The Study On Evaluation Method Of Binocular Vision In Weld Appearance

Weld appearance quality is an important part of welding production qualityguarantee system for pressure vessel, cars, spiral welded pipe. Weld appearancequality is a main factor in production costs and production efficiency.Manufacturers’ requirements become more rigorous for weld appearance formingquality. The requirements of its testing and evaluation method to further improve.At present, artificial visual is main method to determine weld surface undercut,root concavity, blowhole, cracks and other defects. Due to the different test skillsand experience, the quality inspection standards have inevitably deviations, andtest results are unstable by the inspectors subjective factors, and the quantitativeaccuracy of the evaluation are difficult to guarantee.This evaluation method cannot meet the development needs of welding production automation and precisionto quantify. Binocular stereo vision technology is widely used in targetidentification and precise measurement, the path guidance and planning, real-timetracking and other industrial applications. In this article using the image a weldsurface forming obtain by binocular stereo vision, and to explore a visualmeasurement and evaluation method of forming a weld appearance. The maincontents include:(1) Built a set of computer binocular stereo vision system for the evaluationand detection of weld shape. The system’s hardware includes industrial digitalcamera (MV-3000UC), lens (VS-0814M), the white strip light source (VS-RL200)and its controller, computer, etc. Put the weldments on platform below thebinocular camera. Under the irradiation of active light sources to obtain visualimages of weld. The collected visual images transported to the computer throughthe USB. The visual image processing software can be realized the functions ofweld image display, storage, pretreatment, and the weld shape parameterextraction.(2) Established the relationship between the image pixel position and3-Dscene location. Obtained the conversion relationship between measured weld inwhich the physical coordinates and its imaging coordinate system. From cameracalibration, gotten the inside and outside parameters of two cameras and theprojection matrix. According to the projection matrix of the calibrated cameramodel, it become idealistic and greatly simplifies the later calculations. From Image correction, sharpening, smoothing, and other pretreatment, the outlinecharacteristics of the weld is more apparent. After split out of the weld area of theimage can be extracted the plane information such as length and width. In theextraction of height information of the weld surface, it carried stereo matching andsurface reconstruction for the weld image. The actual measurement proved that theweld width, reinforcement forming parameters have a higher accuracy by thecomputer binocular stereo vision technology, forming the basic parameters of thejudge as the weld.(3) Analysis of the main factors affecting of the weld formation, and the maindefects of weld appearance. According to evaluation method and evaluationsystem of the weld forming, established a quality evaluation system of the weldappearance base on computer binocular vision platform. From the three aspectssuch as weld width, reinforcement, and quality of deposition make the weldevaluation as matrix, parameterized.(4) Analysis of the relationship between welding parameters and the mainindex of weld, two prediction model of evaluation index is established base onSVM. It make welding current, welding voltage, welding speed as the input vector,and each of the average weld width and the average reinforcement as the outputvector. Analysis showed that the predicted results have a high correlation.