Abstract
This article introduces multiangular imagery into physics-based bathymetry in order to compensate for the shortage of bathymetric spectral bands caused by the low spectral resolution of current high-spatial-resolution satellite multispectral imagery. The focus is to propose a selective bathymetric retrieval method to eliminate the negative effect of nonoptimal image data on depth retrieval in multiangular imagery-based bathymetry. The elimination of the negative effect is implemented by excluding nonoptimal pixels in every individual image from bathymetric retrieval. An empirical criterion is designed for the determination of nonoptimal pixels. The proposed method can use multiangular image data selectively, avoiding situations where bathymetric retrieval results from the whole multiangular imagery are poorer than that from a part of the individual images. The method was tested in two typical areas within the Xisha (Paracel) Islands of the South China Sea using two-angle WorldView-2 multispectral images. The test showed that the derived depths of the method (i.e., depths derived from the selective image data) provided a better fit to the validation depths than those from the entirety of both images. The underestimation of depths derived from the entirety of both images was also improved to some extent.
Highlights
B ATHYMETRY is the study of the underwater depth of lake or ocean floors [1]
Three image data options were available for every pixel in selective bathymetric retrieval: the pixel data of one single image, the pixel data of the other single image, and the pixel data of both images
The image dataset used at all pixel locations after the exclusion of nonoptimal pixels is referred to as “selective image data,” which are a part of the multiangular imagery but do not include its nonoptimal pixels
Summary
Manuscript received July 7, 2020; revised September 13, 2020 and October 29, 2020; accepted November 18, 2020. Date of publication November 24, 2020; date of current version January 6, 2021. Physics-based bathymetry is one of the most widely used SDB approaches [4]. It can derive bathymetric information from satellite multispectral or hyperspectral imagery with an accuracy of 1–2 m root-mean-square error (RMSE), down to a depth of about 20 m [5]–[11]
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
