| 多光谱遥感影像的两种浅海水深反演模型对比与分析 |
| |
| 引用本文： | 戚甲伟,任照宇,赵金秀,朱金山.多光谱遥感影像的两种浅海水深反演模型对比与分析[J].海洋学研究,2020,38(1):50-58. |
| |
| 作者姓名： | 戚甲伟 任照宇 赵金秀 朱金山 |
| |
| 作者单位： | 1.地理信息工程国家重点实验室,陕西 西安 710054;2.山东科技大学 测绘科学与工程学院,山东 青岛 266590;3.自然资源部 海洋测绘技术重点实验室,山东 青岛 266590 |
| |
| 基金项目： | 地理信息工程国家重点实验室开放研究基金资助项目(SKLGIE2017-Z-3-3);国家重点研发计划课题资助(2018YFC1405903) |
| |
| 摘 要： | 利用WorldView-2高分辨率卫星影像,以南海北岛附近海域为研究区,研究了两种水深反演模型——对数变换模型(Stumpf 2003)和双波段线性回归模型(Lyzenga 1985)。分析了不同底质情况下水深与各波段的相关性,并利用L-M(Levenberg-Marquardt)算法求解模型参数,然后对两种模型反演的水深结果的精度进行了对比分析。对于珊瑚底质,Lyzenga 1985模型水深反演的决定系数和均方根误差分别为0.902和1.651,均优于Stumpf 2003模型(0.882,6.421);对于砂质底质,Lyzenga 1985模型水深反演的决定系数和均方根误差分别为0.897和0.529,均优于Stumpf 2003模型(0.779,0.723)。可见,在水体清澈的珊瑚底质和砂质底质区域,Lyzenga 1985模型的水深反演精度均优于Stumpf 2003模型,Lyzenga 1985模型普适性更强,能够呈现出较为稳定的反演效果。
|
| 关 键 词： | 水深反演 对数变换模型 双波段线性回归模型 L-M算法 反演精度 |
| 收稿时间： | 2019-06-13 |
Comparison and analysis of two shallow water depth inversion models using multispectral remote sensing images |
| |
| QI Jiawei,REN Zhaoyu,ZHAO Jinxiu,ZHU Jinshan.Comparison and analysis of two shallow water depth inversion models using multispectral remote sensing images[J].Journal of Marine Sciences,2020,38(1):50-58. |
| |
| Authors: | QI Jiawei REN Zhaoyu ZHAO Jinxiu ZHU Jinshan |
| |
| Institution: | 1. State Key Laboratory of Geo-information Engineering, Xi'an 710054, China;2. College of Geomatics of Shandong University of Science and Technology, Qingdao 266590, China;3. Key Laboratory of Marine Surveying and Mapping, Ministry of Natural Resources, Qingdao 266590, China |
| |
| Abstract: | Using the WorldView-2 high-resolution satellite images, two models were studied: log transform model(Stumpf 2003) and dual bands linear regression model(Lyzenga 1985). Firstly, the correlation between water depth and each band was analyzed under different sediment conditions. Then the L-M(Levenberg-Marquardt) algorithm was used to solve the model parameters. Finally, the accuracy of the two models was compared and analyzed. The results show that for coral sediment, the determination coefficient and root mean square error of Lyzenga 1985 model are 0.902 and 1.651 respectively, which are better than those of Stumpf 2003 model(0.882, 6.421); For sandy sediment, the determination coefficient and root mean square error of Lyzenga 1985 model are 0.897 and 0.529 respectively, which are better than those of Stumpf 2003 model(0.779, 0.723). It can be seen that Lyzenga 1985 model is better than Stumpf 2003 model in the area of clear coral and sandy sediment, and Lyzenga 1985 model has stronger universality and can present relatively stable inversion effect. |
| |
| Keywords: | bathymetric inversion logarithmic transformation model two-bands linear regression model Levenberg-Marquardt algorithm the inversion accuracy |
|
| 点击此处可从《海洋学研究》浏览原始摘要信息 |
| 点击此处可从《海洋学研究》下载免费的PDF全文 |
|
