期刊名称：American Journal of Geographic Information System
印刷版ISSN：2163-1131
电子版ISSN：2163-114X
出版年度：2018
卷号：7
期号：4
页码：99-106
DOI：10.5923/j.ajgis.20180704.01
语种：English
出版社：Scientific & Academic Publishing Co.
摘要：Estimation of water depths plays an important role in monitoring water level for solving a wide variety of engineering problems. Currently, bathymetric data are acquired based on single- or multi-beam echo-sounding and airborne Light Detection and Ranging (LiDAR) techniques. The field data collection at a site is expensive and time consuming. On the other hand, it is sometimes extremely difficult in shallow water regions. Satellite data can be a valuable alternative for determining shallow water depths. In this research, a new approach has been applied to determine the water depths in shallow area using Landsat-8 images. The method compromises three different steps: 1) selection of a set of pixels on the images with known water depths from the echo-sounding process as a training sample, 2) applying the Multi-layer Feed-forward (MLF) neural network and Binary Encoding (BE) classification algorithms to estimate the probability that each pixel belongs to each height in the training sample, ii) applying the fuzzy majority voting (FMV) algorithm to combine the probabilities from MLF and BE, iii) using the Inverse Probability Weighted Interpolation (IPWI) algorithm to convert the probabilities from FMV into water depths. When compared with the echo-sounding data, the proposed method showed reasonable results with mean error of 0.457m and standard deviation (SD) of 0.286m.
