Identification method of plot-scale rice cultivation in southern hilly regions
Aiming at the hilly area with cloudy rain and complex plot distribution in southern China,a new method for multi-season rice plot level structure information recognition is proposed.By utilizing high-resolution optical imagery and time-series Sentinel-1A SAR imagery,integrated with the Psi-Net model for multi-task semantic segmentation,the study analyzes the relationship between rice growth phenology and backscatter coefficients.The method achieves precise extraction of multi-season rice planting distributions through thresholding,validated in Pucheng County,Fujian Province.Results demonstrate superior performance in shape preservation and boundary accuracy,with Hausdorff distance of 21.368,notably better than the single-task U-Net network.Overall,the method achieves 88.6％and 87.7％accuracies for mid-season and late-season rice identification,respectively,with Kappa coefficients of 0.752 and 0.738.These findings underscore the significant application potential and practical value of the proposed approach for rice planting plot recognition under complex climatic and terrain conditions in southern regions.
万方数据
维普
