Research And Implementation Of Gray Fabric Defect Detection System Based On Machine Vision

Gray cloth refers to the cloth that has not been dyed and printed after textile processing.It is the initial form of clothing.At this stage,many domestic textile manufacturers still use manual visual inspection to detect the defects of gray cloth,which will not only damage the workers' health,but also ensure the accuracy and speed of detection.Based on machine vision,the detection method is not only safe,but also fast and accurate.In view of this,this paper studies the related content of grey fabric defect detection system.The main work is as follows:(1)The preprocessing algorithm in image processing is studied.In the stage of brightness compensation,the commonly used brightness compensation algorithms are studied and a self-adaptive brightness compensation algorithm is proposed according to the particularity of defective images.The comparison of various algorithms proves that the algorithm can achieve brightness compensation of images very well and has certain advantages.In the phase of filtering and denoising,several commonly used filtering and denoising algorithms are studied and compared and analyzed in time and effect,and the most suitable filtering algorithm is selected.(2)The application of dictionary learning in gray fabric image defect detection is studied.According to the principle of dictionary learning and the characteristics of gray fabric defects,appropriate algorithms related to dictionary learning and image reconstruction are selected.In the optimization of dictionary size and sub-window size,the best parameters are selected through the analysis of experimental results and related data.The algorithm realizes the defect image detection accurately,which not only meets the accuracy requirements of the system,but also ensures real-time performance.(3)The classification of gray fabric defect image is studied.Firstly,the region of interest(ROI)of defective image is extracted,and the common feature selection methods of gray image are studied.The first moment feature,second moment feature and Gabor feature of ROI of defective image are selected and calculated.Support Vector Machine(SVM)is used to realize classification.One-to-one multi-classification strategy is selected and parameters are optimized to furtherimprove the accuracy and efficiency of classification.(4)The overall architecture design of gray fabric defect detection system is studied.The hardware includes the design of image acquisition module and synchronization control module,and the software includes the design of image processing module and human-computer interaction module.The performance test of gray fabric defect detection and defect classification is carried out for the detection system.The test results show that the system designed in this paper can run steadily,the accuracy of defect detection can reach 93.5%,and the accuracy of defect classification can reach 91.5%.It meets the design requirements and achieves the predetermined goals.