Design And Implementation Of Fabric Defect Detection System Based On Vision

In each production process of fabric,quality monitoring is responsible of the final task of check,and fabric quality analysis results also play an important role in the discovery,positioning and solution of problems in fabric production.At present,the existing fabric quality inspection system on the market generally has problems such as incomplete function,poor adaptability,poor portability,and high cost,which can not meet the needs of most small and medium-sized fabric production enterprises.So a fabric defect detection system with low cost and strong adaptability is proposed based on the actual requirements of fabric quality detection.The system design mainly involves the sorting and analysis of the fabric quality detection requirements,the system process design and hardware selection,the design and implementation of the defect location algorithm and the defect classification algorithm,as well as the system construction,testing and improvement.In terms of system flow design,the system hardware adopts two CMOS cameras to form a binocular camera,and uses an industrial computer to control the camera and collect the image data of the fabric.Furthermore,the system uses the encoder to collect the state of the fabric,and selects the backlight supplementing scheme for supplementing the light.In view of the real-time demand of fabric quality inspection process,the system software designed the offline and online detection process.In the online detection process,the suspected fabric with defects can be quickly found out and retained to meet the requirements of detection speed.In the process of offline detection,the suspected defects in the images left by online detection are accurately located and classified to meet the requirements of detection accuracy.In the defect location algorithm,an improved region growing algorithm is proposed to improve the defect location and segmentation accuracy.The difference between the defect and the background is enhanced by background compensation,nonlinear transformation and other image enhancement methods.On the basis of the traditional region growth,the row and column gray histogram and BP neural network are used to obtain the growing seed points adaptively,so as to realize the automatic localization and segmentation of the defect region.A defect duplicate check algorithm is designed to solve the problem of repeated defects detected in the image.On the premise that the longitudinal position of the defect is similar,the algorithm extracted the transverse position and shape features of the defect,and weighted fusion was carried out after coding the features.By comparing the fused eigenvalues,the repeated defects were removed.In defect recognition algorithm,in order to solve the problem of high misidentification rate caused by high similarity of fabric defects,a cascade classifier is designed by combining gray texture features,shape features and SVM classifier to realize the classification and recognition of three kinds of detects,i.e.,common stains,folds and cloud spots.The data of fabric defects collected is used to establish a database,and the problem of unbalanced number of natural collected defects samples is solved by selecting representative defects and expanding defects.Through the laboratory test and the factory field test,it can be seen that this system can effectively detect common types of defects in fabric,and generate quality inspection report after statistics.In order to solve the problems of overheating and aging of the light box and over-sensitivity to special nonwoven fabric in the testing process,a solution of fan temperature control and background compensation is proposed.