岛礁周边水深光学遥感反演模型比较与分析

Comparison and analysis of bathymetric optical remote sensing inversion models around islands

岛礁周边水深精确测量是海洋测绘工作的难点之一,光学遥感技术在浅水区域尤其是无人到达的深远海岛礁区域有其独特的优势,可作为传统测深的有效补充手段。利用蜈支洲岛WorldView-2多光谱遥感影像和实测水深数据,对比了多波段回归模型、对数波段比值模型、支持向量机回归模型(support vector regression,SVR)、BP神经网络模型4种光学遥感水深反演模型。结果表明,BP神经网络和SVR具有精度优势,而多波段回归模型和对数波段比值模型具有效率优势,经综合分析,多波段回归模型具有较高精度和效率,适用于岛礁周边的水深反演。

Accurate bathymetry around islands and reefs is one of the difficulties in marine surveying and mapping.Optical remote sensing technology has its unique advantages in shallow water areas,especially in uninhabited deep-sea islands and reefs,and can be used as an effective supplementary means to traditional bathymetry.In this paper,using WorldView-2 multispectral remote sensing images and actual bathymetry data of Wuzhizhou Island,four optical remote sensing bathymetry inversion models,namely,multi-band regression model,log-band ratio model,support vector regression(SVR)model and BP net model,are compared,and the results show that BP net and SVR have the advantage of accuracy,and multi-band regression and log-band ratio have the advantage of efficiency.The multi-band regression model has high accuracy and efficiency,and is suitable for bathymetric inversion around islands and reefs.