Research On Defect Detection Algorithm Of Polyethylene Gas Pipe Based On Image Recognition

The urban gas pipe network is one of the lifeline projects related to people's livelihood,and has widely used gas polyethylene(PE)pipes.Once the gas polyethylene pipe leaks,it will have very serious consequences.Therefore,regular inspection of buried gas polyethylene pipes and maintenance based on the inspection results will greatly reduce the hazards caused by pipe defects.This paper proposes an algorithm for automatic detection of defects in polyethylene gas pipes.It can accurately and efficiently automatically determine whether there are defects in the polyethylene gas pipe or not and the types of defects.In this paper,according to the special characteristics of the PE gas pipes with a small diameter and polymer materials,and in combination with the knowledge of image recognition,an image acquisition device was designed for the internal inspection of PE gas pipes.Firstly,the mainboard selection and matching components of the hardware system were strictly selected,then the database of the software platform and the Open CV function library required for image recognition were installed,and finally the internal parameters of the camera was calibrated.Then a preprocessing algorithm for PE gas pipes was proposed.Pre-processing algorithm study was performed for the acquired images.After preliminary screening of the acquired images,the contrast between the background and defects of the pipe was enhanced using gamma correction method,and the type of noise present in the PE pipe defect image was analyzed to facilitate the design of the subsequent filtering algorithm,a double filtering algorithm was proposed to remove the pretzel noise using adaptive median filtering,using the Bilateral filtering to remove Gaussian noise.After that,improving the classic Sobel algorithm by adding new templates to the existing ones,comprehensive coverage of all directions,then adaptive threshold segmentation was used to obtain sub-images containing defects,and finally the morphological processing algorithm was used to fill the non-closed area and corroded the interference points with small area,which was convenient for the measurement of defect feature parameters.According to the criteria for extracting features,and a program was designed to store the extracted features to prepare for the next training data.Finally the defect classification algorithm suitable for PE gas pipes was studied.The support vector machine classifier with the advantage of small sample classification was used to distinguish various defect types,then the parameters of the kernel function used were optimized,and the classifier parameters with high accuracy were obtained.