Development of UAV-based river surface velocity measurement by STIV based on high-accurate image stabilization techniques

Ichiro Fujita, Yuichi Notoya, Mitsuru Shimono

Thursday 2 july 2015

14:50 - 15:05h at Africa (level 0)

Themes: (T) Special session, (ST) Acoustic monitoring of flow, turbulence and river discharge

Parallel session: 12F. Special session: Acoustic monitoring of flow, turbulence and river discharge

In recent years, the so-called unmanned air vehicles (UAV), remotely controlled airplanes or multi-copters, have become available for various civil engineering purposes. In the field of river engineering, although they have been used to examine river features such as the area of vegetation zone or grain size distributions of a bar, measurement of river flow has not been conducted by using them, probable due to the difficulty of stabilizing vibrating pictures inherently included in airborne images. Therefore, we developed a novel method to stabilize the airborne images with a high accuracy and the stabilized images were used to measure river surface distributions by a space time image velocimetry (STIV) technique. In the image stabilization process, novel computer vision techniques such as the Scale Invariant Feature Transform (SIFT) and the Rotation Invariant Phase Only Correlation (RIPOC) were combined to generate highly stable images by utilizing ground features outside of the river surface zone. The developed method was applied to investigate a snowmelt flood of the Uono River in 2014, in which UAV-based airborne video images measured at about 300m from the ground were used that covers a river reach of about four hundred meters. Although there appeared significant standing waves on the water surface, the aerial STIV succeeded in measuring two dimensional velocity distributions with a reasonable accuracy when compared with a concurrently conducted ADCP measurement or conventional STIV measurements using obliquely viewed images from a riverbank. It should be noted that another image analysis method, the large scale particle image velocimetry (LSPIV) is difficult to apply to such images due to the existence of standing waves. Another river surface structures such as a deflection of surface flow direction or variation of standing wave lengths were detected from multiply superposed river surface images.