Abstract: The semantic segmentation of high-resolution remote sensing images has achieved notable precision through deep learning methods, albeit demanding a substantial number of parameters and computations. To address the challenge of reducing computational costs without sacrificing precision, this paper advocates a lightweight edge-enhanced network for semantic segmentation of high-resolution remote sensing images. Initially, the network employs the lightweight MobileNetv3 as its backbone feature extractor. The bottleneck block is enhanced by designing a residual dual-channel feature extraction module. Additionally, a contextual feature enhancer and depthwise separable convolution are introduced, effectively strengthening the feature fusion capability. To address the multi-scale challenge in remote sensing segmentation, a pooling fusion module is proposed, achieving complementary information exchange between deep and shallow features.To validate the effectiveness of the proposed method and demonstrate its superiority, we evaluate it against six segmentation models, including U-Net and DeepLabV3+, on public dataset ISPRS Potsdam and Vaihingen. Experimental results demonstrate that the proposed model achieves mean Intersection over Union (mIoU) scores of 78.16% and 76.78% on the two datasets, with an inference time of only 0.75 seconds. All evaluation metrics outperform the competing models, indicating a favorable balance between segmentation accuracy and computational efficiency.