Research on Image Recognition Method of In-service Pipeline Corrosion Fault

摘要

In this paper, for the practical demand of in-service pipeline detection, a new way called mathematical morphology wavelet de-noising innovative method has been developed, based on separate defect points. This method needs to extract the edge of defective parts by wavelet transform modulus maximum method, select some key characteristic parameters in favor of defects identification like fine length, moment invariant, gray energy and so on, and recognize patterns by means of single-output BP neural network. This method has been successfully applied to differentiate the weld joints and corrosion defects of pipelines, and quantitatively recognize the corrosion defects.