Research on Segmentation Features in Visual Detection of Rust on Bridge Pier

The detection of bridge rust is an essential work in the daily maintenance of bridges. For the detection and recognition of rust in bridge pier images, firstly, image segmentation and erosion expansion are preprocessed. Then, the area of rust is counted and the rust grade is determined according to the standard. The quality of bridge image is easily affected by environmental factors, which leads to the increase of segmentation error. The selection of segmentation features is the key to improve the segmentation effect. Gray value, R, G, B single color feature and the combination of hue and saturation are selected as segmentation features respectively to detect and recognize the rust of bridge piers. The experimental results of 35 pier images show that when the combination of hue and saturation is used as segmentation feature, the rust disease can be effectively detected and identified, and the accuracy rate is more than 94%.