HRIC: A High-Resolution Remote Sensing Interchange Classification Dataset

High-accuracy classification of interchange structures from remote sensing imagery is crucial for urban transportation planning, intelligent transportation systems, and autonomous driving technologies. However, current research on classifying high-resolution remote sensing images of interchanges remains limited, particularly due to the lack of dedicated high-resolution datasets. To address this issue, this study constructs a high-resolution interchange classification dataset (HRIC dataset) using Gaofen-2 and Jilin-1 satellite imagery with a spatial resolution of 0.5–0.75 meters. The dataset covers six common types of urban interchanges, including cloverleaf interchange, diamond interchange, roundabout interchange, T-interchange, trumpet interchange, and turbine interchange. Based on this dataset, we evaluate the performance of multiple deep learning models on the interchange classification task and propose a structural feature–enhanced classification framework tailored for accurate interchange recognition. The framework incorporates core structural modeling concepts from the STAR module into the ResNet18 backbone to strengthen structural feature representation, while integrating the ELA attention mechanism to enhance sensitivity to local features. Experimental results demonstrate that the optimized model achieves strong performance in identifying various interchange structures, reaching an overall classification accuracy of 79.10%. This study provides an effective method for the automatic recognition and analysis of complex transportation infrastructure, and lays a foundation for future research on road segmentation and topological structure extraction.