Research On Image Acquisition And Processing Of Molten Pool In CO₂ Gas-shielded Welding Process

With the appearance of industrial robot and its application in welding field,welding automation has become a developing trend.Most of the welding robots used at present are demonstration and reproduction type,and are based on artificial welding.Compared with manual welding,robot welding lacks flexibility and can not be found and corrected in time for the problems existing in the welding process.With the development of computer technology and image processing technology,it is combined with welding robot to collect and process the weld pool image of welding process,obtain the relevant characteristic information of weld pool image,and establish the mapping relation with weld depth and width,etc.In order to realize the control of welding process and adjust and realize the intelligence and flexibility of welding robot.Edge extraction of weld pool image is a key step to obtain weld pool feature information.In this paper,the passive vision sensing technology is used to collect and process the weld pool image of CO₂gas-shielded welding with robot welding,so as to obtain the complete weld pool edge and lay the foundation for the future research.A suitable image acquisition system was established to collect the images of the welding process.Aiming at the abnormal trigger shooting due to the influence of welding voltage,the external trigger circuit of the camera is designed to enable the camera to trigger shooting normally,so as to realize the clear pool image of CO₂gas shielded welding.For the collected molten pool image processing,the traditional four filtering algorithms are used to Denoise the image.After comparative analysis,the average filtering method is selected.In order to improve the overall brightness of the image and display the details of the molten pool,the image is enhanced,and the results of the six traditional enhancement algorithms are compared and analyzed,and the results are not ideal.And it is not suitable for most of the molten pool images collected.Therefore,this paper chooses to process the original molten pool image on the basis of color image,and carries on the average filtering method to the component graph,and carries on the algebra operation to the component graph after denoising.When the red component graph is multiplied by the difference between the green component graph and the blue component graph,the processing image with enhanced effect can be obtained.The results of traditional gradient edge detection algorithm and second-order differential edge detection algorithm are compared and analyzed,and the complete weld pool edge can not be obtained and there are more pseudo-edges.Therefore,in this paper,the mathematical morphological edge detection algorithm is selected,which needs to binary the image after denoising and enhancement,and the iterative method and the maximum inter-class variance method are used,and the effect is the same.In this paper,the maximum inter-class variance method with less running time is selected to binarize the image.In this paper,the multi-structure element morphological edge detection algorithm is used to extract the edge of the processed binary image.The specific process is to open the target image first,on the basis of which the closed operation is carried out to obtain the new image.After subtracting the image after corrosion operation on the basis of the new image,the edge of the molten pool image is obtained,and the edge of the molten pool image is complete and smooth,which is basically consistent with the original image.