Shell Body Flaw Detection Method Research Based on Wavelet Transform

摘要

The key of automatic recognition system is feature extracting, especially in product quality automatic lossless detecting system with image sensor device. Wavelet transform which has the characteristics of processing faintness signal is developed based on the Fourier Transform. This paper researches the shell body flaw product line quality detecting system which requires non-contact measurement, gives its characteristic information, displays in computer and automatic alarm functions. This system realizes the quantitative analysis for the physical features such as length and location by trailing and extracting, line draft joining and other methods to flaw image of post pretreatment. It is proved that this method using wavelet transform is better than traditional methods such as through smoothing, sharping and so on in edge extracting. The reason is that image noise and edge is high frequency signal and can’t be differed by frequency band methods. As regards shell body, flaw and edge signal is high frequency weight and noise is low frequency weight. Therefore, the author reduces shell noise, extracts edge and flaw using wavelet transform according to the above characteristic. In this paper, a kind of filtering algorithm based on domain correlation is prompted so that this method processes threshold coefficient after wavelet-domain filtering and get rid of remained noise coefficient. The result shows that the wavelet coefficient has better continuity, high accuracy and easy construct signal. Especially, the effect of reducing noise is excellent. Moreover, this method also has specialty of low compute complexity and high rate.