Study On Nonwoven Fabric Defects Inspection

The defects detection is an important factor in nonwoven fabric manufacturing. The traditional methods depend mainly on human version. With the development of IT technology and the decrease of the cost of computer hardwares, it is important to develop new techniques, low-cost, high-speed and accurate test results detection devises. It is meaningful to promote the quality of nonwoven fabric.The results of the inspection equipment based machine vision are fast and objective and will not be influenced by the environment such as temperature or humidity. The subject is to make research on the feasibility and usability of nonwoven defects inspection based on first-order statistics on histogram of grayscale image, thresholding methods for image segmentation and 2-D Gabor filters. Static algorithm research is launched through matlab. C++ is used to release the interactive nonwoven fabric defects detection system to explore the statement of the algorithm usage. The main contents and the results are listed as followed:1.The method of first-order statistics on histogram of grayscale image is to compare the parameters of fabric with the ones of normal nonwoven. Because of the elementary operation,the speed of the method is high which can reach the standard of real time inspection. The detection rates of variance, distortion, kurtoses and entropy are between 60-70%. It is not a self-adaptation inspection method because it must change the ranges of the parameters by huge amounts statistics if surrounding or the fabric changes.2.In thresholding methods for image segmentation, the defects are inspected through preprocessing, threshold segmenting and extracting characteristic values. The detection rates on holes, oil and most of the impurities are higher than 90%.But it failed to detect the invisible defects which mix in the fabric. 3.The method based on 2-D Gabor filters is researched to detect the invisible deficits. After being proceeded by the Gabor filters of three scales and four directions before the threshold segmentation, the effectiveness is improved. Not only can it detect the invisible defects, but also its detection rates on other defects is higher than 95%. Its speed is low, which can not reach the real time detection standard.4.The three methods mentioned above are compiled by C++ within the nonwoven fabric defects inspection system. The user can choose the methods according to the actual through the interactive user interface. The test initially prove the detection ability of the system.