IMPROVING THE RESOLUTION OF SHALLOW-RIVER DISCHARGE MEASUREMENT BY DENOISING OUTLIERS USING BINARIZED IMAGE SUBTRACTION FEATURE

抄録

 Accurate measurement of river discharge for complex and shallow bathymetries needs sophisticated monitoring systems. On the other hand, observation site in most cases is considered as the most important challenge that must be resolved, otherwise additional uncertainties and errors will be introduced to discharge estimates. In this work, author tested a new version of the Fluvial Acoustic Tomography (FAT) system operated by high-frequency 53-KHz underwater acoustic transducers. Monitoring site, however, was very stringent and fixed. Consequently, arrival times data recorded by the FAT system was highly scattered which impaired the quality of discharge estimates and had produced imperfect results.

 In order to overcome this challenge, authors used a straightforward imagery solution, to retain only the desired data and eliminate outliers using image subtraction feature. The results showed that streamflow estimated by FAT had very good agreement with discharge estimated by the Rating Curve (RC) method. Also, it can be deduced that imagery methods are very powerful in eliminating noisy data when employed to time-series analysis.