Research On Detection Method Of Fabric Defects Based On Gabor And Visual Information

The traditional fabric detection strategy relies on artificial detection,and the existing computer vision fabric detection methods are mainly based on learning the flawless fabric image feature,then we can process the fabric detection.These methods require a comparative study of the flawless fabric images,and have strict requirements for the experimental images which should collect at specific angles.The above strategies rely on the prior knowledge learning of the flawless fabric image,which can not be applied to different scale of fabric defect detection at the same time,and those strategies could lead to false detection.In this paper,we propose a fabric detection method based on Gabor and visual information for the above problems.Based on the phenomenon that the human eye can accurately identify the defective area from different scale and different texture of fabric images,the human eye vision is introduced into the fabric detection.In addition,this paper proposed an improved multi-channel Gabor fabric detection method and an improved visual saliency detection.The main contents of this paper are as follows:(1)Aiming at the problem that the defect detection method relies on the prior knowledge learning of the flawless fabric image,it can not detect the flaw of the fabric image simultaneously,which collected at different scale textile factories.To solve this problem,this paper proposes an improved multi-channel Gabor defect detection method.In this method,the improved multi-channel Gabor filter is used to filter the fabric image,and the multi-channel Gabor filtering result is selected by sub-block scoring method.The selected multi-channel is fused and segmented by threshold to obtain the defective region.The comparison experiments of the traditional method of fabric detection,the morphological method and the MRF method,shows that the proposed method can effectively improve the accuracy.(2)According to the problem of low dectection efficiency of the improved multi-channel Gabor defect detection method,a method of fabric defect detection with improved visual significance is proposed in this paper.The method described in this paper is based on the classical visual saliency model to obtain the bottom-up saliency feature of the fabric image,then proposed a top-down entropy and energysaliency feature calculation method.The method is used to fuse the saliency feature map,and the maximum interclass variance method is used to segment the saliency feature map,then we can obtain the visual area of the fabric image.The experimental results show that the method we proposed in this paper can improve the typical defects dectection efficiency and ensure the recognition rate.