基于遥感云计算平台的四川盆地水稻种植面积提取

Rice Planting Area Extraction Method in Sichuan Basin Based on Remote Sensing Cloud Computing Platform

四川盆地是中国重要的水稻产区,提取水稻种植面积有助于水稻产量估算和农业政策制定。本文提出一种基于Sentinel-1 SAR数据和遥感云计算的四川盆地水稻种植面积提取方法,首先利用PIE-Engine平台提供的充足存储算力进行大区域遥感时空分析,然后基于农业普查信息进行水稻物候窗口划定,采用多时相均值滤波抑制时序噪声,最后按照层次分类思想对非水稻特征进行逐步分离,并选择大津法实现分类阈值的自适应确定。使用该方法提取了四川盆地2020年水稻种植面积,样本检验精度达到82.08%,与农业部门统计数据有较好的一致性。研究结果表明提出的方法在不依赖训练样本和经验阈值的基础上实现了水稻种植面积的高效提取,有望在更大区域推广应用,为保障国家粮食安全和社会稳定提供信息支撑。

Sichuan Basin is a key region for rice cultivation.Identifying rice planting area contributes to rice yield estimation and agricultural pol-icy formulation.We proposed a novel method to identify rice planting area in Sichuan Basin by using Sentinel-1 SAR data and remote sensing cloud computing platform.Then,we conducted large-scale spatio-temporal analysis through the abundant storage and computational resources provided by the PIE-Engine platform.We defined the key phenological windows based on agricultural census information and applied a multi-temporal mean filter to suppress temporal noise.Meanwhile,we adopted a hierarchical classification scheme to distinguish non-rice fea-tures and used the OSTU algorithm to adaptively determine classification thresholds.Applying this method to identify rice planting areas in Sich-uan Basin for the year 2020,we achieved an overall accuracy of 82%,aligning well with official agricultural statistics.This method efficiently identifies rice planting area without the need for training samples or empirical thresholds,and holds promise for broader applications on a large scale,providing essential information to ensure national food security and social stability.