Abstract: To aim at the characteristics of the hydrological monitoring video images, such as the high
signal to noise ratio and the obvious color differences among targets and the specific location of target
regions, a water level reading method is proposed with strong practicability and little interference
from external factors. Due to the fixed content of cameras and only blue and red scales of
hydrological rulers, it extracts color components according to the characteristic color of the ruler.
After determining the adaptive threshold, the image is changed with the binary processing, denoised
and morphological processing. The ruler successfully is segmented using Hough transform to
determine the boundary of the hydrological ruler. All these work for reading real-time water level by
edge detecting. The experiment results show that this method can quickly segment and get the region
of interest efficiently.

Key words: water level monitoring, hydrological ruler characteristics, adaptive threshold, Hough
transform, image segmentation