Research On Image Detection Algorithm Of Fillet Weld Based On Visual Sensor Using Three-line Combined Structured Light

Laser vision sensor is widely used in welding seam tracking,which has the advantages of high precision,non-contact and fast measurement.At present,most of the vision sensors can only be used for welding seam tracking in one plane.In this paper,a visual sensor using three-line combined structured light is designed.On the basis of the double-line laser tracking welding seam,a horizontal laser is added to detect the fillet of the turn ahead,which provides the required information for the robot to track the fillet weld seam.This paper first introduces the composition of the robot system,the control system is based on PC104 bus development,using image acquisition,motion control,data acquisition and other board cards to achieve control.A three-line structured light monocular vision sensor is designed to collect image information for measurement,including the selection and installation of camera and laser,the shape and size of sensor design.The software part added Open CV computer visual library,using MFC design interface to display the collected images and control the robot,to achieve human-computer interaction.Secondly,the conversion of each coordinate system in image processing is introduced.The camera in the sensor is calibrated,and the internal and external parameters and distortion parameters are obtained,and the lens distortion is corrected.The calibration method based on the three-point perspective model is used to calibrate the laser plane equation to ensure the precision of the calibration,and the calibration parameters can be directly used in the subsequent image processing.Finally,the image is processed.The threshold binarization of the image is carried out to segment the laser line and the background.Gaussian filter smoothing laser fringe;The laser center line is extracted by thinning algorithm;The Hough transform detection line algorithm is used to fit the line.By adding the line discrimination algorithm,the laser lines are discriminated and the linear equations are obtained.The coordinates of the feature points are obtained by simultaneous linear equations to realize the recognition of the feature points.The coordinate of feature points is transformed to the base coordinate system,and the Angle is calculated by the law of cosines.Through image processing and detection algorithm,to test the different angles and distance,the detection error within 3%,and proves the applicability of the sensor and image processing algorithms.