Research On Weld Characteristic Information Extraction Method Of Line Structured Light

At present,with the rapid development of science and technology and productivity,welding technology as an indispensable part of production is more widely used in all walks of life.The detection of weld feature information is an indispensable part of the welding process,but the traditional method generally uses tools such as stoppers for manual measurement,which has low measurement accuracy and large error.The detection technology based on image processing provides a new way for non-contact measurement of weld.However,the actual welding process is very complex,and the collected images contain a large number of complex noise,such as high light reflection,splash slag and laser scattering.Therefore,how to accurately extract the weld feature information under complex conditions is the key to the realization of automatic welding technology.In this paper,the feature extraction method of line structured light weld under complex noise is studied by combining visual sensors,line structured light emitters as hardware and digital image processing algorithms as software.The main work is as follows :Firstly,according to the common welding scene,a vision sensor system based on linear structured light is designed.By summarizing the advantages and disadvantages of various methods of seam feature extraction at home and abroad,the basic scheme of this paper is selected,which can be stable,convenient and fast to collect images of welding seam in the welding process.Secondly,in view of the large amount of scattering and splashing welding slag generated in the welding process,an Adaptive top-hat Transform based line structured light welding image denoising method was proposed.The CCIH(cross-correlation Coefficient of Image Histogram)was used to select the optimal structural element size L.At the same time,an evaluation Index RSB(Ratio of Structural Similarization Index to Average Brivity)based on OTSU improved SSIM algorithm is proposed to select the optimal number of iterations I.The experimental results show that the proposed method is better than the traditional method in both subjective visual effects and objective indicators,and the noise is effectively removed while the linear structured light region in the image is well preserved.Then,a "limited contrast" adaptive histogram equalization enhancement algorithm(CLAHE)was proposed.The difference between the amplified line structured light bars and the background was beneficial to the segmentation and extraction of the line structured light bars,and was compared with the traditional enhancement methods.The results show that the binary image of weld seam with linear structure light obtained by this method contains more effective pixels.Secondly,a multi-direction dual tree complex wavelet transform(M-DTCWT)edge detection method is proposed to obtain single pixel edge points.Finally,the least square line fitting method was used to get the center line,and the center line and the boundary line were combined to get the final characteristic points of the weld.