Optimized Extraction Method of Fruit Planting Distribution Based on Spectral and Radar Data Fusion of Key Time Phase

With China’s fruit tree industry becoming the largest in world, accurately understanding spatial distribution of growing areas is crucial for promoting socio-economic development and rural revitalization. Remote sensing offers unprecedented opportunities monitoring. However, previous research has mainly focused on UAV near-ground remote sensing, with limited accuracy obtaining information through satellite sensing. In this study, we utilized Google Earth Engine (GEE) cloud platform integrated data from Sentinel-1, Sentinel-2, SRTM sources. We constructed a feature space by extracting original band features, vegetation index polarization terrain texture features. The sequential forward selection (SFS) algorithm was employed optimization, combined machine learning object-oriented classification model used to extract crop distributions comparing key temporal phases trees. results revealed that backscatter coefficient features Sentinel-1 had highest contribution classification, followed while relatively smaller contribution. jujube plantation observed November (99.1% user 96.6% producer accuracy), whereas lowest found pear same month (93.4% 89.0% accuracy). Among four different methods, random forest (RF + OO) exhibited (OA = 0.94, Kappa 0.92), support vector (SVM) method 0.52, 0.31). total area Aksu City 2022 estimated be 64,000 hectares, walnut, jujube, pear, apple trees accounting 42.5%, 20.6%, 19.3%, 17.5% area, respectively (27,200 13,200 12,400 11,200 respectively). SFS optimization RF OO-combined selected study effectively mapped planting areas, enabling estimation based image data. This approach facilitates accurate real-time monitoring provides valuable management relevant departments.