基于DLformer-UNet网络的UAV航拍图像分割方法OA

Unmanned Aerial Vehicles(UAVs)are increasingly pivotal in military operations,with aerial imagery serving as a critical source of intelligence.However,UAV image datasets pose challenges due to their rich contextual information and significant scale variation among objects.To address the inherent limitations of Convolutional Neural Networks(CNNs)and Transformers in semantic segmentation,we propose an enhanced architecture named DLformer-UNet.First,a lightweight self-attention module is introduced to reduce computational overhead.Second,the feedforward network is replaced by a dual-path hybrid gated network to boost segmentation performance.Extensive experiments conducted across multiple datasets,employing comparative analyses with different network backbones,demonstrate that our model achieves superior segmentation results.