Study Of The Steel Pipe Welded Defects Detection Based On The Computer Vision Image

The steel pipe is one of the main equipment of the transportation of the oil industry, it's welded defects seriously influence oil and gas transportation security. For a long time, the steel pipe welded defects detection completely depended on the manual work. Questions of the detection are the poor consistency of the quality and the low efficiency of the detection. Therefore, the need to improve the detection means is urgent, in order to ease the contradiction between the quick processing of the steel pipe welded seam and the low speed detection. The paper researches the algorithms of the digital image processing of the steel pipe welded seam defect detection based on the computer vision technology. The algorithm achieves the purpose of the recognition of the steel pipe welded seam defect.(1) Because the steel pipe welded seam image acquisition is in the process of the movement of the detection of the welded seam, the movement process exists the unavoidable problems of the motion blurred image. The study makes the rotation of the image on the base of having detected the fuzzy direction of the image, and then does the differential of the image, then calculates the fuzzy measure of the image utilizing the autocorrelation of the image signal, finally establishes the diffusion function using the fuzzy direction and the fuzzy scale.(2) The paper does experimental research on the steel pipe welded defects' type, analyses the gray feature of the type of defect. Considering that the contrast of the defects and background is little, it puts forward the evaluation function of image enhancement. It brings forward the local image enhancement algorithm based on region variance increased using the small pixel gray variance and the differences between the defects and the background entropy. The algorithm can effectively enhance the contrast ratio of the defects and the background.(3) In the foundation of in-depth study of the gray scale characteristic, it proposes defect recognition algorithm aiming at three defects of pores, inclusions and cracks. The algorithm can effectively segment the image defects, pores, inclusions and cracks whose area of not less than 0.3 mm2 can be identified automatically, and can meet the practical testing requirements as well.(4) The paper uses the MATLAB software to establish a series of the algorithms procedures of the image processing, the image analysis and the extraction of the image feature, furthermore, they have been made the experimental verification.The task makes the exploration and the attempt in the steel pipe welded seam detection using the computer vision technology, in addition, lays the foundation for the developing of the expert testing system in the future.