面向国产高分辨率卫星的浒苔绿潮遥感监测算法对比  

Comparison on remote sensing monitoring algorithms of Ulva prolifera green tide using domestic high-resolution satellite images

作　　者：秦泉 胡先锋[1,2] 李峰 王晗 段金馈[1,2] 韩东枫 顾琛 QIN Quan;HU Xianfeng;LI Feng;WANG Han;DUAN Jinkui;HAN Dongfeng;GU Chen(Key Laboratory for Meteorological Disaster Prevention and Mitigation of Shandong,Jinan 250031,China;Shandong Climate Center,Jinan 250031,China;College of Geodesy and Geomatics,Shandong University of Science and Technology,Qingdao 266590,China)

机构地区：[1]山东省气象防灾减灾重点实验室,山东济南250031 [2]山东省气候中心,山东济南250031 [3]山东科技大学测绘与空间信息学院,山东青岛266590

出　　处：《海洋气象学报》2023年第2期64-75,共12页Journal of Marine Meteorology

基　　金：卫星海洋环境动力学国家重点实验室青年访问学者开放课题(QNHX2209);山东省气象局科研项目(2022sdqxm01);江苏省基础研究计划青年基金项目(BK20210667)。

摘　　要：利用高空间分辨率卫星影像准确监测浒苔绿潮对灾害早期发现、动态跟踪以及沿岸防御具有重要应用价值。目前面向高分辨率遥感影像已有多种浒苔提取方法,但不同遥感算法受到海水背景(浑浊和清澈海水)和外界观测环境(如云层、太阳耀斑、观测几何条件)等干扰,其监测效果可能受到不同程度影响。为此,以国产高分辨率GF-WFV和HJ-CCD影像为例,充分对比了归一化差值植被指数(normalized difference vegetation index,NDVI)、漂浮藻类高度虚拟基线指数(virtual baseline floating macroalgae height,VB-FAH)和绿度指数(tasseled cap greenness,TCG)在常见多种环境背景下提取浒苔的优势和不足。研究结果表明:在清澈海水、浑浊海水和弱太阳耀斑背景下,NDVI、VB-FAH和TCG三种算法均有较好的浒苔识别能力,其精度评价指标F1-score和总体分类精度(overall accuracy,OA)分别超过95.6%和95.2%。对于几何观测条件,VB-FAH和TCG算法对观测几何角度的变化不敏感并表现较高的稳定性,要优于NDVI方法。在云层覆盖和强太阳耀斑背景下,TCG算法的浒苔判识能力最好,并可有效排除云覆盖和强太阳耀斑的干扰,其精度评价指标F1-score和OA分别超过95.2%和95.0%。Using high spatial resolution satellite images to accurately monitor the Ulva prolifera(Ulva)green tide has important application value for early disaster detection,dynamic tracking,and coastal defense.Although there are many methods to extract Ulva from high-resolution satellite images,the performances of different remote sensing algorithms are influenced by the common various observing conditions including seawater background(turbid and clear seawater)and external observation conditions(such as cloud cover,sun glint,and observation geometry).Therefore,taking domestic high-resolution GF-WFV and HJ-CCD images as examples,this work compares the advantages and disadvantages of the normalized difference vegetation index(NDVI),virtual baseline floating macroalgae height(VB-FAH),and tasseled cap greenness(TCG)index in extracting Ulva under different environmental backgrounds.The results show that NDVI,VB-FAH,and TCG algorithms have good performance of Ulva extraction under clear water,turbid water,and weak sun glint conditions.Their accuracy evaluation indicators,F1-score and OA(overall accuracy),are greater than 95.6%and 95.2%,respectively.Under different geometric observation conditions,VB-FAH and TCG indexes are not sensitive to the change of observation geometric angle with high stability,performing better than NDVI.Under the background of cloud cover and strong sun glint,TCG method has the best ability to identify Ulva,and can effectively eliminate the interference of cloud cover and sun glint,whose accuracy evaluation indicators,F1-score and OA,are greater than 95.2%and 95.0%,respectively.

关 键 词：高分辨率卫星遥感 浒苔绿潮 遥感监测算法 高分卫星 环境卫星 

分 类 号：P76[天文地球—海洋科学]