U-Net and Transformer-Based Remote Sensing Image Urban Road Recognition Method

In recent years, with the rapid development of remote sensing technology, urban remote sensing images have been widely used in urban planning, traffic management, disaster monitoring and other fields. However, the directions, colors, and shapes of urban roads are diverse. If a dataset that exhaustively enumerates all samples cannot be prepared, the recognition ability of the corresponding deep model will be very low. In order to solve the above problems, this paper proposes a U-Net and transformer-based remote sensing image urban road recognition method (UT-Road), a typical characteristic of this method is that it can recognize roads in cities based on “linear” recognition in the sequence through the transformer structure on the basis of the traditional semantic segmentation structure. In the experiment, we introduced multiple sets of remote sensing images for testing. The results showed that our method is significantly better than the traditional method in terms of accuracy and stability. It shows that UT-Road has huge application value in urban road management.