Investigation On Computer Vision Inspection Of Laser Welding Of Inlet Tube Based On Digital Image Processing Technology

There are surface welding defects such as incomplete welding,surface hollow,surface heave and welding spatter in the inlet tube of laser welding in the Auto-mobile oil injection nozzle production line. The traditional eddy current testing results are not so ideal, the misjudging rate is relatively high. In order to avoid misjudging, workers will have a second time check, but such kind of operation can neither ensure the product quality nor decrease the production efficiency.In this study, we acquire welding image by using the existing rotation function,line CCD sensor and lamp-house. Image preprocessing including wiping off the background in the welding,image denoise and image enhancement. After Image preprocessing, we use the Improved Otsu arithmetic to realise image segmentation. Finally, we judge the welding quality through statistics and comparison. The median filtering and image enhancement are two main functions in image preprocessing that are used to solve the problem of unstable impact which results from line CCD sensor; After comparing the results of regional image segmentation algorithm (traditional Otsu and improved Otsu Methods) and marginal image segmentation algorithm, we find that marginal image segmentation algorithm does not apply to the image with complicated margin and uneven lighting. The objectives and background cross-distributed in the image, improved Otsu applied this situation successfully. It is one of the practical and effective image thresholding segmentation methods.Through this on line image processing software, we will identify product quality and finish the existing production line information exchanges. Through improving judging efficiency and meeting existing production rhythms, we control the time of image acquisition and image processing within 1.5 second / piece.