Application of Masked Two-dimensional Fourier Spectrum for Improving Measurement accuracy of River Surface Flow

For establishing a proper management scheme of a river, it is crucial to obtain accurate and long-standing hydrological information such as rainfall intensity, water level and river discharge. Among them, discharge measurement is the most difficult task to perform especially in case of a flood under extreme flow condition. The conventional way to measure the discharge in Japan is to use pole-type floats but it is difficult dangerous to execute the measurement works in such a condition. Therefore, the so-called non-contact methods such as image-based techniques and surface velocity radar (SVR) have come to be available. The space-time image velocimetry (STIV) is one of the image-based techniques for measuring river surface flow from video images and the technique, available as a commercial software KU-STIV, is being used in the actual discharge measurements in Japan and Australia. However, in the course of measurements it is reported that there encountered several cases in which measurement accuracy is less reliable depending on the image shooting conditions.

To overcome such difficulties, we classified the image conditions into ten types; i.e. normal, shadow, light, wavy, standing wave, whitecap, random, obstacle, noise, and vague, and developed an image filter mainly based on the wavenumber-frequency spectrum (WFS) to improve the texture quality of space-time image (STI) that represent the flow information as a form of striped pattern. Specifically, we masked the region in WFS representing information other than the flow. The inverse transform of the masked spectrum can yield the quality of STI significantly. The whole process can be treated as a kind of filter for improving the measurement accuracy and robustness of the measurement system. Finally, the developed technique was applied to the ten types of STIs picked up from the actual image measurements with success of yielding more reliable flow data.