Detecting flow regime changes from the spatial RGB distribution of turbid water images using a three-dimensional vector space model

Kazuhide IGARASHI, Tasuku MIZUTANI, Atsuhiko KINOSHITA, Shinji IBUKA, Akihide HIGO, Tsuyoshi IKESHIMA, Hiromi AKITA and Hideaki MIZUNO


In this report, we introduce a new method for automatically detecting flow‐regime changes using unit vectors and the normal unit vector of RGB values from camera imagery. Although the method is still under development, it is expected to lead to further enhancement of the monitoring system for sediment disasters. From the results of a casestudy at the Inari river in Nikko, Japan, it is suggested that the RGB unit vectors may have eigenvalues according to different types of water color: e. g., clear water, turbid water, and debris flow color. It is also suggested that the RGB unit vectors can draw a curve among the unit vectors of pre and post‐turbidity colors. Furthermore, even if the amount of solar radiation changes, the RGB unit vectors can indicate almost stable values.

Key words

image analysis technique, auto‐sensing, stream‐monitoring