The rapid acquisition of information on winter wheat planting areas and their changes using remote sensing technology is of great significance for ensuring national food security and promoting sustainable regional agricultural development. Temporal feature curves of winter wheat, winter rapeseed, and other land cover types were established by using Sentinel-1 SAR and Sentinel-2 optical images during the phenological period of winter wheat on the Google Earth Engine (GEE) platform. By qualitatively and quantitatively analyzing the differences among the temporal feature curves of various land cover types, the optimal features for winter wheat extraction were identified. Based on these optimal features, a hierarchical classification approach for winter wheat was developed. This method firstly distinguished winter crops from other land cover types using vegetation index time series, and then applied the OTSU algorithm to separate winter wheat from winter rapeseed, thereby achieving high-precision extraction of winter wheat. The results showed that among the three hierarchical classification algorithms, the "NDPI_TWDTW & NDSVI_OTSU" algorithm achieved the highest extraction accuracy for winter wheat (OA of 0.927, Kappa of 0.854), which was comparable to that of the "RF_Full Feature" algorithm (OA of 0.931, Kappa of 0.863). The "decision-level fusion" of the three algorithms had a limited effect on improving accuracy. Due to speckle noise, SAR images exhibited lower extraction accuracy for winter wheat in the VV (OA of 0.879, Kappa of 0.757) and VH (OA of 0.898, Kappa of 0.796) bands compared with the NDSVI optical vegetation index (OA of 0.927, Kappa of 0.854). Sentinel-1 SAR and Sentinel-2 optical data complemented each other in spatiotemporal resolution and characteristic information. Under cloudy and rainy conditions, SAR data effectively supplemented optical remote sensing data for winter wheat extraction, enhancing the timeliness and stability of regional winter wheat monitoring. The proposed hierarchical classification approach demonstrated mechanistic interpretability in feature analysis, low complexity, high accuracy, and strong robustness, and can be effectively applied to the identification and change monitoring of winter wheat over large areas and long time series.