Car Wheel Internal Defects Automatic Detection Technology

X-ray detection is the main approach to detect quality of automobile hub. At present,most manufacturers use artificial mode to evaluate the quality of the product. It greatlyreduces the degree of automation of production and the evaluation results are unreliable. So Itis of great significance that realize the automation of car wheel hub defect detection toimprove production efficiency and reduce manufacturing cost, etc. This paper has designedthe car wheel defects automatic detection system that can meet the needs of current industrialdetection to solve the existing problems of testing mode and made the deep research work onthe following aspects:(1) Usually direct acquisition of X-ray image has low contrast, defect edge blur and noise,there is a big background fluctuation characteristics, this paper analyzes the reasons affectingthe quality of the image, the image acquisition when using multiframe superposition noisereduction greatly improve the SNR, good suppresses noise; At the same time introduced themedical images on a window wide gray dynamic range adjustment algorithm to lift the image,and use the linear inverse sharpening mask algorithm to enhance image edges and details;(2) X-ray imaging device can detect only a portion of each hub, by means of amechanical device must regularly rotating hub to ensure that all parts can be detected, thepresence of errors due to mechanical movement, leading to the hub during Teach ROI alsomarked deviation which caused the final defect detection Missing and misjudgment. Thepaper used the image registration on the ROI automatic tracking, which uses a wheel centerfeature-based image registration method of registration to good effect, and time-consuminglevel have dropped ms; spokes edge due to the structure of their own error again plus thickedge disturbances such image registration method based feature less effective, while the useof intensity-based normalized cross-correlation registration algorithm achieved good results,and on this basis, based on the use of one-dimensional projection improved algorithm, theregistration effect with the traditional NCC algorithm, but in some time-consuming aspect of the algorithm greatly improved, and more to meet the online real-time detection systemrequirements;(3) In full consideration of defect detection based on real-time and accuracy, through theactual test, using the dynamic threshold segmentation algorithm based on ROI for defectsegmentation, and corrosion, expansion algorithm of mathematical morphology for imagesegmentation processing, make the edges of the defect more clearly, continuously to ensurethe integrity of the defects, can effectively solve the wheel defects segmentation problem.