Real-Time Speed Bump Detection Using Image Segmentation for Autonomous Vehicles

Autonomous vehicle technology, which is evolving at a faster pace than predicted is promising to deliver higher safety benefits. Detecting the obstacles accurately and reliably is important for safer navigation. Speed bumps are the obstacles installed on the roads in order to force the vehicle driver to reduce the speed of the vehicle in the critical road areas, such as hospitals and schools. Autonomous vehicles have to detect and slower the speed appropriately to drive safely over the speed bump. In this paper, we propose a novel method to detect the upcoming speed bump by using a deep learning algorithm called SegNet, which is a deep convolutional neural network architecture for semantic pixel-wise segmentation. The trained model will give segmented output from the monocular camera feed placed in front of the vehicle.