Monitoring chenille yarn defects using image processing with control charts

Abstract

In this paper, a control chart is introduced for monitoring various defect types occurring on chenille yarns. To implement the control chart, a grey level image of chenille yarn is captured as an image matrix. Image preprocessing is applied and this involves thresholding to a binary image and a morphological opening operation for removing small objects from the image. The height of the pile yarn, measured from the processed images, is selected as the monitored quality characteristic. Since the monitored quality characteristic was highly autocorrelated, a first-order autoregressive AR(1) model was found to be appropriate for modelling the autocorrelation structure. Due to estimation of the AR(1) process parameters, a modified exponentially weighted moving average (EWMA) control chart for residuals is implemented as a tool for monitoring and detecting defects. It is shown that the modified EWMA control chart can be used successfully for monitoring different types of chenille yarn defects.