Real-Time Stair Detection Using Multi-stage Ground Estimation Based on KMeans and RANSAC

Multiplane estimation from three-dimensional (3D) point clouds is a necessary step in the negative obstacle detection. In recent years, different Random Sample Consensus (RANSAC) based methods have been proposed for this purpose. In this paper, we propose a multi-stage algorithm based on RANSAC plane estimation and KMeans clustering, and apply it to the negative stairs detection. This method contains two steps: first, it clusters the point clouds and downsamples them; second, it estimates the planes by iteratively using RANSAC algorithm with the downsampled data. Finally, according to the relationship between regions to determine whether there is an obstacle in front of the autonomous vehicle. Our experimental results show that this method has satisfactory performance.