Bump/Hump Detection Using CNN with Stereo Camera Road Height Profile as Input

Abstract

We developed a method to detect bumps/Humps, which are small bumps placed on the road, using an in-vehicle stereo camera. Therefore, we developed a CNN discriminator that uses depth-height waveforms obtained from 3D shape measurements as input. We investigated the performance and processing time of a configuration that can be scaled to the processing capacity of an embedded device.