| Because the mistaken-estimate, forgotten-estimate and low efficiency influence the output of product and productive efficiency when X-ray is used in the security detection of products, this paper indicates that the software of artificial intelligent recognition, generally used by flaw-image, is advanced to solve these problems referred when products are tested.This paper focused on the pretreatment of images firstly, and then recognized them. The main contents of the pretreatment part include the following parts: the appropriate reinforce of wavelet transform was used to solve the low contrast;the median filter and the de-noising of outline extraction were adopted to solve the noise of pictures;Aiming to these blur pictures, sharpening of gradient with threshold was used, based on analyzing reasons, all these methods referred before got the perfect effect. During the segmentation, iterative threshold was used according to the features of pre-dealing images, while during the marginal extraction of images, profile method was used according to the features of images after segmentation and de-noising, the ideal effect was got during both phases. The recognizing part of pictures mainly studied how to use fuzzy mathematics to categorize the different vectors effectively, and how to use the ability of genetic algorithm to look for the best way to make the image matching come true, the ability of BP neural network to deal with high-speed and uncertain information was got according to the non-linear but largely paralleled feature of BP neural network. The key is that outline extraction and genetic algorithm should be adopted in BP neural network to recognize images comprehensively using the combination of advantages of three methods.This study indicates that instead of judging pictures by man, judging pictures on-line is scientific, normative and intelligent; meanwhile the phenomenon of mistake-estimate and forgotten-estimate could be decreased effectively. |
PDF Full Text Request