ABSTRACT Small water bodies have always been an important part of water ecology systems. In the past, due to the limitations of satellite spatial resolution and recognition method precision, there have been few satisfactory remote sensing small water bodies extraction methods. In this article, a method based on index composition and HSI (hue, saturation, and intensity) colour space transformation is proposed to precisely extract small water bodies. An easy-to-deploy, fast, universal, and effective algorithm is used to accurately identify paddy fields and exclude shadows. This method is tested and verified with Sentinel-2 MSI (MultiSpectral Imager) images in seven cities in the Guangdong-Hong Kong-Macao Greater Bay Area. Compared with the traditional modified normalized difference water index (MNDWI) and enhanced water index (EWI) water extraction methods, the proposed HSI method has shown a better performance in small water bodies mapping with a kappa coefficient of 0.94, overall accuracy of 97%, producer’s accuracy of 96%, and user’s accuracy of 98% in test regions, which is significantly higher than the benchmarking water extraction methods. It provides a powerful supplement for the remote sensing monitoring of water resources in surface water bodies. The method proposed in this study exhibits extendibility, it also has the potential to extract other small features with minor modifications of the method.

中文翻译：

---

从 Sentinel-2 MSI（多光谱成像仪）图像绘制的小水体具有更高的精度

摘要 小水体一直是水生态系统的重要组成部分。过去，由于卫星空间分辨率和识别方法精度的限制，很少有令人满意的遥感小水体提取方法。本文提出了一种基于指标组合和HSI（色调、饱和度和强度）颜色空间变换的方法来精确提取小水体。采用易于部署、快速、通用且有效的算法来准确识别稻田并排除阴影。该方法在粤港澳大湾区七个城市的 Sentinel-2 MSI（多光谱成像仪）图像上进行了测试和验证。与传统的改进归一化差异水指数（MNDWI）和增强水指数（EWI）水提取方法相比，所提出的HSI方法在小水体测绘中表现出更好的性能，kappa系数为0.94，总体准确率为97%，生产者准确率为 96%，测试区域用户准确率为 98%，明显高于基准水提取方法。为地表水体水资源的遥感监测提供了有力的补充。本研究中提出的方法具有可扩展性，它还具有通过对该方法进行微小修改来提取其他小特征的潜力。生产者准确率为 96%，测试区域用户准确率为 98%，明显高于基准水提取方法。为地表水体水资源的遥感监测提供了有力的补充。本研究中提出的方法具有可扩展性，它还具有通过对该方法进行微小修改来提取其他小特征的潜力。生产者准确率为 96%，测试区域用户准确率为 98%，明显高于基准水提取方法。为地表水体水资源的遥感监测提供了有力的补充。本研究中提出的方法具有可扩展性，它还具有通过对该方法进行微小修改来提取其他小特征的潜力。