Fast vanishing point detection for unstructured road using Haar texture

• Published in: 2012 IEEE International Conference on Signal Processing, Communication and Computing (ICSPCC) pp. 167 - 170
• Main Authors: Yongzhong Wang, Xiaoyun Wang, Chenglin Wen
• Format: Conference Proceeding
• Language: English; Japanese
• Published: IEEE 01.08.2012
• Subjects: Accuracy; Computer vision; diversity voting; Estimation; Feature extraction; Gabor filters; Haar texture; Histograms; orthogonal rectification; Roads; vanishing point detection
• ISBN: 9781467321921; 1467321923
• DOI: 10.1109/ICSPCC.2012.6335718

The unstructured road detection is a challenging problem by the fact that many rural or desert roads are not well-paved, lack clear lane markers or boundaries. However, the ruts and tracks left by previous vehicles on the path commonly appear to converge into a single point, called as vanishing point, which is successfully used to determine the road region. In this paper, we propose a novel and fast texture extraction method by utilizing the carefully designed Haar templates and the orthogonal rectification technique, and based on which the vanishing point of unstructured road is reliably estimated by the diversity voting method. Compared with the state-of-the-art algorithms in a variety of difficult conditions ranging from desert to snow environments, the experiment results demonstrate the effectiveness and efficiency of the proposed methodology in quantitative and qualitative analyses.