Semantic segmentation of UAV remote sensing images based on improved U-Net

Semantic segmentation of unmanned aerial vehicle (UAV) remote sensing images has been widely used in various fields of remote sensing; however, manual visual interpretation methods are inefficient and highly dependent on ground investigation and a priori knowledge, which consume huge human and material resources. Therefore, we propose a deep convolutional neural network based on improved U-Net to solve the semantic segmentation problem of UAV remote sensing images. The network is a U-shaped structure with a symmetric encoder and decoder. We conducted experiments on the UAV remote sensing dataset Aeroscapes, and compared it with several other popular semantic segmentation networks, our method achieved significant improvement in segmentation accuracy.