基于Sentinel-2A MSI数据的水体信息提取算法对比研究  

Comparison of Water Information Extraction Algorithms Based on Sentinel-2A MSI Data

作　　者：张磊 韩秀珍[2] 翁富忠 丘仲锋 Zhang Lei;Han Xiuzhen;Weng Fuzhong;Qiu Zhongfeng(School of Marine Sciences,Nanjing University of Information Science&Technology,Nanjing 210044,Jiangsu,China;National Satellite Meteorological Center,Beijing 100081,China;State Key Laboratory of Hazardous Weather,Chinese Academy of Meteorological Sciences,Beijing 100081,China)

机构地区：[1]南京信息工程大学海洋科学学院,江苏南京210044 [2]国家卫星气象中心,北京100081 [3]中国气象科学研究院灾害天气国家重点实验室,北京100081

出　　处：《激光与光电子学进展》2022年第12期495-505,共11页Laser & Optoelectronics Progress

基　　金：国家重点研发计划(2018YFC1506501);国家卫星气象中心山洪地质灾害防治气象保障工程2020年建设项目(IN_JS_2020041)。

摘　　要：地表水体信息的准确提取对于水资源规划管理与相关灾害监测等具有重要意义。以Sentinel-2A MSI影像为研究数据,基于10种典型光谱波段指数算法对鄱阳湖周边区域的水体进行了提取,并以结合目视解译的synthetic aperture radar(SAR)影像水体识别结果作为真值进行了对比分析。实验结果表明:归一化差异水体指数、高斯归一化差异水体指数、伪归一化差异水体指数的识别效果最好,平均F1-score超过95%,且阈值稳定程度较高;改进的归一化差异水体指数适用于城镇区域的水体提取;组合水体指数在水田以及河流区域识别效果较好;增强型水体指数、阴影水体指数均适用于山区,其中增强型水体指数在城镇区域也具备适用性;修订型归一化差异水体指数仅适用于针对大面积水体的识别;面向阴影的自动化水体指数能够有效剔除地形阴影造成的干扰,而非面向阴影的自动化水体指数的整体表现相对较差。Accurate extraction of surface water information is important for water resource planning and management and related disaster monitoring. Herein, using Sentinel-2A MSI images as the research data, the water body around Poyang Lake was extracted based on 10 typical spectral band index algorithms and the water body recognition results of synthetic aperture radar images combined with visual interpretation were compared and analyzed as the true value. The experimental results demonstrate that the recognition effect of the normalized,Gaussian normalized, and pseudo normalized difference water body indexes is the best. The average F1-score is greater than 95%, and the threshold stability is high. An improved normalized difference water body index is suitable for water extraction in urban areas. The combined water body index has good recognition effect in paddy fields and river areas. The enhanced water body and shadow water body indices are applicable to mountainous areas.The enhanced water body index is also applicable in urban areas. The revised normalized difference water body index is only applicable to the identification of large-area water bodies. A shadow-oriented automatic water index can effectively eliminate the interference caused by terrain shadows. However, the overall performance of a nonshadow oriented automatic water index is relatively poor.

关 键 词：遥感 水体识别 水体指数 算法对比 Sentinel-2A MSI 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]