Domain-adaptive lane detection for transitioning weather conditions

Abstract

Lane detection is a critical task handled by autonomous vehicles to perceive their environment accurately. Ensuring robust and reliable lane detection capabilities is essential for the safety of the vehicle's operations. However, adverse weather conditions like rain can significantly impede the performance of lane detectors. They introduce rain artifacts in images, which can distort the lane features, making them more difficult to distinguish. Rain also causes road surfaces to be more reflective to light, distorting much of the markings on the road, further causing a greater hindrance to lane detection. When it rains, the road surface also becomes more slippery, making correct and accurate lane detection even more crucial in autonomous vehicles for their safety. Therefore, it is crucial to adapt lane detectors to perform reliably in both clear and rainy conditions.