Urban Land Use Classification with Street View Image-Assisted Remote Sensing Images

Accurate information on land use classification is essential for promoting sustainable urban growth, preserving the environment, safeguarding public health, and enhancing socio-economic prosperity. This paper explores the application of street view image (SVI) assisted remote sensing images (RSI) for urban land use classification. In this paper, we introduced a dual-stream network integrating a large model and CNN for feature extraction, termed the Large-VGG Dual-Stream Network (LVDNet), which extract RSI semantic features using a CNN backbone, while SVI semantic sementic features are obtained from a large model trained on real-world data, with feature fusion achieved through cross-learning. We constructed a new dataset $\mathbf{Z Z}$ for experiments, in which the spatial correspond between street view images and remote sensing imagery is established. Our findings indicate that the proposed method performs effectively in datasets, demonstrates SVI significantly enhance the accuracy of urban land use classification, particularly in identifying complex urban functional areas.