Semantic segmentation network-based road adhesion coefficient prediction method

The invention discloses a road adhesion coefficient prediction method based on a semantic segmentation network, and the method comprises the steps: building a semantic segmentation network based on a multi-scale space attention mechanism; secondly, pre-training the established segmentation network on a public data set; then, enriching a semantic segmentation data set and carrying out specific training on a segmentation network; then, extracting a road surface region by using a segmentation network to manufacture a road surface classification network data set; secondly, building and training a road surface type classification network; and finally, establishing a mapping rule to obtain road adhesion coefficient information. According to the method, the generalization ability of the algorithm on rain and snow driving scenes is enhanced, and the precision, the real-time performance and the robustness of driving road surface extraction are further improved; and meanwhile, a lightweight road surface recognition net