Visual detection algorithm of water line based on feature fusion

Image-based method of water level measurement uses image processing technology to detect water line and realize automatic water level acquisition. However, due to the complex light conditions, low imaging resolution and inclined angle of view in the field, the recognition of characters and graduation lines on the surface of water gauge is quite unreliable. Existing image methods are difficult to ensure long-term and effective measurement. In order to solve this problem, a method of water line detection based on Mean Difference Fusion (MDF) of gray and edge features is designed. By calculating the gray mean difference of gray image and edge image, the water line is located by rough to precise method. In-situ experiments under different light conditions were carried out at Qianhancun Hydrological Station in Nanjing, and compared with Otsu method. The accuracy of water level measurement can reach 1 cm.