Research And Application Of Weld Position Recognition Of Longitudinal Welded Pipe Based On Machine Vision

As the application of straight seam welded pipe becomes more and more extensive,its weld,as the weakest position of the entire welded pipe,often needs to be inspected to meet the requirements of actual scenes.According to the requirement of controlling the welding seam position and observing whether the weld is qualified when the hydrostatic machine performs the hydrostatic pressure test on the longitudinally welded pipe,this paper proposes a scheme based on machine vision to judge the welding seam position of the longitudinally welded pipe.The weld position of the longitudinally welded pipe is a circumferential feature on the welded pipe,and the position information is three-dimensional information.This paper uses monocular vision to detect the deflection angle of the welding seam of the rotating longitudinal welded pipe from the target position,and controls the rotation stop time so that the welding seam finally stops at the target position.According to the corresponding coordinate systems under different viewing angles and the relationship between them,the mathematical relationship between the deflection angle of the weld and the corresponding features of the imaging plane is established to calculate the deflection angle.In order to obtain the position of the weld feature,two solutions are designed in this paper.(1)Directly extract the edge information of the welded seam and the welded pipe.Aiming at the problems that the traditional Canny algorithm is susceptible to noise interference and poor connection between strong and weak edges when identifying the weld edges of longitudinally welded pipes,an improved Canny algorithm is proposed to detect the weld position of longitudinally welded pipes.First,the local mean deviation in the gradient direction is introduced in the non-maximum suppression;secondly,the maximum between-class variance method is used to calculate the threshold for the points after the non-maximum suppression;finally,the strong edge boundary is determined by the strong and weak edge distribution characteristics of the welded pipe picture.The connection boundary is used as an additional connection condition.It is verified through experiments that the improved non-maximum value can effectively suppress noise points and retain edge information compared with traditional methods.Compared with the traditional edge detection algorithm,the improved Canny algorithm in the recognition of the weld edge position of the longitudinally welded pipe can reduce the interference of factors such as rust and scratches,and retain more welded pipe edges and weld edges.The peak signal-to-noise ratio and mean square error of the improved Canny algorithm on the commonly used BSDS500 data set are better than other edge algorithms in88.79% of the pictures.Finally,the Hough line feature recognition is used to detect the line and bring into the welding seam position mathematical model to obtain the final welding seam position information.(2)Make use of the certainty of the welding seam position of the longitudinally welded pipe during welding,and spray color calibration at the 90° circumferential position of the welding seam position.Perform color segmentation on the image,and use the center line of the smallest circumscribed rectangle of the color connected area as the color calibration position.The image is binarized using the maximum between-class variance method,and the vertical edge detector is used to obtain the edge of the welded pipe,and the mathematical model of the weld position is brought into the color calibration position,and the weld position is obtained after a delay of 90°.It is experimentally verified that the color calibration feature through image processing is more stable than the weld feature detection result of the longitudinally welded pipe.The hardware of the detection system was analyzed and selected,and the advantages and disadvantages of various light source combinations for metal cylindrical lighting were analyzed,and the hardware selection of the system was completed.The human-computer interaction window of the detection program is designed,through which the parameter input,the monitoring of the running state,and the start and stop control of the system can be carried out.According to the hardware conditions of the inspection site,the welding seam position verification experiment platform is designed to simulate the entire identification control process.Connect the end of the longitudinally welded pipe with an angle sensor,and use the detection result of the angle sensor as the control result of the entire system.The experimental results show that the angular deviations obtained by the two schemes under static measurement are both less than 1.5°,and the dynamic measurement and control results meet the design requirements.In this paper,by establishing the relationship between the image and the detection target,the algorithm used in the image detection is improved,the hardware selection and the program interface design are completed for the detection requirements,and the monocular vision is used to determine the position of the weld of the longitudinally welded pipe.The identification of the welding seam meets the requirements of the hydrostatic pressure test for the position of the weld.