Study on Image Identification Method of In-service Pipeline Corrosion Fault
Liang Zhu, Hongyi Liu, Yuan Pei-xin · 2010
In this paper, a new mathematical morphology wavelet denoising method which based on separating defect points was put forward for the actual needs of the in-service pipeline inspection. This method uses wavelet maximum value algorithm to extract the edge of defect area. It use the single-output mode of BP neural network in pattern recognition by choosing small length, invariant moment, grey energy and other key characteristic parameters which is in favor of defect identifying. This method achieved the classification of pipeline weld and corrosion defect, and achieved the quantitative identification of corrosion defect.