Research On Vision-based Decoupling Detection Algorithm Of Three Dimensional Welding Seam

As the modern welding production requirements for product quality and efficiency is higherand higher, how to guarantee the welding quality and improve the level of welding automationbecome an important topic for human beings. Laser structure light active visual inspectionsystem has played a positive role in improving the automation and intelligence of welding robot.At present, manufacturing at home and abroad is facing precious opportunity of robotslarge-scale application so the demand of visual sensing system which improves the robotflexibility and intelligence is becoming more and more urgent. So studying the visual detectionalgorithm and developing visual inspection system with independent intellectual property rightsare of great significance for improving the robot’s awareness.The visual inspection system prototype in which pipeline docking V groove weld is seen asthe main research object has been designed for algorithm testing. Intelligent cameras based onTexas Instrument C64series DSP and the laser light source with line power of100mw, thewavelength of650nm and appropriate focal length of lens with filter are chosen for the system.The hardware structure of prototype has been designed according to space constraints and visualmeasurement principle of visual inspection system.The structure light processing is extracted based on a least interest in real-time weld seamimage area (ROI), an adaptive image segmentation, the stripe extraction of center line, thedifferential and filtering of the center line, Hough transform. Weld feature points are extractedwith the use of the method based on distance search to obtain the point coordinates accurately onthe image plane which reflect the characteristics of the weld shape and location.Decoupling detection algorithm is put forward for circumferential weld, that is to saythree-dimensional weld characteristic information extracted by the time sequence adjacent frameis decoupled in order to calculate the torch welding deviation value in the two-dimensional planerelative to the center of the weld to achieve the accurate and rapid welding torch motion controlbased on image. Moreover, the control method based on image has been designed to achieve fastaccurate localization of welding torchA large number of field test results show that the visual inspection system set up in thispaper is compact and reliable, because image processing algorithm designed in the system has abetter stability and real-time performance, high reliability and anti-interference performance sothat weld recognition and real-time control problem can be solved in the environment withstrong interference during the welding process.Our visual inspection system can meet therequirement of industrial production and has important scientific and use value...