“In the field of deep learning semantic segmentation, this paper proposes a street view image semantic segmentation network MDFNet that integrates multidimensional features. Through the target area enhancement module, feature pyramid grid, and dual solution dock technology, the segmentation accuracy of complex street view images is significantly improved, with an average intersection to union ratio of 80.11%, providing a new solution for street view image understanding.”