The Research On Fabric Defects Of Coarse Positioning And Detection

Fabric defect detectio n plays a crucial role in the quality control of manufacturing in the textile industry.With the rapid development of computer vision and image processing technology,the research on fabric defects has been an important focus in the field of image processing in recent years.The projection method is a method of the localization for fabric defects by determining the anomaly range of the corresponding projection graph,while the projection graph is obtained by projecting the fabric image along a certain direction.Gabor filters have the selectivity of directions and frequencies,and have the excellent locality in the spatial and frequency domains,then are widely applied to the fabric defect detection.This paper applies the projection method to localize fabric defects,at the same time,proposes an algorithm of the fabric defect detection based on optimal Gabor filters:(1)For fabric defects which approximatively appear in an oblique direction,based on the projection method with the horizontal and vertical directions,we propose an oblique projection method to localize fabric defects.Firstly,we obtain the principal diagonal and anti-diagonal projection graphs by projecting along the corresponding directions of the defect-free and defect images.Secondly,we determine the normal ranges of the features,such as the distances between maximum and maximum values,minimum and minimum values,maximum and minimum values,minimum and maximum values,according to these features of the corresponding projection graphs of the defect-free image.Lastly,we judge the abnormal ranges of these features of the defect image through the ir normal ranges,then localize the defects.The experimental results show our method can validly localize the fabric defects which appear in an oblique direction.In addition,for fabric images that defects appear in the horizontal and vertical directions,based on the projection method,we apply different features to localize fabric defects.The results further indicate our four features are more reasonable,and the results which apply our four features to localize the fabric defects are superior to the ones which apply peak values,valley values,the distances between maximum and maximum values,minimum and minimum values.(2)For the commom fabric defects,we propose an algorithm for detecting fabric defects based on the optimal Gabor filters.On one hand,we choose two optimal Gabor filters : we determine 2 optimal directions by the information of the images that the defect image is convoluted by applying 4 direction masks with 5 ? 5,then choose respectively 1 optimal scale in 8 Gabor filters along 2 optimal directions by applying the Signal to Noise Ratio and cost function,thus we determine 2 optimal Gabor filters.On the other hand,on the basis of 2 optimal Gabor filters,we propose an algorithm of detecting fabric defects.The experimental results show the algorithm can effectively detect fabric defects.For defect images,we combine the projection method with algorithm for detecting fabric defects based on optimal Gabor filters to detect fabric defects,this method not only effectively detects fabric defects,but reduces the computational complexity in all the steps of the detection process.