| With the wide rang of applications and flourish of advanced technology in manufacturing industry, such as robot technology and digital manufacturing, quality assurance system and technology become the most important process to decide product life. In the manufacturing industries, such as aircrafts, vehicles, steamships and so on, weld surface quality also become the main part to decide the product competition. However, surface weld quality detection is mainly finished by manual work after weld now, and it can’t meet the automatic and efficient needs of industry production. So developing digital quality assurance system and technology become the need of automatic weld production. Because of the simple structure, high accuracy and efficiency, binocular stereo vision have been used into many fiels, such as target tracking and visual measurement. Aming at SAW joint, a non-contract, digitized and on-line judging method for weld forming was proposed based on binocular stereo vision, and the main research work include:(1) A specialized visual system for weld surface quality was set up based on computer. Based on weld characteristics and the special surrounding of image acquisition, the selection and construction of hardware system was finished. Weld images acquisition and processing software system was also developed, which can realize the acquisition, display, storage, print, invocation, pre-processing and data acquisition of weld images.(2) The calibration of visual detection system was finished, according to the obtained internal and external camera parameters, a mathematical model which can calculate the depth of object based on disparity was set up. Weld images were rectified using epipolar rectification. And weld segmentation was finished by using a series of processing methods, such as contrast enhancement, grey level normalization, gauss filtering, dynamic threshold segmentation, mathematical morphological processing and so on, the acquired weld images can reflect the weld contour exactly.(3) Stereo matching algorithms of weld forming was set up. By analysing the characteristics of weld and the influcing facors of stereo matching, stereo matching algorithms for weld shape matching were set up based on region and characteristic. By the contrast of mismatching results of the two methods, matching method based on region was more suitable for the weld stereo matching. On the basis of matching result,3d-information of weld shape was calculated and3d-reconstruction of weld shape was finished. The reconstruction result showed that the acquired information could reflect the changes of weld shape exactly.(4) Visual judgement indexes of weld shape were established. Through the analysis of geometrical characteristics, gray level distribution, location of weld, aiming at the shape changes of weld width and hight,7weld forming judgement indexes were established, such as weld edge deviation degree A e, change rate of weld width Kw, center line flatness Kc, weld hight deviation degree δh, change rate of weld hight Kt, filling degree Kδ, kurtosis n. Measurement results showed that the established indexes could judge the weld shape effectively.(5) A intelligent pre-judging algorithm for the weld shape of weld width and hight was explored. For the macroscopical changes of weld shape in width and hight. Taking weld edge deviation degree A e, change rate of weld width Kw, center line flatness Kc as input and forming quality of weld width as output, pre-judging GABP algorithm for the anomaly of weld width was established. Meanwhile, taking weld hight deviation degree δh, change rate of weld hight Kt, filling degree Kδ, kurtosis n as input and forming quality of weld hight as output, pre-judging GABP algorithm for the anomaly of weld hight was established. Actual testing results showed that the linear dependent coefficients between judged results and measured results were0.95and1.00, the established pre-judging GABP algorithms could be used to judging the changes of weld shape in weld width and hight efficiently. |
PDF Full Text Request