Classification of landscape architecture design based on dual-channel attention improved FCN

Landscape architecture design integrates natural and artificial landscapes, and needs to accurately categorize diverse landscape elements. To solve the problems of low efficiency and high subjectivity of traditional design, the study proposes an improved fully convolutional network model that combines the U-Net structure, multi-scale hopping connection network, and dual-channel attention mechanism to enhance the ability of detail capture and feature fusion. The experimental results show that the model achieves the highest classification recall of 0.92, the shortest inference time of 0.17 s, the precision of 92.96 %, 93.97 % and 92.94 % on vegetation, sky and building categories, respectively, and the feature extraction is stable with good robustness in pixel value interval. The results validate the efficiency and adaptability of the model in complex landscape scenes and provide effective support for landscape design intelligence.