High-resolution remote sensing image road extraction method based on RBAM attention mechanism and RBELoss

The invention discloses a high-resolution remote sensing image road extraction method based on an RBAM attention mechanism and RBE Loss, and the method comprises the following steps: S1, building an improved RBAU-Net network structure through adding the RBAM attention mechanism at a jump connection structure on the basis of a U-Net network; s2, selecting a pre-extracted high-resolution remote sensing road image as an intermediate feature map, and obtaining a reconstructed enhanced feature map through a channel attention mechanism and space attention; s3, constructing a self-adaptive road type judgment module by analyzing the morphological characteristics of the weak road; and S4, using a loss function RBE Loss of road information compensation to complete the learning of the road type determination module on the weak road target. According to the method, the problem of information redundancy during feature extraction can be well solved, and the weak road extraction capability of the model is improved. 本发明公开了一种