Abstract: Accurate identification and spatial mapping of high-standard farmland (HSF) are essential for ensuring national food security, optimizing agricultural land management, and evaluating the effectiveness of land consolidation policies. However, recognizing HSF through remote sensing remains challenging in hill–plain transition zones, where fragmented landscapes and heterogeneous land cover obscure spectral signals. Traditional classification methods mainly rely on natural attributes such as spectral and phenological features and therefore fail to effectively distinguish HSF from general cultivated land, which is defined not only by crop type but also by engineering and management standards such as parcel regularity, irrigation, and drainage infrastructure. To address this limitation, this study developed and validated a multidimensional feature fusion methodology that integrates policy-derived spatial attributes with traditional remote sensing features for automated and high-precision HSF extraction. Zhangfeng Town in Dehong Prefecture, Yunnan Province, China, a representative hill–plain transition zone, was selected as the study area. Multi-source datasets, including preprocessed Sentinel-2 and Gaofen-2 imagery combined with a 10 m Digital Elevation Model (DEM), were used to construct a comprehensive 20-dimensional feature set comprising four feature types: (1) spectral features from ten Sentinel-2 bands; (2) spectral indices including NDVI, RVI, NDWI, and OSAVI to enhance vegetation and moisture information; (3) phenological features—Start of Season (SOS), End of Season (EOS), and Length of Season (LOS)—derived from time-series NDVI using a threshold and linear interpolation method; and (4) regional features—Patch Density (PD) and Ditch Density (DD) to quantify parcel fragmentation and irrigation infrastructure, together with slope from the DEM. Five feature combination schemes (S1–S5) were designed to evaluate the contribution of each feature type, and two machine learning algorithms, Random Forest (RF) and eXtreme Gradient Boosting (XGBoost), were employed for classification using identical training and validation samples obtained from field surveys and high-resolution image interpretation. The results show that the introduction of regional features significantly enhanced classification performance. Compared with the spectral-only scheme (S1), the scheme incorporating regional features (S4) improved overall accuracy by 9.41% for RF and 5.60% for XGBoost, highlighting the importance of quantifying policy-related spatial characteristics such as parcel regularity and ditch networks. The full-fusion scheme (S5) achieved the best performance, with overall accuracies of 96.15% (Kappa = 0.91) for RF and 96.79% (Kappa = 0.93) for XGBoost. Feature importance analysis indicated that DD and PD were the most influential predictors, contributing more to the model than any single spectral or phenological variable. Spatial pattern analysis revealed that HSF in the study area covers 58.44 km² and exhibits a distinct pattern of “basin aggregation and hilly fragmentation.” Specifically, 84.43% of HSF occurred on slopes less than 15°, and 83.13% was located within 300 m of major ditch networks, reflecting the combined effects of topographic suitability and irrigation accessibility and consistent with national HSF construction principles. This study demonstrates that quantifying and integrating policy-derived regional features with multi-source remote sensing data and advanced machine learning algorithms such as XGBoost can greatly improve the accuracy of HSF identification in complex terrains. The proposed framework provides a reliable and automated technical pathway for monitoring HSF construction and evaluating policy implementation effectiveness, offering scientific support for precision land management and sustainable agricultural development in mountainous agricultural regions.