Spatial distribution information extraction of winter wheat in Henan Province based on multi-source sentinel remote sensing data

Timely and accurately acquiring spatial distribution information for winter wheat is crucial for formulating agricultural policies. However, the information extraction for traditional spatial distribution for winter wheat relies primarily on manual field surveys. This method is labour-intensive and subjective and less accurate. With that remote sensing technology continues to develop, new avenues for extracting distribution information for winter wheat have emerged. Nevertheless, previous studies often did not fully leverage the advantages of remote sensing data from multiple sources, resulting in lower extraction accuracy. To address this, this study focuses on Henan Province, where Sentinel-2 optical imagery and Sentinel-1 radar imagery. Then support vector machine and random forest algorithms were used to construct winter wheat extraction model. The results indicate that the extraction accuracy of Sentinel-2 optical imaging surpasses Sentinel-1 radar imaging, with an overall accuracy of 82.04%, the producer's accuracy for winter wheat reaches 86.25%, the accuracy of user is 85.19%, the Kappa coefficient is 0.770. Further integration within the Sentinel-2 optical imagery with Sentinel-1 radar imagery significantly enhances extraction's accuracy. The overall accuracy improves to 90.19%, the accuracy for producers increases to 96.25%, user's accuracy rises to 95.06%, and the Kappa coefficient increases to 0.874. Additionally, by comparing different machine learning algorithms, it is observed that the algorithm random forest consistently outperforms the algorithm support vector machine in extraction accuracy regardless of the dataset. In this way, this study provides valuable guidance for obtaining information on the spatial distribution of winter wheat in Henan Province.