Bridge Crack Image Segmentation Based on Improved Watershed Algorithm
作者单位：School of Electrical and Information Engineering Guangxi University of Science and Technology
会议名称：《第30届中国控制与决策会议》
会议日期：2018年
学科分类：081406[工学-桥梁与隧道工程] 08[工学] 080203[工学-机械设计及理论] 0802[工学-机械工程] 0814[工学-土木工程] 082301[工学-道路与铁道工程] 0823[工学-交通运输工程]
基 金：supported by Guangxi Key Research and Development Program(No.AB17292053) Guangxi Graduate Education Innovation Program(Project No.GKYC201624)
关 键 词：Bridge Crack Watershed Algorithm Butterworth Low-pass Filter Over-segmentation Local Minimum
摘 要：In order to solve the problem of over-segmentation of bridge fractures by the watershed algorithm, an improved watershed segmentation algorithm based on threshold marker is proposed. Firstly, the original gradient image is obtained from the gray image by using morphological operation, and the second-order Butterworth low-pass filtering is performed on the original gradient image to obtain the low-frequency component of the gradient image. And then apply the best threshold for the use of Otsu method to be applied to the morphological extended minimum transform technique, and the technique is used to extract the local minimum value which is highly correlated with the bridge crack feature information from the low frequency components of the gradient image;Finally, the original gradient image is used to compose the minimum gradient of the original gradient image to obtain the modified gradient image, and watershed segmentation is performed on these gradient images. The experimental results show that the crack image of the bridge is accurate and clear, and the over-segmentation phenomenon is effectively reduced.