Research On Some Problems Of Fabric Defect Detection

"Clothing,eating,living,and traveling" are the four basic elements of human survival.Among them,"clothing" has been ranked first since ancient times.Therefore,in China's traditional industries,the textile industry has always been China's pillar industry and important national production industry,and the quality of textile products is also related to the lifeline of textile enterprises.Therefore,the monitoring and testing of textile quality is an indispensable part of the production process.In the production process of textile enterprises,the traditional artificial fabric defect detection method not only consumes a lot of manpower,but also has the disadvantages of low detection efficiency,low accuracy,high labor intensity and high cost.In recent years,with the rise and widespread application of computer vision and image processing technology,research on fabric defect detection has become one of the research hotspots in the field of image processing.In order to improve the efficiency of fabric defect detection,we use image processing technology to study several problems of fabric defect detection in this thesis.In view of the difficulty in detecting narrow spots of fabrics,we combines the sparse expression of images with image processing technology as the background,and uses FCM clustering method to detect.The basic idea is to divide the fabric detection model into a learning module and a real-time detection module.The innocent fabric sample library is input into the learning module for learning,and the sample input real-time detection module containing the defects is compared with the learned sample,and the threshold is adjusted to segment the missing image for inspection purposes.Aiming at the problem of fabric defect detection,this thesis proposes a non-uniform incremental LLE detection algorithm based on Gabor filter cluster.The Gabor wavelet has better resolution in the time domain and the frequency domain.This feature is used to effectively extract the local directional features of the image at multiple scales,and then the non-uniform incremental supervised LLE algorithm is applied to the extracted high-dimensional features.Dimensionality reduction,and then F-KNN algorithm is used forclassification,and finally the defect recognition result is obtained.Compared with the LLE algorithm and the supervised LLE algorithm,this method has a significant improvement in detection accuracy and reduces the detection time to a certain extent.In order to extract the intermediate breakpoints and the non-smooth problem for the image edge,we uses the cubic B-spline as the filter,and uses the wavelet transform method to improve the edge discontinuity problem from the global perspective,and then combines the wavelet transform method based on the modulus maxima.Maximal value suppression,finally using K-means clustering adaptive double threshold method to detect the edge of the image,this method improves the positioning accuracy of the edge.