Three-Dimensional Measurement For Weld Based On Reliability Evaluation And Structured-Light Vision

With the continuous development of manufacturing technology,three-dimensional measurement technology has become an importantresearch direction in the entire manufacturing process, which occupies akey position in industrial inspection, reverse engineering, and other fields.Welding as the final stage in the manufacturing process, the weldparameters will directly affect the final quality of the product. Therefore,the three-dimensional measurement of the weld is one of the importantdirections of today’s welding intelligent automation development.Using line structured-light active vision technology can clearlycapture the original image information of the weld surface, followed by adigital image processing, which can obtain the parameter values of theweld. However, the different reflective properties of the weld surface andthe complex contour structure will result in uneven thickness ofstructured-light stripe, partial loss of information, gray change dramatically,more noise and other issues which will affect the calculation accuracy ofthree-dimensional measurement. Therefore the design of a suitable opticalimaging system, and using the effective image acquisition and processingalgorithm is critical.This thesis is based on the technical difficulties, to design a set ofoptical imaging system with line laser and CCD camera to capture the distinctive characteristics and clear stripes of welds original image, andcalibrate the entire imaging system; to analysis imaging principle andmathematical model of line structured-light and the energy transfer processwhen CCD camera capturing images; to introduce the concept of thereliability value of the light bar, in order to evaluate the exposure quality ofvarious locations throughout the image, and analyze the relationshipbetween the reliability value of light bar and the image gradation; based onthis reliability evaluation mechanism, to design an adaptive multipleexposure method, after each exposure to retain the position with higherexposure quality, i.e. the higher reliability value, and set the next exposuretime, after this multiple exposure method, stitching to obtain thestructured-light image with good exposure quality; in addition, to design aset of digital image processing algorithms for the inspection object in thethesis, fillet weld, with dynamic median filtering, binarization processing,morphological repair and other methods, accurate extraction of a contourof the weld surface, and to retrieve the feature point of the weld, tocalculate the relevant weld parameters; Finally, through the experimentalresults to analyze the influence of adaptive multiple exposure methods andits internal parameter on the detection speed and accuracy of the entiresystem.