Research On Image Recognition Algorithm Of The Surface Cracks On The Billet

Surface defects of the steel billet is an important element to determine the quality of thesteel and its service life. But in the process of producing and detection of the steel at home andabroad currently, it is a common practice to identify the defects of the steel billet by means ofhuman and offline work, which is neither good for promoting the producting efficiency, nor forthe reliability of the quality of the steel, because of some artificial uncertain reasons. This papercarries out a basic research detection system of identifying steel billet surface defects wit digitalimage processing technology and pattern recognition technology, and provides an integral andeffective evaluation system of identifying steel billet surface defects.Firstly, this paper has made a series of image processing for the effects of fuzzy, noise andthe unbalanced distribution of the gray value, which is caused by the environment or the ways ofphotographing. Based on the analysis of the grey level histogram of the picture, this paper takesa subsection linear stretch for the picture of the steel billet, in order to compress the noise at bothends of the gray value area and stretch the useful details linearly, improving the comparisons ofthe useful details and decrease the visibility of the noise. Then, the image smoothing is carried onwith the eliminating noise and masking template method, which can not only smooth the noisebut also can keep the details as much as possible. Medial value filter is a kind of nonlinear filterwhich can come over the details fuzzy of the image caused by linear filters like LMS filter ormean filter under certain conditions, and is the most efficient filter to filter the impulseinterference and the scanning noise of the image. It is the best filter for the image of defects.Secondly, by comparing the results calculated with varied edge detection operators of thedefects image, the Sobel operator has a better result as the edge detection operator of the Surfacedefects image of the steel billet, which can get a complete and clear edge without furtherrefinement. The image erosion based on edge detection can finally gets the defects imagebinaryzation which is calculated with the horizontal structural element.Thirdly, this paper has proposed the means of boundary tracking to extract the contourfeatures of the objects and gets the length, width and areas of the defects in the image of the steelbillet. The introduction of the run length encoding theory realizes the location and realization ofthe defects, and builds the realization pattern of the steel billet surface defects image primarily.Finally, this paper has designed and developed a software system of the realization of thesteel billet surface defects image based on image recognition technology by Microsoft VisualC++6.0, which has4function modules of file management, image processing, feature extraction,realization and location of the defects.