Automatic pattern separation of jacquard warp-knitted fabric by supervised multi-scale Markov model

Abstract

In this study, an automatic pattern separation approach using supervised multi-scale Markov model has been proposed. Gaussian low-pass filter has been used to smoothen the jacquard texture produced by various lapping movements, and the noise appearing during the capturing procedure is eliminated. Then the pyramidal multi-scale wavelet decomposition is adopted to lessen calculation burden and prepare specimens for subsequent pattern separation. In view of the non-stationary jacquard fabric image signals, the modified multi-scale MRF model is presented, which can fully capture and utilize correlations over sets of both inter-scale sub-images and intra-scale neighborhoods, and take the influence of weave structure and illumination condition into account. Finally, a supervised parameter estimation method is put forward to carry out pattern separation in Bayesian frame, in which the cost function changes with the decomposition scale, and parts of parameters are obtained by training in advance. Experimental results show that the proposed method is suitable for the pattern separation of jacquard warp-knitted fabric.