Research And Experimental Verification Of Defect Detection And Recognition Algorithm In Weaving Gray Fabric

After the weaving enterprises have finished weaving,they generally carry out defect detection on the fabric.At present,most domestic enterprises still use manual methods.Scholars in China and abroad have carried out a lot of research on automatic fabric detection systems,and have also launched some machine vision-based fabric detection prototypes and related commercial equipment,but the market is not very accepting of them.The main reasons are: there are many kinds of fabrics and defects,it is difficult to guarantee the versatility,at the same time,the detection efficiency is low,and the overall equipment cost is high.In addition,the current weaving and fabric detection processes are separated,which is not attractive in reducing the availability of auxiliary processes and information data.Based on the innovative ideas proposed in the subject patent(CN201910525250.X),this paper designs and studies three white fabric defect detection algorithms integrated on the loom to complete the detection and identification of fabric defects during the weaving process.The main contents include:Firstly,the design requirements of the defect detection system for weaving gray fabric are proposed,the overall scheme of the experimental verification platform is designed,the image acquisition module is designed,and the design and selection of the shooting scheme,precision,industrial camera,lens and light source are completed.Second,study the defect detection algorithm based on threshold segmentation.Firstly,median filter is used to suppress texture information,and then a contrast limited adaptive histogram equalization algorithm is designed to enhance defect features,which improves the shortcomings of global enhancement algorithm that cannot specifically enhance defect features when dealing with fabric image defect features with uneven illumination.In the threshold segmentation stage,the adaptive threshold method is designed to segment the defect features,which improves the defect that the fixed threshold method and the global threshold automatic selection method cannot segment the defect features of uneven illumination images.Afterwards,clear defect information is obtained through operations such as area filtering and dilation of connected areas.The algorithm has a high defect detection rate and real-time performance,and can also complete the detection when the shooting environment has uneven lighting.However,the generality of the algorithm is weak,and the algorithm can be practically applied for small projects with fewer types of fabrics after simple debugging.Thirdly,research the fabric defect detection algorithm based on statistics.Firstly,an algorithm for calculating the gray-level co-occurrence matrix in blocks is designed,which improves the shortcomings of the classical gray-level co-occurrence matrix,which cannot meet the needs of practical engineering due to its poor real-time performance.After that,an algorithm for cropping images in multiple positions is designed,which improves the defect that the feature information of the edge defect of the block is easy to lose after the block is divided.Then use histogram equalization,calculate eigenvalues,and analyze eigenvalues to obtain defect information.This algorithm has high versatility and real-time performance,but its detection efficiency is low.Finally,the defect detection algorithm of template matching which integrates statistical algorithm is studied.A sub-regional detection method is designed to detect the same position in different images to improve the detection efficiency.Then,a template design algorithm based on Gaussian distribution model is designed,which automatically obtains the feature threshold and eliminates the influence of accidental errors on template matching.In the detection process,the template is used to judge whether the target position contains defects.This algorithm retains the advantages of generality and realtime performance of statistical algorithms,and at the same time improves the problem that it requires too much lighting conditions,and has a high defect detection rate.