Unstructured Road Segmentation based on Otsu-entropy Method

摘要

Unstructured road segmentation plays an important role in visual guiding navigation for intelligent vehicle. A novel vision-based road segmentation method that combined the Otsu double-threshold method with the maximum entropy double-threshold method was proposed to handle those problems caused by illumination variations and road surface dilapidation. Spatial correlation by analyzing the grey-level histogram of the original image and temporal correlation by matching of the selected referenced region was used to estimate the coarse range of the road region. Road segmentation experiments executed in different road scenes have demonstrate that the method proposed in this paper is robust against illumination variations and surface dilapidation.