Detection Algorithms On The Surface Edge Defects Of Galvanized Sheet Based On Digital Image

More and more steel enterprises realize that the effect of recognizing surface defect of the galvanized sheets on improving the quality of product and increasing economic output with aggravating competition in the market currently. The quality of the surface of the galvanized sheet directly related to the quality and performance of the product. It is low that the traditional detecting accuracy operator used in the surface defects of galvanized sheet. The edge detection algorithm under the digital image of the surface defect of the galvanized sheet was studied in this paper. Two hybrid edge detection algorithms were put on the basis of the analysis of the surface defects of the galvanized sheet, and the detailed content is as follows:1. The Sobel operator processed images is characterized by high speed, little calculating and smoothing image noise, etc. But the image contained the pattern on the surface of galvanized sheet, which made the image more complex and stronger noise. It would make the pseudo edge obviously if the Sobel operator directly. Therefore, in this paper the digital image of the surface defects of the galvanized sheet was deposed by a gradient sharpen combining with Sobel operator forming mixed edge examination. The grey value morphological technology was used to remove noise, followed by simulation experiment. The results showed that this algorithm contained good ability to remove the noise and edge extraction for the images of edge containing obvious defect( such as holes, shells and coating ripples).2. A mixed edge detection algorithm was brought on the basis of wavelet transform for images of obscure edge (such as plating scratch, surface impurity and edge stripes). By using wavelet transform, the outline and detail were clearer, more accurate orientation on edge of the images. Comparing with the traditional edge detection algorithm, the results showed that the image of the edge defect was more clearly and accurately, improving the defect recognition rate as well.