Woven fabric linear defect research based on AR model

As the detection effect of woven fabric linear defects is far from satisfaction,the spectral estimation method of AR model was used to research the problem.Firstly,split fabric image obtained into certain size according to child windows.And also from both the vertical and horizontal directions,turn gray value of the child windows into variance sequence by the way of using the variance projection.Secondly,choose the appropriate AR model order and use Burg algorithm to estimate spectral data.Finally,according to the correlation coefficient with the spectral estimation between a defect image and a normal one,the defect and its location will be detected.The effectiveness of the method of the linear defect detection on woven fabric is proved by experiment.