时序SAR与光学影像的南方丘陵区水稻种植识别

Rice extraction in southern hilly areas by fusing time-series SAR features and optical images

针对南方地区的水稻提取存在长时序光学遥感应用受限的问题,该文基于Sentinel-1 SAR时序数据和Sentinel-2数据,以江西省南昌县为例,提出一种融合SAR时序特征及光学影像的南方地区水稻种植识别方法。该方法通过组合时序SAR特征、红边波段、EVI和LSWI指数特征,采用随机森林算法构建水稻提取模型进行水稻种植信息的提取,并与不同分类方法及数据集的提取结果进行对比分析。结果表明,基于多源数据集的随机森林水稻提取方法可以更加有效提高水稻种植信息提取精度,总体精度和Kappa系数分别达到92.67%和0.91。研究结果可为多云雨的南方地区水稻种植提取提供参考,具有一定的应用价值。

Aiming at the problem of limited application of long time series optical remote sensing for rice extraction in southern regions,this paper proposes a method to identify rice cultivation in southern regions by fusing SAR time series features and optical images based on Sentinel-1 SAR time series data and Sentinel-2 data,taking Nanchang County of Jiangxi Province as an example.The method uses a random forest algorithm to construct a rice extraction model for the extraction of rice planting information by combining time-series SAR features,red-edge band,EVI and LSWI index features,and compares and analyzes the extraction results with those of different Classification methods and datasets.The results show that the random forest rice extraction method based on multi-source datasets can more effectively improve the extraction accuracy of rice planting information,with the overall accuracy and Kappa coefficient reaching 92.67%and 0.91,respectively.The results of the study can provide a reference method for rice planting extraction in southern regions with cloudy rains and have some application value.