The Research On Detection Algorithm Of Fabric Defects

We all know that during production of the fabrics, there must be defects, which directly affect the quality of the fabrics and then they affect the sales and export of the products. Therefore the perching link is an important process in controlling the quality of the fabrics. So far, the traditional perching work proceeds with simple hand tools. Since the problems may occurred by the subjective influence of the personnel, like error, omission and so on, the efficiency is very low. With the development of the computer imaging processing method and the industrial automation, the automation of the fabrics production has become an inevitable trend. Based on machine vision, the automatic perching system has become the focus of attention.Currently, foreign fabrics testing has already applied on hardwares, but the price is very expensive. Finding an automatic detection algorithm that based on PC platform can detect the defect on the fabrics surface effectively, rapidly, accurately and with low cost.The cord fabric image defects detection algorithm is mainly composed of four parts-- image preprocessing, comparative analysis, image segmentation and feature exaction. The first step begins with image preprocessing. After the histogram equalization of the fabric image and comparative analysis of some smooth sharpen algorithms, we take the median filter to smooth and noise-removing the images. And with the transformation operator--Top-Hat, we can sharpen the fabric images. Then through comparative analysis, we take the gray histogram arithmetic to quickly determine whether the fabric images exist defects. For the image segmentation, it presents a method based on mathematical morphology, which utilizes the autocorrelation function and FFT to identify the repeat units of the fabric longitude and latitude structure. Then mathematical morphological erosion and dilation methods are used to detect the defects information in terms of those repeat units .After using the traditional method of morphological operations, he author uses morphological opening again in order to remove noise and highlight the defects. Finally through the five feature constants of the defects-- the length(L), the width (W), the transit level elongation (R), the area(S) and the compactness(C) to draw the characteristic of the fabric images. Experimental results show the algorithm that this paper presented is validated and effective.