Study On Recogniziation Technology Of Weld Pool Image Character Based On Fourier Transform

Real-time weld arc control is important to welding automatization and to improve joint quality. Vision sensor is one of the most important sensors available in real-time weld control, the key to vision sensor method lies in the dynamic valid identification of weld pool character information from welding visual image captured in welding. The geometrical parameters of the weld pool reflect the current welding state, include the depth of the weld pool and the figuration of the seam, etc. The warps between ideal welding state and the current welding state can be obtained by the geometrical parameters of the weld pool, then rectify the control system to guarantee the weld quality. Methods emphasized in extracting weld pool character technology are filtering, image enhancement, image segmentation and edge location as well. This experiment will try transforming the weld image from spatial domain into frequency domain through Fourier transform, and process the weld image with some frequency methods to enhance the image and give prominence to the weld pool characters.This article analyzed the image processing methods in spatial domain, according to the weld pool image process flow - image filtering, image enhancement, image segmentation and edge location. The comparative experiment shew a project of weld pool image character extracting. Average filter, median filter and Wiener filter were used to get rid of the noises, and the experiment also tried combining every two of the three filters. Image segmentation based on region and image segmentation based on edge were carried out in image segmentation. At last the experiment shew that the project that including combined filter, image segmentation based on region, histogram equalization and Canny edge detection, had the best result in weld image character extraction.This article also studied the Fourier transform and the filters in frequency domain, and analyzed the understanding of image in frequency domain - the relationship between spatial domain and frequency domain, in detail. Butterworth high-pass filter was carried out in frequency domain to enhance the weld pool image. According to the problem that the weld pool image became dim after the Butterworth high-pass filtering, the Butterworth high-pass filter was impoved in two ways. The improved Butterworth high-pass filter having the best effect was finally found, by changing the parameters of the filters. Then, the extraction of the weld pool image characters was finished by the frequency domain method combined with spatial domain methods. Compared with the spatial domain only method, the project of spatial domain combined with frequency domain method is better, it is good to make the next process of computing the parameters of the weld pool.