FPB-UNet++: Semantic Segmentation for Remote Sensing Images of reservoir area via Improved UNet++ with FPN

Abstract

In order to improve the accuracy of semantic segmentation of remote sensing images in the reservoir area, this paper improves UNet ++, and proposes a UNet ++ semantic segmentation network model fused with feature pyramid network, called FPB-UNet ++. First, in order to fully extract the semantic information of different scales and enhance the recovery ability of the spatial information of remote sensing images, this paper uses the improved feature pyramid structure as the basic unit of the UNet ++ coding structure. Then, the pooling of position information will be lost between each coding unit To remove the layer, use convolution instead. Finally, in order to make full use of multi-scale feature information in the multi-sided output part, all the side output feature maps are stitched and fused in the channel dimension. Through experiments on the open and self-built remote sensing image semantic segmentation data set of Xiaolangdi Reservoir area, the results show that the network model has a good segmentation effect on feature information.