Detection algorithm of defects on polyethylene gas pipe using image recognition. (June 2021)

• Record Type: Journal Article
• Authors: Li, Cong; Lan, Hui-Qing; Sun, Ya-Nan; Wang, Jun-Qiang
• Is Part Of: International journal of pressure vessels and piping. Volume 191(2021)
• Publication Date: 2021-06
• Subjects: Image recognition -- Polyethylene gas pipeline -- Defect detection -- Support vector machine; Pressure vessels -- Periodicals; Pipe -- Periodicals; Récipients sous pression -- Périodiques; Tuyaux -- Périodiques; Pipe; Pressure vessels; Periodicals; 681.76041
• DOI: 10.1016/j.ijpvp.2021.104381
• Languages: English
• ISSNs: 0308-0161

Abstract

Abstract: Aiming at the defects in the polyethylene (PE) gas pipeline, a defect detection algorithm based on image recognition is proposed. Firstly, the collected image is preliminarily screened to obtain the image with defects. Gamma correction algorithm is used to enhance the image, and then dual filtering is used to eliminate the interference of noise on image recognition. Then the defect edge is extracted by the improved traditional Sobel edge detection algorithm. Image defects are segmented by adaptive threshold method, and defect contours are effectively extracted through an open operation. Finally, the defect feature parameters are extracted and inputted into an optimized support vector machine for defect recognition. The experimental results show that the detection algorithm in this paper can effectively improve the observability of the image. Through the trained support vector machine, it can effectively identify the open joint, crack, and deformation defects. The recognition rate of pipeline image defects reached 96.3%. Highlights: The preliminary screening algorithm for PE gas pipeline defect images is proposed. PE pipeline image filtering and denoising are realized by dual filtering method. The edges of image defects are effectively extracted by the improved Sobel operator. The image defects in the PE gas pipeline are accurately classified and identified by the algorithm proposed in this paper.