An Optimalizing Threshold Segmentation Algorithm for Road Images Based on Mathematical Morphology

Considering the requirements of real time and robustness for the segmentation algorithms for mobile robots, an optimal threshold segmentation algorithm based on mathematical morphology was studied. Two improved measures, i.e., introducing an improved combination filter of median and multi-scale, and increasing morphological opening and closing operation, were proposed for the general optimal threshold segmentation algorithm. The experimental results indicated that the algorithm can not only accurately segment road and off-road regions within the unstructured road images that contain three common types of noise, but better overcome the influence of the road shadow, water marks, cracks and damage on the segmentation.