Research On Laser Scanning Vision Measurement System For Welding Seams

Due to the thermal deformation of the workpiece and the error of the fixture,the traditional teaching and reproduction welding robot cannot meet the requirements of modern high-precision automatic welding,while the seam recognition and tracking functions are firmly bound with the main bodies of the robots,which are extremely expensive on the market.Therefore,in this paper,a visual measurement sensor based on linear structured light will be studied and designed for practical requirements.It can accurately and rapidly recognize and locate the weld seam by acquiring images of the weld seam in real-time so that the robot’s welding gun can be guided to do the welding job.The sensor is characterized by small size,simple use and perfect function,but with no compromise in accuracy and real-time at meantime.With a bit of simple setup,a traditional welding robot can be upgraded to one with automatic welding function by just installing it in front of the welding gun.The system studied in this paper used a linear structured light and a vision sensor to recognize and measure the weld seam.It is mainly based on laser triangulation.By analyzing its principle,hardware devices were chosen to build the main structure of the measurement system.Then we derived the relative formulas between the intrinsic parameters and the unknown variables and calibrated the parameters using a range of methods,and also studied the welding seam image processing method and recognition algorithm.Finally,an application software is programmed to work with this sensor,which performs a series of operations such as laser weld image acquisition,image processing,weld feature point extraction,coordinate transformation,etc.Finally,with the collected information it can guide the welding gun to move forward along the seam while further scanning the working area,so as to realize automatic welding.This paper studied the laser triangulation method,weld feature point extraction algorithm,sensor calibration method,etc.,and described and summarized the experiences and results of the system in detail.After a series of experimental tests,the system can be working at an average rate of 15 FPS under the Ubuntu system running on the virtual machine,the average relative error of the measurement is about 4%,and the average absolute error in the experimental range is about 0.54 mm.The welding feature point extraction algorithm and calibration method of the system was proved to be feasible by the experiment results.