ROAD IMAGE SEGMENTATION BASED ON THRESHOLD WATERSHED ALGORITHM

摘要

Road segmentation has important applications in many areas, especially in the context of scene understanding and autonomous driving. However, in the existing segmentation methods, there are still over-segmentation and under-segmentation. To solve this problem, this paper proposes a watershed road segmentation algorithm based on threshold marking. The method first performs noise filtering and histogram equalization to preprocess the street view image, and then computes the maximum gradient image using the multi-scale morphological gradient algorithm. Then it optimises the gradient amplitude image through the obtained threshold marks. Finally, the watershed transform is used to extract the road information in the streetscape. The experimental results show that the algorithm performs well and the segmentation accuracy and efficiency are significantly better than the existing segmentation methods.