Research On Image Processing Algorithms Of Moving Droplets In Magnesium Alloy Welding

The application of machine vision and image processing technology to melt drop behavior study has important significance to update welding control methods and to improve welding quality.Taking double arc welding process of magnesium alloy as the research object, according to the characteristics of the spectrum, this paper designed a set of the composite filter based on passive vision system, and combining with CCD camera, PC, image acquisition card, the magnesium alloy double arc welding system and filter to reduce the light system this paper set up the image acquisition system, Using the equipment the original drop image was collected finally.According to the characteristics of motion blurred image, this paper puts forward a kind of image restoration algorithm based on frequency domain transform. After the image was transformed to the frequency domain, respectively using the Radon transform and differential autocorrelation, this paper estimated the fuzzy Angle and length of molten drop images, which results could be to to recover the drop image based on wiener filtering.Results for motion blur image restoration, firstly a median filter is used for data-processing, and then a histogram equalization method is used to do gradation conversion so as to get the suitable image intensity distribution. Then using the maximum variance threshold method to determine the threshold for image threshold segmentation to get binarized image. Finally using Canny edge detection operator to extract the clear drop edge profile by morphological method for its image fitting processing, The access to get outline of the droplet laid a solid foundation for the future study of the droplet size and improving welding process.In addition, in order to identify and track droplet motion in the welding video, a algorithm based on frame difference and Mean-shift was designed for its gray image and narrow background characteristics. Firstly frame differential method is used for difference processing of the welding video image of the first two frames, identifying the movement of the droplets from the background image and framing, these have solved the problem that the Mean-shift method needs to manually box the target in the starting frame. And then combined with the Mean-shift algorithm based on gray level histogram so as to realize the automatic identification and tracking for the moving droplets, all of these laid a foundation for the study of droplet transfer behavior and welding process control.