Printed Fabric Defects Detection Research Via Gabor Filter Construction Based On Genetic Algorithm And Characteristics Of Regular Band

The quality of printed fabrics can be affected by fabric defects, thus the value of printed fabrics can be reduced. Now defects detections of textile fabrics are mainly inspected by textile workers. Low detection efficiency and poor accuracy problems are existed in artificial detection. However, realization of automated defect detection can solve the existed problems of low detection efficiency and poor accuracy, which is an important research field of textile industrial sustainable development.Defect detection method of irregular patterned fabrics can be divided into two steps: the stage of training and testing phase. Stage of training is to use genetic algorithm for choosing the optimal parameters of Gabor, the objective function evaluated parameters in genetic algorithm. Selection, crossover and mutation are to construct a new parameter set. Finally, the direction and frequency of the optimal parameter are obtained via the rotation of the specific application in the testing stage. In the detection phase, the extreme value of the standard samples of printed fabrics via Gabor filter act as threshold value. Binarization based on threshold would be operated in printed fabrics. Execution of fabric detection in industrial production can be realized with high speed.Equalization pretreatment is performed at first stage of regular patterned fabrics. The size of fabric traversal window is equal to the fabric cycle in next step. Fabric cycle can be calculated by distance matching function. The two characteristic values of fabrics window are extracted for describing fabric characteristics. The two characteristic values are obtained via calculating the difference of normal fabrics and sample fabrics. Threshold of fabric defects discriminant is the characteristic value of non-defective fabrics. Speed of proposed method can satisfy the demand of detection system.In the light of the problem of fabric defects, a third type method of fabric defect detection automatically based on the eigenvalue of fabric weighted covariance is put forward. First of all, fabric grey value is set as the center of every window in fabric image, and the size of window in fabric image is set. The eigenvalues of weighted covariance matrix of fabric window is calculated as fabric window eigenvalues. Eigenvalues of weighted covariance in fabric defective window is less than the eigenvalues of non-defective fabric window, thus the characteristic value of weighted covariance of standard fabric acts as threshold of fabric defects discriminant. Fast detection of fabric defects can be realized accurately via the calculation of eigenvalue from weighted covariance matrix of fabric window.