Research On Double Arc Welding Molten Pool Image Processing And Image Fusion

In recent years, with the rapid development of computer technology and visualsensing technology, extracting weld pool image using the CCD sensor, achieving theinformation of weld pool edge through the digital image processing and then realizingthe procession control has become a vital research area in modern weldingautomation.Through the analysis of the typical DE-GMAW principle system and parentmetal heat input, this paper designed the double arc welding system which is suitablefor the AZ31B magnesium alloy welding. According to the arc spectrum distributionof the AZ31B magnesium alloy, this paper selected the characteristic spectral linewavelength of magnesium as the filter window. According to the arc interferenceproblems in the process of magnesium alloy double arc welding, this paper designedmultiple sets of composite filter system which is combined by the differenttransmittance neutral filters with magnesium alloy filter. Through contrasting a lot ofextraction experiments of magnesium alloy weld pool image, the suitable compositefilter system to magnesium alloy double arc welding has been determined. Using thepassive sensing, according to the pool’s radiation and the reflected light to the moltenpool at the center of the arc, the weld pool image of magnesium alloy was acquired.This paper mainly introduced the influence of median filtering window to imagefiltering effect, the difference between gradient sharpening with Laplasse sharp andthe characteristic of maximum variance threshold segmentation algorithm andinteractive threshold segmentation algorithm. The right weld pool image processingalgorithms of magnesium alloy were determined by experiment.According to the imaging theory of image focusing, the necessity of the weldpool image fusion was expounded. This article introduced some typical image fusionalgorithm, such as the pixel gray value take large or small, weighted average andregional energy algorithm, and proved the effect of image fusion through thisalgorithm. The theory of wavelet transform and wavelet transform multi-resolutioncharacteristics were elaborated. According to the respective characteristics of wavelet’s low frequency sub band and high frequency sub band, a new fusion rule ofweld pool image was proposed. Through the comparison of experimental results andimage fusion quality parameters, the superiority of the proposed algorithm wasdemonstrated and the information amount of weld pool edge was improved.This text introduced the basic algorithms of morphological image processing andoptimized the information of weld pool image through mathematical morphologyalgorithm. Using the weld pool image edge extraction algorithm, the accurate poolprofile was obtained. Through calculating the size of weld pool, this paper finally getthe character information of AZ31B magnesium alloy weld pool’s maximum width、half length、area and bank angle.