain. This simulator is used to systematically investigate the effects of rain on camera, lidar, and radar sensor data. This paper aims at (1) comparing the performance of simple object detection algorithms under clear weather conditions, (2) visualizing/discussing the direct negative effects of the same algorithms under adverse weather conditions, and (3) summarizing the identified challenges and pointing out future work.