结合底质分类与SVR算法的多光谱影像测深  

Multi-spectral Water Depth Inversion Based on Bottom Sediment Classification and SVR Algorithm

作　　者：朱金山[1,2,3] 宋珍珍 赵露露 ZHU Jinshan

机构地区：[1]山东科技大学测绘科学与工程学院,山东青岛266590 [2]地理信息工程国家重点实验室,陕西西安710054 [3]自然资源部海岛(礁)测绘技术重点实验室,山东青岛266590

出　　处：《地理空间信息》2019年第11期44-46,74,I0001,共5页Geospatial Information

基　　金：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);2015年测绘地理信息公益性行业科研专项资助项目(201512034)

摘　　要：针对目前遥感水深反演方法都是以海底底质均匀为前提而缺乏对混合海底底质研究的问题,提出了一种结合底质分类与SVR算法的水深反演模型。利用WorldView-2多光谱遥感影像对南海北岛岛礁周围混合底质的浅海海域进行底质分类,并对底质分类后的不同海域分别建立线性回归与SVR非线性回归多种测深模型。通过水深分段验证后结果表明,结合底质分类的SVR非线性回归方法更适合混合底质的浅水水深反演。In view of the fact that the current remote sensing water depth inversion methods are based on the premise of uniform sea floor sediments and lack of research on mixed sea bottom sediments, we proposed a water depth inversion model combining bottom sediment classification with SVR algorithm in this paper. We used the World View-2 multi-spectral remote sensing images to classify the bottom sediments of the mixed-substrate shallow sea area around the Bei Island reef in the South China Sea at first. And then, we established multiple sounding models of linear regression and SVR non-linear regression for the different sea areas after bottom sediment classification. The subsection verification result of water depth shows that the SVR non-linear regression method combining with bottom sediment classification is more suitable for shallow water depth inversion of mixed sediments.

关 键 词：水深反演 底质分类 SVR算法 多光谱 

分 类 号：P237[天文地球—摄影测量与遥感]