Magneto-optical Image Restoration Algorithm For Detecting Micro-gap Weld

With the development of laser technology, the development trend of modern welding technology has become intelligent automation, advanced welding process was undoubtedly the most widely used laser welding. High-power laser welding can be used for a variety of metallic and non-metallic welding, which is characterized by a high concentration of energy may be guaranteed of welding, faster, smaller aspect ratio and the weld heat affected zone. When welding thick plate material with high-power laser welding can be welded once, but do not fill material and beveling, therefore saving a lot of work time and input costs.In the laser welding process, to achieve an accurate seam tracking is to ensure the quality of an important part of laser welding, laser beam must be strictly controlled and always focused on weld center. Since the laser welding process encountered close butt, no groove, the weld gap is less than, the conventional method is more difficult to realize micro 0.1mm gap weld automatic detection and tracking. An approach of detecting the weld joint based on magneto-optical (MO) imaging is investigated.Principle of magneto-optical imaging detector is a combination of Faraday and Faraday effect, with excitation magnetic field induction welding the weld, the magnetic field component of the detection area is changed and the angle of rotation of polarized light to produce the image changes accordingly. In actural welding, however, even when various welding control parameters are kept constatnt, the welding process still shows instability as affected by different interuptive factors. For example, variation of magnetic field intensity and the interference of noises may result in image degrdation, poor image clarity and low indentificability, which make it hard to conduct in-depth analysis of image features. Therefore, weld seam surrounding should be detected by using magneto-optical sensor. By woking out the algorithm of image restoration, the degraded images can be recovered, based on which the accurate real-time detection of welding process status can be realized.This paper sets the laser welding process of low-carbon steel as the research object and investigates restoration algorithm of welding seam magneto-optical images. The magneto-optical sensor, by interacting with the magnetic field, creates magneto-optical images. The measurement value of weld seam position is obtained by analyzing the features of its magneto-optical images. Recovering processing is performed on the degraded images, and the image quality is much improved by integrated application of constraints least squares filter and blind deconvolution. The comparison between the true value of weld seam deviation, its measured value and the deviation value after image restoration shows that the weld seam deviation value after image restoration is closer to the true value. Therefore, using magneto-optical sensor to collect weld seam images and restoring the images help to realize more accurate real-time detection of welding process status.A total variation image restoration model is established based on energy function and visual features. According to the variational principle, the overall existence of the model can be proved by the comparison principle of partial differential equation. It then turns to the optimization of constraint and sets up functional. Euler-Lagrange formula of the model is constructed based on functional calculus of variation and the variational principle. Experimental image restoration is performed the discrete version of value approximate solution. Experimental data shows that the total variation image restoration model can not only reduce noise interference and help retrain desirable verge and textural features, but also effectively improve magneto-image quality. Experimental result indicates that weld seam can be more accurately positioned after total variation image restoration.