Rapid and accurate acquisition of crop acreage information is helpful for crop growth monitoring, yield estimation, pest monitoring and early warning. Sentinel-1/2 was used as the data source for the characteristics of cloud and fog in Jiangnan region of China，and multi-source data was used, such as optical remote sensing images and synthetic aperture radar (SAR) images. Then, according to the characteristics of crop mulching in early spring, the SAR plastic-film vegetation index (SPVI) was constructed. A crop identification method based on XGBoost algorithm was studied by using time series spectrum and vegetation index characteristics.Finally,taking Xuanzhou District, Xuancheng City, Anhui Province as the research area, a case study was carried out to verify the results. Sentinel-2 images were used as the main method to obtain the time series index, and four Sentinel-1 radar images were added to supplement the optical images missing in the cloud and rain period (May to July). The overall accuracy in cloud coverage area was 84.87%, which was better than that of random forest (83.93%).And the accuracy of tobacco identification mapping was 88.69%, and the user's accuracy was 95.51%. The overall accuracy of the Sentinel-2 image was 79.01%, the mapping accuracy of tobacco was 82.30%, and the user's accuracy was 93.49%. The research results showed that the multi-source remote sensing crop identification method based on the constructed XGBoost algorithm can meet the accuracy requirements.