Research On Fabric Defects Segmentation Based On FT Method And Frangi Filter

It is important to improve the accuracy of defects segmentation that has scientific significance to improving the accuracy of cloth quality evaluation.In this paper,we focus on improving the accuracy of defects segmentation.We take the enhancement of the defect area contrast as the main line,and improve the frequency-tuned salient(FT)algorithm and the multi-scale linear(Frangi)filtering method as the bridge.The influence of filter on the accuracy of defect segmentation is explored in the frequency-tuned salient algorithm.The relationship between the non local mean(NLM)filter and its structural parameters and the accuracy of defect segmentation is revealed.A marked defect segmentation method based on improved frequency-tuned salient algorithm is proposed.The bilateral filtering algorithm is applied to improve the defect segmentation method based on Frangi filter,and the way to enhance the defect contrast by using Frangi filter is explored.A fabric defect segmentation algorithm combining bilateral filtering and Frangi filtering is proposed.It lays a theoretical foundation for improving the accuracy rate of textile quality evaluation.In view of the problem that the smoothing ability of Gauss filter is weak in the FT algorithm,NLM filter is used instead of the Gauss filter in FT algorithm,and the influence of NLM filter parameter h on the accuracy of defect segmentation is explored.The mean maximum variance is used to study the optimization method of filter parameter h,and the optimization model of filter parameter h is established.An improved frequency-tuned salient algorithm is proposed,which is used for defect image preprocessing,improving the contrast of defects,and achieving the purpose of enhancing segmentation accuracy.The experimental results show that the improved frequency-tuned salient algorithm can significantly improve the accuracy of defect segmentation for image enhancement of coarse defects such as coarse warp,slub,knot,broken weft,oil stain and hole.The average filtering parameters based on the Otsu h optimization model is not suitable for side weft,sloughed-off weft and thin weft defects such as thin and problem research using image variance ratio of improved filter parameter optimization model of h,and explore the effects of h on the variance ratio of filter parameters,a filtering parameter optimization model of h,puts forward the optimization method of image filtering parameters based on the ratio of the average variance.A uniform threshold segmentation method for defect saliency map is proposed by using the optimization model of filter parameter h,and a fixed threshold segmentation method based on filter parameter model is proposed.The experimental results show that this method is not only suitable for common defects such as coarse warp,slub,knot,broken weft,oil stain and hole,but also for defects such as side weft,weft removal and dilute weft,which improves universality of marked defect segmentation method based on improved frequency-tuned salient algorithm.According to the Frangi filter to enhance the effect of fabric defects in regional contrast and background of extremely low case,the advantages of edge preserving denoising using bilateral filter to weaken the fabric image background texture information while maintaining the defect information,improve the defects and the contrast of the background region,provide technical support for further enhancement of defects by using Frangi filter.On this basis,a fabric defect segmentation algorithm based on bilateral filtering and Frangi filtering is proposed for defect image preprocessing,which significantly improves the defect area contrast.The experimental results show that the combination of bilateral filtering and Frangi filtering can enhance the accuracy of defect segmentation,such as lifting,double weft,slub,broken and broken ends.