陈安娜, 马毅, 张靖宇. 水深被动光学遥感反演模型适用性研究[J]. 海洋环境科学, 2018, 37(6): 953-960. DOI: 10.12111/j.mes20180623
引用本文: 陈安娜, 马毅, 张靖宇. 水深被动光学遥感反演模型适用性研究[J]. 海洋环境科学, 2018, 37(6): 953-960. DOI: 10.12111/j.mes20180623
CHEN An-na, MA Yi, ZHANG Jing-yu. Study on applicability of water-depth passive optical remote sensing inversion models[J]. Chinese Journal of MARINE ENVIRONMENTAL SCIENCE, 2018, 37(6): 953-960. DOI: 10.12111/j.mes20180623
Citation: CHEN An-na, MA Yi, ZHANG Jing-yu. Study on applicability of water-depth passive optical remote sensing inversion models[J]. Chinese Journal of MARINE ENVIRONMENTAL SCIENCE, 2018, 37(6): 953-960. DOI: 10.12111/j.mes20180623

水深被动光学遥感反演模型适用性研究

Study on applicability of water-depth passive optical remote sensing inversion models

  • 摘要: 目前常用的水深遥感反演模型有对数线性模型和对数转换比值模型,近年新发展了一种改进的对数转换比值模型。本文分别选取不同水体类型的蜈支洲岛和槟城为研究区,基于GeoEye-1多光谱影像开展如下水深反演研究:首先对改进的对数转换比值模型进行水深反演能力验证,然后从三个维度对三种模型反演能力进行剖析与比较。研究表明:(1)改进的对数转换比值模型基本可以真实的反映水下地形变化,且反演水深与真实水深差值绝大部分控制在2 m以内,在10~20 m反演效果尤佳,水深差在1 m水深处上下浮动。(2)多维度剖析:①整体精度评价:在清澈海域蜈支洲岛周边水深反演精度最高的是改进的对数转换比值模型,MAE为1.56 m,MRE为13.5%,对数线性模型MAE为1.71 m,MRE为14.8%,对数转换比值模型次于前两者,在浑浊海域槟城周边水深反演精度最高的亦是改进的对数转换比值模型,而对数线性模型反演精度有所下降;②分水深段精度评价:随着水深值的增加,三种模型在蜈支洲岛反演精度均呈上升的趋势,在槟城研究区,对数转换比值模型和改进的对数转换比值模型反演规律与蜈支洲岛相似,而对数线性模型在15~20 m水深段,反演精度明显下降;③模型稳定性:改进的对数转换比值模型在不同水深段反演精度变化幅度最小且反演精度最高,且在不同的水质环境中具有相对稳定的反演能力,对数转换比值模型其次,而对数线性模型易受海水环境影响,尤其在浑浊水体反演精度不够稳定。

     

    Abstract: The Log-linear model and Stumpf model are most commonly water depth remote sensing inversion models, and a kind of improved stumpf model has been developed these years.In this paper, We choose Wuzhizhou and Penang as research areas based on the different water types and carry out the following water depth inversion based on GeoEye-1 multispectral image:Firstly, verify the inversion ability of the Improved Stumpf model.Secondly, do the analysis and comparison of three models' inversion ability from three dimensions.Here are the results:(1) The improved Stumpf model can reflect the changes of underwater terrain, the differerces between the inversion water depth and real water depth are mostly controlled within 2 m, especially the inversion effect in the 10~20 m is better, and the water depth differences are float around 1 m. (2) Multi-dimensional analysis:①The overall accuracy evaluation:the improved stumpf model has the highest inversion accuracy in Wuzhizhou with clear water, whose MAE is 1.56 m and MRE is 13.5%, the stumpf model's MAE is 1.71 m and MRE is 14.8%, while The logarithmic band ratio model is inferior to above.In Penang with turbid water, the improved stumpf model also has the highest accuracy while the accuracy of the logarithmic linear model is decreased.②The accuracy of differenct depths:The accuracy of three models in Wuzhizhou showed an increasing trend with the increase of the water depth.In Penang, the accuracy of Stumpf model and the improved Stumpf model are similar to those of the Wuzhizhou, while the accuracy of the logarithmic linear model has a significant decrease in 15~20 m depths.③Model stability:The improved stumpf model has the smallest variation range and the highest inversion accuracy, and has relatively stable inversion ability in different water environment; The stumpf model follows next, while The log-linear model has a large variation about inversion results in different water quality environments especially in the turbid water.

     

/

返回文章
返回