Abstract: The fine-scale extraction of crop planting structures (CPSs) provides a crucial foundation for accurately determining irrigation water volumes and ensuring food security. In southern China, characterized by cloudy and rainy conditions, single-source remote sensing data are insufficient to simultaneously characterize the spatial details and temporal variations of crops, thus limiting classification accuracy. This study investigated the CPS of Jiangning District, Nanjing City, southern China. Based on Sentinel-1 SAR and Sentinel-2 optical images, combined with the U-Net and bidirectional long short-term memory (BiLSTM) neural network models, this study acquired the spatiotemporal differences across crops, achieving fine-scale CPS identification in the study area. The results indicate that land in the Jiangning District is primarily used for gardens, paddy rice, rapeseed, wheat, and vegetables. Specifically, garden land is distributed in the southeastern hilly area; paddy rice is concentrated in polder areas along rivers and lakes; wheat and rapeseed are planted in traditional farming areas with favorable irrigation conditions, serving as preceding crops in rotation before paddy rice, and vegetables are primarily grown around residential areas. Compared to individual neural network models, the classification method based on both BiLSTM and U-Net models proposed in this study achieved higher accuracy and stability in crop identification. The Bi-LSTM model allowed for the effective characterization of the radar time-series characteristics of paddy rice in its typical growth stages, while the U-Net model enabled the accurate discrimination among crops with similar phenological traits, such as wheat and rapeseed, by capturing their spatial morphological differences. Owing to the synergy of both models, the proposed method achieved an overall classification accuracy of 93.20%, representing improvements of 3.16 percentage points and 4.97 percentage points, respectively, compared to individual BiLSTM and U-Net models.