Convolutional neural network applied to traversability analysis of vehicles

Abstract

We focus on the need for traversability analysis of vehicles with convolutional neural networks. Most related approaches to traversability analysis of vehicles suffer from the limitations imposed by extracting explicit features, algorithm scalability, and environment adaptivity. In views of this, an approach based on the convolutional neural network (CNN) is presented to traversability analysis of vehicles, which can extract implicit features. Besides, in order to enhance the training speed and accuracy, preprocessing and normalization are adopted before training. The experimental results demonstrate that our method achieves high accuracy and strong robustness. © 2013 Li Linhui et al.